Masanari Asano
Masanari Asano
Quantum Adaptivity
in Biology: From
Genetics to Cognition
Quantum Adaptivity in Biology:
From Genetics to Cognition
Masanari Asano Andrei Khrennikov
•
Ichiro Yamato
Quantum Adaptivity
in Biology: From Genetics
to Cognition
123
Masanari Asano Yoshiharu Tanaka
Liberal Arts Division Information Science
Tokuyama College of Technology Tokyo University of Science
Tokuyama Tokyo
Japan Japan
Masanori Ohya
Information Science
Tokyo University of Science
Tokyo
Japan
Unlike any other discipline in the natural sciences, Biology has benefitted
tremendously from intriguing ideas and novel concepts from outside the subject’s
area throughout the last century. The apparently distinct topics of chemistry,
physics, mathematics and informatics became integral and indispensable matters of
biological research that blended surprisingly well with organismal studies of the last
centuries’ Biology. In this very aspect, Biology is reminiscent of the principal ideas
of ancient Philosophy, as both fields specialize in their quest for understanding the
essence of ‘life’, ‘meaning’ and ‘truth’.
As an inherent consequence of organismal plasticity and diversity, ‘truth’ in
biological findings is given support by studying the probability of a discrete or
gradual trait in a population, which uses stochastic expressions of classical math-
ematics. Here, again, striking similarities between biology and philosophy exist:
While modern deductive biology uses mathematics to describe processes as pre-
cisely as possible such as the dynamics of chemical reactions inside a cell or even
growth and development of organs, pure mathematical formalism dominated phi-
losophy and ruled deductive reasoning for almost two millennia.
Established by Aristotle’s logic, especially in his theory of syllogism about 300
BC, the meaning of words and thoughts can be expressed in common mathematical
formalism to deliver significance. This pure logic is bijective and unbiased from
feelings or subjective observations. Besides its influence on natural sciences,
Aristotle’s formal logic and the resulting Aristotelian philosophy had a great impact
on theology, especially Islamic and Christian religion. In addition, he introduced
deductive reasoning also into his own biological studies and, hence, Aristotle can
be considered as the founder of modern deductive biology. He was even ambitious
enough to adopt his theory of pure logic to the operational processes in the brain.
Ironically, the period of Aristotle’s logic formalism faced an end during the
epoch of Enlightenment that climaxed in Charles Darwin’s and Alfred R. Wallace’s
biological ideas of speciation and evolution!
During a period of almost 200 years, theories of modern mathematical logic and
Aristotelian logic were seemingly incompatible. In recent years, however, it is
stated that modern mathematical formalism and Aristotle’s theories of logic disclose
vii
viii Foreword
striking similarities. Moreover, biological conceptions merged with these ideas and
formed the current basis of modern biology that superficially splits into overlap-
ping, but yet distinct disciplines such as bioethics, biochemistry, biophysics,
biostatistics or bioinformatics.
Modern biologists still aim at finding scientific ‘truth’, at identifying how ‘life’
functions or how nerve cells compute ‘meaning’. Intriguingly, with high-through-
put methods at hand, biologists accumulate an enormous amount of data within a
short period of time that gives a renaissance to descriptive biology of pre-Aristo-
telian reasoning or of the epoch of Enlightenment. Methods of classical probability
are imposed to mine all kinds of observables for statistical significance, but a large
amount of information remains cryptic.
Especially, next generation sequencing technologies provide strings of genomic
DNA-sequences that encode for all the genetic make-up that determines an indi-
vidual or a species at extraordinary speed. As a consequence, an overwhelming
volume of genomic sequences is generated that await detailed analysis by bioin-
formatics. This large amount of information produces huge problems in data storage
and evaluation that might benefit from novel ground-breaking ideas or progress in
technology.
The present monograph by Masanari Asano, Andrei Khrennikov, Masanori
Ohya, Yoshiharu Tanaka and Ichiro Yamato pursues such novel ground-breaking
ideas in approaching phenomena of modern biology by quantum probabilistic
formalism. Unlike the description of quantum physical processes at molecular scale,
e.g. the processes of exciton transfer in photosynthesis or fluorescence resonance
energy transfer between different molecules, the authors use the operational
formalism of quantum theory to address biological problems at diverse macroscopic
biological scales. They discuss and describe the application of quantum-like
probability to various biological examples such as sequence and gene expression
analysis, bacterial growth or epigenetics studies as well as cognitive science.
The book highlights the similarities between mathematical formalism of quan-
tum probability in physics and statistical experimental data by quantum bio-infor-
matics. The authors carefully discuss current limitations of the general quantum
information theory and propose that an extended formalism might be required to
suit special biological problems. More importantly, however, they convincingly
explain that some biological observations violate classical mathematic formalism,
which might successfully be addressed by quantum-like probability or quantum
bioinformatics.
This monograph emphasizes the great potential of quantum-like probability in
life science, especially for the many dynamical processes in biological studies.
While classical stochastic formalism is rather static, quantum probability is more
advanced and allows the representation of dynamical biological processes as
different states in a framework of quantum fluctuations.
Finally, the authors provide a vital debate about how evolutionary aspects can
possibly be described by quantum-like models. Therefore, it is not surprising that
the authors address how quantum-like formalism and probability can support bio-
logical research in clarifying what ‘life’, ‘meaning’ and ‘truth’ is. Following
Foreword ix
Aristotelian logic, they concluded that quantum-like models hold the potential for
unifying Neo-Darwinism and Neo-Lamarckism theories in modern deductive
biology.
Although this book does not solve the problems of modern biological data
analysis, it constitutes an extraordinary attempt to show the applicability of quan-
tum theory and quantum-like formalism to macroscopic observables, which is novel
and a very stimulating read.
Dierk Wanke
Saarland University and University of Tübingen
Germany
Preface
xi
xii Preface
(even exotic and mystical) features. In particular, they violate the main laws of
classical Kolmogorovian probability. As was emphasized by R. Feynman (one
of the physical genies of the twentieth century), the quantum interference phe-
nomenon demonstrated in the two slit experiment (where a quantum particle
interferes with itself) can be probabilistically interpreted as violation of the for-
mula of total probability. The latter is one of the most fundamental laws of
classical probability theory, the heart of Bayesian analysis with corresponding
applications to the game theory and decision making. Thus quantum physics has
demonstrated that the classical probability model (as any mathematical model) has
a restricted domain of application. In particular, it can be used in classical sta-
tistical mechanics, but not in quantum mechanics.
The situation in probability theory is similar to the situation in geometry. During
two thousand years Euclidean geometry was considered as the only possible
mathematical model of physical space. (We recall that E. Kant even claimed that
this geometry was one of the basic elements of reality.) However, the discovery of
N. Lobachevsky showed that other consistent mathematical geometric models were
possible as well. Later, Riemann by inventing geometric models known nowadays
as Riemann manifolds opened the doors to a variety of geometric worlds. Finally,
the genial mind of A. Einstein coupled these mathematical models of geometry to
the physical reality by developing the theory of general relativity. (And Lob-
achesvky’s geometry has applications in special relativity).
Physics was one of the first scientific disciplines that were mathematically for-
malized. Plenty of mathematical theories, which were born in physical studies, have
later found applications in other domains of science. The best example is the
differential calculus originally developed by Newton for purely physical applica-
tions, but nowadays is applied everywhere, from biology to finances. A natural
question arises: Can quantum and, more generally, non-Kolmogorov probability be
applied anywhere besides quantum physics? In this book we shall demonstrate that
the answer is positive and that biology is a novel and extended field for such
applications. We noticed that biological phenomena, from molecular biology to
cognition, often violate classical total probability conservation law [26]. There is
plenty of corresponding experimental statistical data.1 Therefore, it is natural to
apply non-Kolmogorovian probability to biology (in the same way as non-
Euclidean geometric models are applied in physics). Since the quantum probabi-
listic model is the most elaborated among non-Kolmogorovian models, it is natural
to start with quantum probabilistic modelling of biological phenomena. However,
since biological phenomena have their own distinguishing features, one can expect
that the standard quantum formalism need not match completely with biological
applications. Novel generalizations of this formalism may be required. And this is
really the case, see Chap. 4.
1
A lot of data has been collected in cognitive science and psychology; unfortunately, in molecular
biology we have just a few experimental data collections which can be used for comparison of
classical and nonclassical probabilistic models. We hope that the present book will stimulate
corresponding experimental research in molecular biology.
Preface xiii
xv
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 1
1.1 Complexity of Information Processing
in Biological Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Towards the Operational Formalism in Biological Systems. . . . 3
1.3 Contextuality of Quantum Physics and Biology . . . . . . . . . . . 4
1.4 Adaptive Dynamical Systems . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Breaking the Formula of Total Probability
and Non-Kolmogorov Probability Theory. . . . . . . . . . . . . ... 7
1.6 Quantum Bio-informatics . . . . . . . . . . . . . . . . . . . . . . . . ... 9
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 10
xvii
xviii Contents
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Chapter 1
Introduction
where A is a measurable quantity and x1 , ..., x N are intrinsic variables. In such a sit-
uation one would give up attempts to connect the intrinsic variables with measurable
quantities and to realize this is the first step towards the usage of quantum formalism.
One of the new approaches to the above problems is based on the adaptive dynamics,
which will be discussed in Chap. 4.
1 Sometimes genome is compared with a computer’s program and epigenome with a (huge) set of
instructions on the usage of this software, see Chap. 8 for details.
2 We treat the process of decision making in a very general setup. For example, in our terminology
the E-coli bacteria “decide” whether to consume lactose or not in the presence or absence of glucose,
see Chap. 4.
1.2 Towards the Operational Formalism in Biological Systems 3
3Originally this formalism was developed for observables in microsystems. The main point of this
book is that the formalism is not rigidly coupled to microsystems. This is a very general formalism
describing observables. And it can be applied to observables of any kind.
4 1 Introduction
quantum formalism are not sufficient to model all possible biological phenomena.
And we shall show that this is really the case. In this book (Chap. 4) we present
a novel extension of the quantum formalism based on such an advanced tool of
quantum information theory as lifting.
self-measurements. For example, the brain need not get questions and problems
from the outside world to start preparing answers. It can ask itself and answer to
itself. Hence, in quantum bio-informatics we cannot avoid the consideration of self-
measurements. This is a real mismatch with the ideology of quantum physics (at
least with the operational interpretation). This foundational problem of quantum
bio-informatics has to be studied in more detail. In this book we proceed very
pragmatically. We stress the functional complexity of human brain. Therefore it
is natural to suppose that one functional unit can “ask questions and get answers”
from other functional units. (Of course, this viewpoint needs more justification from
neurophysiology.) Thus, opposite to quantum physics, in quantum bio-informatics
self-measurements are acceptable.
The aforementioned arguments work well for cognitive systems. For cellular bi-
ology, the idea of self-measurement implies a fundamental interpretational problem.
Opposite to biological organisms of higher level, the problem of cell cognition has
not been studied so much, cf. Karafyllidis [7]. It seems that a cell has some form
of cognition, it seems that it can “ask questions” to itself (e.g., about its own state
or the states of other cells) and “get answers”. However, it is not clear how far we
can proceed with the analogy between the cognition of, e.g., animals and “cogni-
tion” of cells. (In principle, we can define cell cognition as the ability to perform
aforementioned self-measurements.)
Finally, we discuss the contextuality in cellular biology. Cell behavior is evidently
contextual. For example, consider cell differentiation. This process depends crucially
on cellular context, especially in the form of signaling from other cells. The same
situation we have for genes expressions; the level of the expression of a special gene
cannot be considered outside the context of expressions of other genes.
Does the contextuality of bio-observables imply that bio-systems do not have ob-
jective properties at all? The answer is definitely negative! Bio-systems have objective
properties, but QL states (encoded via pure states or in general density operators)
do not specify these properties. Suppose that the levels of gene expressions in an
ensemble of cells are represented as the pure state ψ. This state does not describe the
real situation in a single cell. It only describes potential levels of genes expressions
in the ensemble of cells. It describes predictions for coming measurements. One
may say that each single cell “knows” its levels of genes expressions (for this cell
these are objective quantities). However, a geneticist does not know and ψ represents
uncertainty in the geneticist’s knowledge about genes’ expressions in this ensemble
of cells.
This is appropriate to note that on many occasions Niels Bohr emphasized that
quantum mechanics is not about physical processes in microworld, but about our
measurements [8]:
Strictly speaking, the mathematical formalism of quantum mechanics and electrodynamics
merely offers rules of calculation for the deduction of expectations pertaining to observa-
tions obtained under well-defined experimental conditions specified by classical physical
concepts.
6 1 Introduction
Bio-systems are fundamentally adaptive. They could not survive without developing
adaptive skills. They live in the permanently changing environment and the adaptivity
to new contexts is really the question of survival. Therefore it is natural to model, e.g.,
the process of evolution by adaptive dynamical systems, Chap. 8. In the same way
processing of information by the brain is described by the adaptive dynamics. The
brain permanently updates its dynamical system by taking into account variability of
mental contexts. Roughly speaking, by asking a question to Alice, we immediately
change the dynamical system operating in Alice’s brain. The question is processed
by a new dynamical system which takes into account a new context, the context of
this question. (See Chap. 6 for the corresponding mathematical model of decision
making in the games of the Prisoners Dilemma type.)
In physics the mathematical formalization of the adaptive dynamics (AD) has
implicitly appeared in series of papers [4, 9–17] for the study of compound quantum
dynamics, chaos, and the quantum realization of the algorithm on the satisfiability
problem (SAT algorithm). The name of the adaptive dynamics was deliberately used
in [17]. The AD has two aspects, one of which is the “observable-adaptive” and
another is the “state-adaptive”. We now present very general statements about these
two types of adaptivity. At the moment these statements can make the impression of
cryptograms. The precise contents of these cryptograms will become evident from
their mathematical representation in Chap. 4.
The observable-adaptive dynamics is a dynamics characterized as follows:
(1) Measurement depends on how to see an observable to be measured.
(2) The interaction between two systems depends on how a fixed observable exists,
that is, the interaction is related to some aspects of observables to be measured
or prepared.
The state-adaptive dynamics is a dynamics characterized as follows:
(1) Measurement depends on how the state and observable to be used exist.
(2) The correlation between two systems depends on how the state of at least one of
the systems exists, e.g., the interaction Hamiltonian depends on the state.
The idea of observable-adaptivity comes from studying chaos. We have claimed
that any observation will be unrelated or even contradictory to mathematical uni-
versalities such as taking limits, sup, inf, etc. The observation of chaos is a result
due to taking suitable scales of, for example, time, distance or domain, and it will
not be possible in limiting cases. Examples of the observable-adaptivity are used to
understand chaos [10, 15] and examine the violation of Bell’s inequality, namely, the
chameleon dynamics of Accardi [18]. The idea of the state-adaptivity is implicitly
1.4 Adaptive Dynamical Systems 7
used in the construction of a compound state for quantum communication [9, 11,
19, 20]. Examples of the state-adaptivity can be seen in an algorithm solving NP
complete problem, i.e., a pending problem for more than 30 years asking whether
there exists an algorithm solving a NP complete problem in polynomial time, as
discussed [4, 13, 16].
The mathematical details of the adaptive dynamics will be discussed in Chap. 4.
6 The Kolmogorov probability model based on the representation of events by sets and probabilities
by measures will be presented in Chap. 2. For the moment, we operate with probabilities at the
formal level. We remark that in the rigorous framework FTP is a theorem.
8 1 Introduction
7 The basic quantum experiment demonstrating the violation of FTP is the two slit experiment
demonstrating interference of probabilistic patterns related to two incompatible experimental con-
texts: one of slits is open and the other is closed, see Sect. 4.1.1. In general, violation of FTP means
the presence of nontrivial interference of probabilistic patters.
1.5 Breaking the Formula of Total Probability … 9
References
1. Haven, E., Khrennikov, A.: Quantum Social Science. Cambridge Press, Cambridge (2012)
2. Dirac, P.A.M.: The Principles of Quantum Mechanics. Clarendon Press, Oxford (1995)
3. Von Neuman, J.: Mathematical Foundations of Quantum Mechanics. Princeton University
Press, Princeton (1955)
4. Ohya, M., Volovich, I.V.: Mathematical Foundations of Quantum Information and Computation
and Its Applications to Nano- and Bio-systems. Springer, Berlin (2011)
5. Khrennikova, P., Khrennikov, A., Haven, E.: The quantum-like description of the dynamics of
party governance in the US political system (2012). arXiv:1206.2888 [physics.gen-ph]
6. Khrennikova, P.: Evolution of quantum-like modeling in decision making processes. AIP Conf.
Proc. 1508, 108–112 (2012)
7. Karafyllidis, I.: Quantum mechanical model for information transfer from DNA to protein.
Biosystems 93, 191–198 (2008)
8. Bohr, N.: The Philosophical Writings of Niels Bohr, vol. 3. Ox Bow Press, Woodbridge (1987)
9. Ohya, M.: On compound state and mutual information in quantum information theory. IEEE
Trans. Inf. Theory 29, 770–777 (1983)
10. Ohya, M.: Complexities and their applications to characterization of chaos. Int. J. Theor. Phys.
37, 495–505 (1998)
11. Accardi, L., Ohya, M.: Compound channels, transition expectations, and liftings. Appl. Math.
Optim. 39, 33–59 (1999)
12. Inoue, K., Ohya, M., Sato, K.: Application of chaos degree to some dynamical systems. Chaos,
Soliton and Fractals 11, 1377–1385 (2000)
13. Ohya, M., Volovich, I.V.: New quantum algorithm for studying NP-complete problems. Rep.
Math. Phys. 52, 25–33 (2003)
14. Inoue, K., Ohya, M., Volovich, I.V.: Semiclassical properties and chaos degree for the quantum
baker’s map. J. Math. Phys. 43 (2002)
15. Kossakowski, A., Ohya, M., Togawa, Y.: How can we observe and describe chaos? Open Syst.
Inf. Dyn. 10(3), 221–233 (2003)
16. Accardi, L., Ohya, M.: A Stochastic limit approach to the SAT problem. Open Syst. Inf. Dyn.
11, 1–16 (2004)
17. Ohya, M.: Adaptive dynamics and its applications to chaos and NPC problem. QP-PQ: quantum
probability and white noise analysis. Quantum Bio-Inform. 21 181–216 (2007)
18. Accardi, L.: Urne e camaleonti: Dialogo sulla realta, le leggi del caso e la teoria quantistica.Il
Saggiatore, Rome (1997); English edn. World Scientific, Singapore (2002); Japanese edn,
Makino, Tokyo (2002); Russian edn. Regular and Chaotic dynamics, Ijevsk (2002)
19. Ohya, M.: Note on quantum probability. Il Nuovo Cimento 38, 203–206 (1983)
20. Ohya, M.: Some aspects of quantum information theory and their applications to irreversible
processes. Rep. Math. Phys. 27, 19–47 (1989)
21. Khrennikov, A.: Contextual Approach to Quantum Formalism. Springer, Berlin (2009)
22. Conte, E., Todarello, O., Federici, A., Vitiello, F., Lopane, M.: A preliminary evidence of
quantum-like behaviour in measurements of mental states. Quantum Theory: Reconsideration
of Foundations. Series Mathematical Modeling Physics, Engineering, and Cognitive Science
10, 679–702 (2004)
23. Conte, E., Khrennikov, A., Todarello, O., Federici, A., Zbilut, J.P.: Mental states follow quantum
mechanics during perception and cognition of ambiguous figures. Open Syst. Inf. Dyn. 16, 1–17
(2009)
24. Shafir, E., Tversky, A.: Thinking through uncertainty: nonconsequential reasoning and choice.
Cogn. Psychol. 24, 449–474 (1992)
25. Tversky, A., Shafir, E.: The disjunction effect in choice under uncertainty. Psychol. Sci. 3,
305–309 (1992)
26. Hofstader, D.R.: Dilemmas for superrational thinkers, leading up to a luring lottery. Sci. Am.
6 (1983)
References 11
27. Hofstader, D.R.: Metamagical Themes: Questing for the Essence of Mind and Pattern. Basic
Books, New York (1985)
28. Croson, R.: The disjunction effect and reasoning-based choice in games. Organ. Behav. Hum.
Decis. Process. 80, 118–133 (1999)
29. Busemeyer, J.B., Wang, Z., Townsend, J.T.: Quantum dynamics of human decision making. J.
Math. Psychol. 50, 220–241 (2006)
30. Busemeyer, J.R., Matthews, M., Wang, Z.: A quantum information processing explanation of
disjunction effects. In: Sun, R., Myake, N. (eds.) The 29th Annual Conference of the Cognitive
Science Society and the 5th International Conference of Cognitive Science, pp. 131–135.
Mahwah (2006)
31. Busemeyer, J.R., Santuy, E., Lambert-Mogiliansky, A.: Comparison of Markov and quantum
models of decision making. In: Bruza, P., Lawless, W., van Rijsbergen, K., Sofge, D.A., Coeke,
B., Clark, S. (eds.) Quantum Interaction: Proceedings of the Second Quantum Interaction
Symposium, pp. 68–74. College Publications, London (2008)
32. Busemeyer, J., Bruza, P.D.: Quantum Cognition and Decision. Cambridge University Press,
Cambridge (2012)
33. Pothos, E.M., Busemeyer, J.R.: A quantum probability explanation for violation of rational
decision theory. Proc. R. Soc. 276(1165), 2171–2178 (2009)
34. Wang, Z., Busemeyer, J.R.: A quantum question order model supported by empirical tests of
an a priori and precise prediction. Top. Cogn. Sci. (2013), to be published
35. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Tanaka, Y., Yamato, I.: Quantum-like model
for the adaptive dynamics of the genetic regulation of E. coli’s metabolism of glucose/lactose.
Syst. Synth. Biol. 6, 1–7 (2012)
36. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Tanaka, Y., Yamato, I.: Quantum-like model
of diauxie in Escherichia coli: operational description of precultivation effect. J. Theor. Biol.
314, 130–137 (2012)
37. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Yamato, I.: Non-Kolmogorovian approach
to the context-dependent systems breaking the classical probability law. Found. Phys. 43(7),
895–911 (2013)
38. Hameroff, S.: Quantum coherence in microtubules. A neural basis for emergent consciousness?
J. Conscious. Stud. 1, 91–118 (1994)
39. Hameroff, S.: Quantum computing in brain microtubules? The Penrose-Hameroff Orch Or
model of consciousness. Philos. Trans. R. Soc. Lond. A 1–28 (1994)
40. Fichtner, K.-H., Fichtner, L., Freudenberg, W., Ohya, M.: On a quantum model of the recog-
nition process. QP-PQ: Quantum Prob. White Noise Anal. 21, 64–84 (2008)
41. Donald, M. J.: A Review of Quantum Evolution (2001). arXiv:quant-ph/0101019
42. Khrennikov, A.: Quantum-like brain: interference of minds. BioSystem 84, 225–241 (2006)
43. Khrennikov, A.: The quantum-like brain on the cognitive and subcognitive time scales. J.
Conscious. Stud. 15, 39–77 (2008)
44. Khrennikov, A.: On quantum-like probabilistic structure of mental information. Open Syst.
Inf. Dyn. 11, 267–275 (2004)
45. Khrennikov, A.: Ubiquitous Quantum Structure: From Psychology to Finance. Springer,
Heidelberg (2010)
46. Khrennikov, A.: On the physical basis of theory of “mental waves”. Neuroquantology 8, S71–
S80 (2010)
47. Khrennikov, A.: Quantum-like model of processing of information in the brain based on clas-
sical electromagnetic field. Biosystem 105(3), 250–262 (2011)
Chapter 2
Fundamentals of Classical Probability
and Quantum Probability Theory
Abstract In this chapter we present briefly the basic notions of classical and
quantum theories of probability and information. This chapter is especially impor-
tant for biologists, psychologists, experts in cognition, and sociologists who were not
trained in quantum theory (but even classical theory is presented in a simple manner).
We start with the presentation of the standard measure-theoretic formulation of the
modern classical probability theory (Kolmogorov, Grundbegriffe der Wahrschein-
lichkeitsrechnung. Springer, Berlin [1]). Then we turn to fundamentals of quantum
formalism, including theory of open quantum systems and its generalizations.
A → P(A). (2.1)
The probabilistic weights are chosen to be nonnegative real numbers and normalized
by 1: P(Ω) = 1, the probability that something happens equals one. An event with
large weight is more probable than an event with small weight. The weight of an
event A that can be represented as the disjoint union of events A1 and A2 is equal to
the sum of weights of these events. The latter property is called additivity. (There is
the evident similarity with mass, area, volume.)
It is useful to impose some restrictions on the system of sets representing events:
(a) set Ω containing all possible events and the empty set ∅ are events (something
happens and nothing happens); (b) the union of two sets representing events rep-
resents an event; (c) the intersection of two sets representing events represents an
event; (d) the complement of a set representing an event, i.e., the collection of all
points that do not belong to this set, again represents an event. These set-theoretic
operations correspond to the basic operations of (Boolean) logic: “or”, “and”, “no”.
And the modern set-theoretic representation of events is a mapping of propositions
describing events onto sets with preservation of the logical structure. At the beginning
of the mathematical formalization of probability theory the map (2.1) was defined
on an algebraic structure corresponding to the logical structure, the Boolean alge-
bra (invented by Boole, the creator of “Boolean logic” [2]). The set-system with
properties (a)–(d) is called the algebra of sets (in the American literature, the field
of sets).
In the case of finite Ω the map given by (2.1) with the above-mentioned properties
is (measure-theoretic) probability. (Since Ω can contain billions of points, this model
is useful in a huge class of applications.) Here Ω = {ω1 , . . . , ω N }. To determine any
map (2.1), it is enough to assign to each elementary event its weight
0 ≤ P(ω j ) ≤ 1, P(ω j ) = 1.
j
Then by additivity this map is extended to the set-algebra consisting of all subsets
of Ω:
P(A) = P(ω j ).
{ω j ∈A}
and to work fruitfully with such maps (e.g., to integrate), one has to impose special
restrictions on the system of sets representing events. It has to be not simply a set-
algebra, but a σ -algebra of sets (in the American literature, a σ -field), i.e., (b) and
(c) must be valid for countable unions and intersections of sets. In logical terms, it
means that the operations “or” and “and” can be applied infinitely many times to
form new events. Of course, this is a mathematical idealization of the real situation.
One of the most important “continuous probability model” is based on Ω = R, i.e.,
2.1 Short Introduction to Classical Probability Theory 15
P = (Ω, F , P).
belong to F .
It is typically assumed that the range of values X a is a subset of the real line.
We will proceed under this assumption practically everywhere, but sometimes, e.g.,
in cognitive and psychological modeling, it will be more convenient to consider
Boolean labels, e.g., α = yes, no.
The probability distribution of a (discrete) random variable a is defined as
We remark that
If the set of values of ξ is infinite, then the average is well defined if the series in the
right-hand side of (2.5) converges absolutely.
16 2 Fundamentals of Classical Probability and Quantum Probability Theory
pa1 ··· am (α 1j1 , . . . , α mjm ) = P(ω ∈ Ω : a1 (ω) = α 1j1 , . . . , am (ω) = α mjm ). (2.6)
We remark that the joint probability is symmetric with respect to permutations; e.g.,
for two random variables a and b, we have
1 We remark that, for a discrete random variable, the integral coincides with the sum for the
mathematical expectation, see (2.5). And a discrete random variable is integrable if its mathematical
expectation is well defined. In general any integrable random variable can be approximated by
integrable discrete random variables and its integral is defined as the limit of the integrals for the
approximating sequence.
2.1 Short Introduction to Classical Probability Theory 17
This is a probability measure on the Borel σ -algebra. And the calculation of the
average can be reduced to integration with respect to pa :
ā ≡ Ea = x dpa (x). (2.12)
R
For simplicity, further we shall proceed only with discrete random variables. The
general case can be treated by using everywhere integrals instead of sums.
We naturally have
where, e.g., Cαa is defined by (2.3). It is, of course, assumed that in the first case
pa (α) > 0 and in the second case p b (β) > 0.
It is useful to consider the matrices of transition probabilities
We remark that these matrices are always left stochastic. The left stochastic matrix
is a square matrix whose columns consist of nonnegative real numbers whose sum
is 1. For example, for the matrix Pb|a , we have
pαβ = P(b = β|a = α) = 1. (2.18)
β β
for any fixed a = α. This is a consequence of the fact that, for any set C of strictly
positive probability, PC is also a probability measure.
We point to the following equality connecting the joint probability distribution of
two random variables a and b with their transition probabilities
In our further considerations the important role will be played by the formula of total
probability (FTP). This is a theorem of the Kolmogorov model. Let us consider a
countable family of disjoint sets Ak belonging to F such that their union is equal to
Ω and P(Ak ) > 0, k = 1, . . .. Such a family is called a partition of the space Ω.
Theorem 2.1 Let {Ak } be a partition. Then, for every set B ∈ F , the following
formula of total probability holds
Proof We have
∞ ∞
P(B ∩ Ak )
P(B) = P B ∩ ∪∞
k=1 A k = P(B ∩ A k ) = P(Ak ) .
P(Ak )
k=1 k=1
or in compact notation,
k1 ψ1 + k2 ψ2 , φ = k1 ψ1 , φ + k2 ψ2 , φ ,
vectors of H , i.e., ψ such that ψ = 1, represent a special (and the most important)
class of states of quantum systems, pure states. Each pure state ψ can be represented
as an operator acting in H . which is the orthogonal projections π onto the vector
ψ. In terms of the scalar product, the orthogonal projections can be written as π φ =
φ, ψ ψ. By fixing an orthonormal basis in H , the pure state is expressed by a
matrix ρ = (ρij ) satisfying
(a) Hermitian: ρij = ρ̄ij , in particular, the diagonal elements are real,
(b) Positive definiteness: ρφ, φ ≥ 0 for any vector φ,
(c) Its trace equals 1: trρ = j ρjj = 1.
The eigenvalues of Hermitian matrix are real numbers. In addition, a Hermitian
matrix with the positive definiteness has non-negative eigenvalues. Finally, from the
condition c), the sum of all the eigenvalues equals 1.
As an example in the two dimensional space H = C2 , we introduce the operator
given by the following matrix
Here α and β are complex numbers satisfying |α|2 + |β|2 = 1. The above ρqubit
expresses the state of quantum bit (shortly called qubit), which is often seen in
quantum information theory (see Sect. 2.2.4). One can easily see that ρqubit satisfies
the following condition:
ρqubit
2
= ρqubit . (2.27)
for the pure state, this postulate was proposed by Born, see (2.32); the form (2.28)
is due to von Neumann [3]).
Thus all information about possible results of measurements is encoded in two
Hermitian operators, the observable A and the state ρ. This is very compact and
convenient representation.
In fact, mathematically, quantum formalism is a linear algebra (with elements of
functional analysis for the infinite-dimensional case). Thus it is very simple; there
is nothing simpler than a linear representation. The corresponding the dynamical
equations (e.g., the Schrödinger’s equation) are linear.
The simplicity of the quantum linear representation of measurable quantities is
one of the reasons to use this formalism, in particular, in biology.
Surprisingly, practically any theory of statistical measurements can be opera-
tionally represented in a linear space. Here we have no possibility to discuss this
problem that was studied in detail already in 1950th by Mackey; recently one of
the authors of this book has proposed a quantum-like representation algorithm [4].
This algorithm constructs the QL representation of statistical data collected for two
observables (random variables) under some restrictions on the matrix of transition
probabilities for these observables. Hence, all “natural operational representations”
are reduced to linear representations (in some situations one has to use number sys-
tems different from complex numbers, e.g., the so called hyperbolic numbers [4]).
2.2.2 Superposition
Let the state space of some system (physical or biological) be represented as a finite-
dimensional Hilbert space H . Consider a pure state ψ and an observable A, denote
its eigenvalues by a1 , . . . , am and the corresponding eigenvectors by e1 , . . . , em .
This is an orthonormal basis in H . (We again proceed under the assumption that all
eigenvalues are different.) We expand the vector ψ with respect to this basis:
ψ = c1 e1 + · · · + cm em , (2.29)
where (c j ) are complex numbers such that the sum of their squared absolute values
are equal to one (this is the coordinate expression of the normalization by one of a
pure state vector):
By using the terminology of linear algebra we say that the pure state ψ is a
superposition of pure states e j .
The density matrix corresponding to ψ has the elements
ρij = ci c̄ j . (2.31)
2.2 Short Introduction of Quantum Probability Theory 23
Hence, for the pure state ψ, the basic probabilistic postulate of quantum mechanics,
(2.28), has the form
P(a j ) = ρjj = c j c̄ j = |c j |2 . (2.32)
This postulate can be written without using the coordinates of the state vector ψ
with respect to the basis of eigenvectors of a quantum observable. We remark that,
since the basis of eigenvectors of a Hermitian operator can always be selected as
orthonormal, the coordinates c j can be expressed in the form: c j = ψ, e j . Hence,
the Born’s rule takes the form:
P(a j ) = | ψ, e j |2 . (2.33)
Degenerate Eigenvalues
Consider now the general case: the eigenvalues can be degenerate and eigensub-
spaces5 need not be one dimensional. Suppose that by measuring an observable
represented by the Hermitian operator A its eigenvalue a was obtained; denote the
projector to this eigensubspace by πa . Then by the projection postulate the input
pure state ψ is transformed into (again pure) state
πa ψ
ψout;a = . (2.34)
πa ψ
We remark that although this is the standard definition used in modern quantum
theory, von Neumann distinguished sharply the cases of non-degenerate spectrum
(i.e., in the finite dimensional case, all eigenvalues are different) and degenerate
spectrum. In the latter case the definition (2.34) is due to Lüders. Originally von
Neumann assumed that a measurement can transfer a pure state into a mixed state
given by a density operator, see [3]. However, we would not disturb biologists by
such foundational problems. Hence, we proceed with the Lüders form of the projec-
tion postulate even in the case of degenerate spectrum (as the simplest form of the
projection postulate).6 If the input state is given by the density operator ρ, then
πa ρπa
ρout;a = . (2.35)
trπa ρπa
Suppose now that by measuring A we are interested in the state (mixed) representing
the ensemble of states of all systems after measurement, so this ensemble is the
mixture of the ensembles corresponding to the results A = a j for different a j . Then
this state is mathematically represented by the density operator
ρout = πa ρπa . (2.36)
a
Self-measurements
We stress that in quantum bio-informatics we interpret the notion of measurement in
a more general way than in quantum physics. In particular, measurements can be self-
measurements, see Sect. 1.3; we also treat decoherence as a form of measurement.
Dirac was one of the inventors of the quantum formalism [5]. He introduced his own
symbolic notation for the Hilbert space linear algebra which was not used in math-
ematics. This notation became very common in quantum physics and especially in
quantum information theory. We shall now present shortly Dirac’s symbolic notation.
In this books we shall use it very often (but not always).
Consider an observable A, denote its eigenvalues by a1 , . . . , am and the corre-
sponding eigenvectors by e1 , . . . , em . Suppose again that all eigenvalues are different.
Then by Dirac e j is denoted as |a j . Even an arbitrary pure state ψ is often written
as |ψ (just for convenience). In Dirac’s notation superposition (2.29) is written as
Consider some pure state |ψ , then the symbol |ψ ψ| denotes the operator of
orthogonal projection onto the vector |ψ . Thus the operator A with the system of
distinct eigenvalues a1 , . . . , am can be written as
m
A= ai |ai ai |,
i=1
7 Decoherence is a complicated interpretational issue of quantum mechanics. Some (but not all)
researchers treat decoherence as a form of measurement.
26 2 Fundamentals of Classical Probability and Quantum Probability Theory
or simply A = a a|a a|. We shall come back to the problem of encoding of
quantum information in Sect. 2.5.3, after the introduction of state spaces of compound
systems (given by tensor products of Hilbert state spaces of subsystems).
2.2.4 Qubit
The basic notion of quantum information theory is quantum bit (qubit), the quantum
analog of classical bit of information. Consider any dichotomous observable A =
a1 , a2 with a1 < a2 , in particular, it acts in the two dimensional Hilbert space.
Let us encode 0, 1 by its eigenvectors. We can always calibrate the pointer of the
corresponding measurement device in such a way that a1 = 0 and a2 = 1, the
encoding rule has the form 0 → |0 , 1 → |1 . The crucial point is that, besides the
states |α , α = 0, 1, a quantum system can be in superposition of these states,
Thus a single quantum system can carry with its state not just either 0 or 1, i.e., one
bit of information, but both 0 and 1, with some probabilistic weights. However, by
measurement of A it is possible to extract just one bit of information by getting either
the result A = 0 or A = 1. The crucial point is that (2.40) represents not simply the
classical probability distribution with two weights p0 = |c0 |2 , p1 = |c1 |2 . Qubit can
generate interference effects, but the classical probability distribution not. We shall
discuss this problem in more detail later.
Up to now we have restricted the presentation to the finite dimensional case. In real
physics the Hilbert space of quantum states is infinite dimensional, the basic example
is given by the space of square integrable complex valued functions, x → ψ(x),
where x is the spatial variable, i.e., functions such that |ψ(x)|2 d x < ∞. We denote
this space by the symbol L 2 (Rn ), where x ∈ Rn . A pure state ψ, i.e., a function
such that
|ψ(x)|2 d x = 1, (2.41)
ex0 (x) = δ(x − x0 ). Thus eigenfunctions ex0 (x) are “enumerated” with the aid of the
continuous index x0 . And formally any square integrable function can be represented
as the superposition of eigenfunctions of the position operator:
ψ(x) = ψ(x0 )ex0 (x)d x0 . (2.42)
“the probability to find a quantum system with the wave function ψ at the point x is
equal to the squared absolute value of this function at this point.” This is the original
formulation of the basic probabilistic postulate of quantum mechanics, Born’s rule.
This is a postulate; it cannot be derived in the conventional quantum theory; however,
cf. [6].
In the real quantum physics state spaces are always of the L 2 -type, i.e., they are
infinite-dimensional. However, in quantum information one typically selects just a
few degrees of freedom, which are then used for the representation of information
and proceed in finite dimensional state space. In this book we shall apply the same
strategy. However, physical state space and wave function will be used in Sect. 4.1.1,
which is devoted to the presentation of interference of photons.
We can assume that, like in a quantum physical system, the evolution of the QL-
state of a bio-system isolated from the environment is described by the Schrödinger
equation (with a minor modification related to the usage of the Planck constant in
physics):
dψ
iγ = H ψ(t), ψ(0) = ψ 0 . (2.44)
dt
Instead of the Planck constant h, we use a constant γ , which has the dimension of
time. We recall that the Planck constant has the dimension of action, i.e., energy ×
time. Here H is a Hermitian operator acting in the space of QL-states of bio-systems.
In quantum physics H is the energy operator; it is called Hamiltonian. The energy
dimension of this operator matches with the action dimension of the Planck constant.
In quantum bio-informatics this operator does not represent the physical energy of
a bio-system. (We state again that quantum bio-informatics is not at all quantum
biology.) It is the generator of purely information dynamics. It is convenient to have
28 2 Fundamentals of Classical Probability and Quantum Probability Theory
it dimensionless and this motivates the usage of the time scale constant γ , instead of
the action constant h. In coming QL-models γ gives the time scale of information
processing in biological systems. We do not claim that there exists some universal
constant scaling dynamics in all bio-systems. Each population of bio-systems can
have its is own temporal scale.
We shall also call the generator of the Schrödinger dynamics Hamiltonian (the
operator H ) and we hope that it will not cause misunderstanding.
The Schrödinger equation is, in fact, a system of linear differential equations with
complex coefficients; in the one dimensional case H is just a real number and the
−itH
general solution has the form of the imaginary exponent: ψ(t) = e γ ψ 0 . In the
general case H is an operator and the solution is represented in the form of imaginary
operator-exponent (for the fixed basis it is simply the exponent of the matrix):
−itH
ψ(t) = Ut ψ 0 , Ut = e γ . (2.45)
Thus this dynamics transfers a pure quantum state to another pure quantum state.
In quantum physics, the Schrödinger equation plays a crucial role. It seems not to
be the case in biology. Here it is very difficult and practically impossible to provide
the complete isolation from environment. Take a single cell and isolate it from other
cells. If we still provide the supply of chemical staffs, this cell will survive, but
its behavior in the absence of signaling from other cells will be very artificial. For
example, such an important process as cell differentiation will be impossible. If we
try to isolate a cell from the supply of all chemical staffs, it will die. The same can
be said about the isolated (mentally or/and physically) brain.
Since any general quantum state, a density operator, can be represented as a
mixture of density operators corresponding to pure states, the Schrödinger dynamics
for pure states implies the following dynamics for density operators:
dρ
γ (t) = −i[H, ρ(t)], ρ(0) = ρ 0 . (2.47)
dt
(In quantum physics the Planck constant h is used instead of the time scaling factor γ ).
This equation is known as the Von Neumann equation [3]. By using representation
(2.45) of the Schrödinger evolution for the pure state we represent the evolution of
the density operator (“mixed state”) in the form
2.3 Schrödinger Dynamics and Its Role in Quantum Bio-informatics 29
where, for an operator W, the symbol W ∗ denotes its adjoint operator. The latter is
defined by the equality
W ψ1 |ψ2 = ψ1 |W ∗ ψ2 ; (2.49)
by denoting the matrix elements of these operators as wij and wij∗ we have wij∗ = w̄ ji .
What kind of the mathematical apparatus can be used to model adaptive biological
phenomena operationally? An important class of adaptive dynamical systems can be
described by the apparatus of the theory of open quantum systems, see, e.g., Ingarden
et al. [7], Ohya and Volovich [8]. In complete accordance with the operational view-
point to the formalism of quantum physics, we can apply the theory of open quantum
systems in biology, from genetics and cellular biology to the brain science, cognitive
science and psychology. By the theory of open quantum systems, the dynamics of the
QL-state of a bio-system interacting with an environment is approximately described
by quantum master equation, the Gorini-Kossakowski-Sudarshan-Lindblad (GKSL)
equation, see, e.g., [7].8
In cellular QL-models at the beginning of interaction with an environment the
QL-state of a population of cells is characterized by a high degree of uncertainty
about possible changes (cf. Sect. 2.2.2), which can be generated via the coupling
with an environment. The quantum master equation describes the process of res-
olution of this state of uncertainty and approaching the complete matching with
the environment, Sect. 2.4. This process can be considered as decoherence of the
QL-state through interaction with an environment. As a result, the state loses its
fundamentally quantum(-like) feature, superposition of a few alternatives, and the
final situation can be described by classical probability theory. Mathematically, this
is the process of approaching a steady state solution of the quantum master equation.
The first application of the theory of open quantum systems to biology was in
the domain of cognitive psychology and game theory, see [9, 10]. The classical
probability distribution corresponding to the steady state was used to model decision
making by cognitive systems; in particular, by gamblers in games of the Prisoner’s
Dilemma type; cf. with QL-modeling performed on the basis of Schrödinger equation
[11, 12].
One of the novel proposals realized in this book is to use the machinery of the
theory of open quantum systems and to describe dynamics of the QL-state of a pop-
ulation of bio-systems by using the quantum master equation (the GKSL-equation).
We state again that this equation can be used to describe transitions from states
dρ
γ = −i[H, ρ(t)] + W ρ(t), ρ(0) = ρ 0 , (2.50)
dt
where H is a Hermitian operator determining the internal dynamics of the QL-state
of a population of bio-systems isolated from the environmental pressure (“cell’s
Hamiltonian”) and the linear operator W describes the environmental pressure.
Opposite to H (which is simply a Hermitian operator; for a fixed basis it is repre-
sented by a Hermitian matrix), in general the operator W has a complex mathematical
structure. It has such a form that starting with a density operator ρ ◦ we shall get den-
sity operators at all instances of time. For a moment, the specific structure of W is
not important for us; see, e.g., [7, 8], and Sect. 8.2.3 for mathematical details. Bio-
logically this operator is determined by the properties of the environment, including
the initial state of the environment.
Here γ is the time scale constant, it determines the temporal scale of the biological
dynamics. By using such a scaling factor of the dimension of time, we are able to
proceed with dimensionless Hamiltonian H and the environmental operator W .
We state once again that for our QL-modeling it is crucial that, for a very general
class of GKSL-equations, the environmental operator W drives (in the limit t →
∞) the QL-state of a biological population, ρ(t), to the steady solution: ρ(t) →
ρst . Typically the uncertainty (in the form of superposition) is eliminated from the
asymptotic state ρst .
In our QL-model such a steady state is considered as the result of the biological
dynamics in the environment (mathematically represented by the operator W ). For
example, we can consider epigenetic evolution (Chap. 8). Here the limiting proba-
bility distribution ρepi;st describes the probability distribution of epimutations which
took place in a cell population as a consequence of interaction with the environment.
9 We treat the notion of decision very generally: from decisions made by people to “cell’s decisions”,
Internal uncertainty, to epimutate or not epimutate, was resolved and a stable phe-
notype was created.
Finally, we remark that under natural restrictions a selection operator produces
the same steady state for all possible initial states, Sect. 8.3.1. Such an open system
dynamics simulates, for example, the spreading of mutations or epimutations in a
population. The pressure of the environment can be so strong that the same genotype
or phenotype is created independently of the initial states of populations. Such a
dynamics can also have applications in social science and political technologies (See,
e.g., [13, 14] for modeling of dynamics of party governance in the USA political
system. Here a QL-model of decision-making of American voters was elaborated.
In the framework of open quantum systems a possibility of driving populations of
voters to the fixed state, e.g., to vote for republicans independently of their initial
preferences was modeled.)
We shall continue our discussion on the theory of open quantum systems and the
theory of adaptive quantum dynamics in Sect. 2.6 after the introduction to the theory
of compound quantum systems.
10 In the coordinate form tensor products of vectors and matrices are also known under the name
Kronecker product. This structure is widely used in various computational algorithms including
computational biology.
32 2 Fundamentals of Classical Probability and Quantum Probability Theory
where
|cij |2 < ∞. (2.53)
Those who work with electromagnetic signals in biology, e.g., in the brain
research, have experience in expanding electromagnetic signals with respect to vari-
ous bases, e.g., using the Fourier expansion or the wavelet expansion. Some bases are
indexed by continuous parameters; integrals take place of sums. Thus the notion of
basis in the L 2 -space is widely known. However, there is a crucial difference between
the classical field and quantum mechanical representations of compound systems.
The state of a classical bi-signal consisting of two components is represented in
the Cartesian product of the corresponding L 2 -spaces. And the state of a quantum
bi-system, e.g., bi-photon, is represented in the tensor product space (Sect. 2.5.1).
One may state that the crucial difference between the classical and quantum physical
models is in the representation of states of compound systems. (Although, as we
have already seen in previous sections, the descriptions of non-compound systems
differ essentially.)
Tensor product, the algebraic definition. Consider now two finite dimensional
Hilbert spaces, H1 , H2 . For each pair of vectors ψ ∈ H1 , φ ∈ H2 , weform a new
formal entity denoted by ψ ⊗ φ. Then we consider the sums Ψ = i ψi ⊗ φi .
On the set of such formal sums we can introduce the linear space structure. (To be
mathematically rigorous, we have to constraint this set by some algebraic relations
to make the operations of addition and multiplication by complex numbers well
defined). This construction gives us the tensor product H = H1 ⊗H2 . In particular,
(k) (1) (2)
if we take orthonormal bases in Hk , (e j ), k = 1, 2, then (eij = ei ⊗ e j ) form
an orthonormal basis in H , any Ψ ∈ H , can be represented as (2.52) with (2.53).
The latter representation gives the simplest possibility to define the tensor product
of two arbitrary (i.e., may be infinite-dimensional) Hilbert spaces as the space of
formal series (2.52) satisfying the condition (2.53).
Besides the notion of tensor product of states, we shall also use the notion of
tensor product of operators. Consider two linear operators Ai : Hi → Hi , i = 1, 2.
Their tensor product A ≡ A1 ⊗ A2 : H → H is defined starting with the tensor
products of two vectors: Aψ ⊗ φ = (A1 ψ) ⊗ (A2 φ). Then it is extended by linearity.
2.5 Compound Systems 33
By using the coordinate representation (2.52) the tensor product of operators can be
represented as
(1) (2)
AΨ = cij Aei j ≡ cij A1 ei ⊗ A2 e j , (2.54)
where α |cα |2 < ∞.
2.5.1 Entanglement
Ψ = ψ1 ⊗ ψ2 . (2.56)
(In particular, for wave functions we have Ψ (x, y) = ψ1 (x) × ψ2 (y).) However, if
there are some correlations between the degrees of freedom in S1 and S2 , then the
state of S cannot be represented in the form (2.56). Such a state is called an entangled
state. The notion of entanglement can be generalized to nonpure states (which are
represented by density operators acting in H .) The interpretation of this notion is
still the topic of hot debates [4]. There can be found variety of interpretations: from
the original interpretation of Einstein et al. [15], who considered entanglement as
the Hilbert space representation of classical correlations between systems to mod-
ern interpretations such as the subjective probability interpretation – entanglement
as the correlation of probabilistic knowledge about subsystems in the mind of an
observer, e.g., Fuchs and coauthors [16–18]. In this book we shall not try to select
34 2 Fundamentals of Classical Probability and Quantum Probability Theory
Tensor product structure can appear not only in the description of a state space of a
compound system, but even in the case of a single system. Sometimes it is possible to
factorize the state space H of a single quantum system, say a neutron, into the tensor
product H = H1 ⊗ H2 in such a way that two compatible observables (Hermitian
operators A1 , A2 acting in H ) can be represented in the form A1 = a1 ⊗ I and
A2 = I ⊗ a2 , where ai , i = 1, 2, are Hermitian operators in Hi and I denotes
the unit operator, I x = x (For neutrons, A1 and A2 can be selected as the position
and spin observables.) Here we also can consider entangled states. However, this is
entanglement of observables for a single system and not states of a few systems.
This approach is very useful for quantum bio-informatics. For example, we can
consider the entanglement of various genetic or epigenetic markers of a single cell
11 “What is beyond this symbolism?”—this is a separate question (Sect. 9.2; see also [19]). In a
series of papers [6, 20, 21, 23, 24] quantum systems were represented by classical random fields.
In this approach elements of the tensor product can be visualized via the functional representation.
2.5 Compound Systems 35
(Sect. 8.5). Using of the entanglement (-like)12 model can explain the high speed
of the epigenetic evolution, in which inheritable epigenetic markers can be created
during only one generation of a cellular population.
The notion of qubit was introduced in Sect. 2.2.4. Now we consider n-qubit states.
One qubit space is two dimensional with the basis encoding 0 and 1: |0 , |1 . It
is isomorphic to C2 . Now we consider the tensor product of n qubit spaces. The
basis in this space has the form |x ≡ |x1 ⊗ · · · ⊗ |xn , where x j = 0, 1. Thus
each basis vector |x encodes the string of zeros and ones of the length n. As in the
one dimensional case, it is possible to form weighted superpositions of these basis
vectors. These are quantum n-bit states. The dimension of the n-qubit state space is
n
equal to 2n . This space is isomorphic to C2 . Typically one omits the signs of the
tensor product and writes |x ≡ |x1 · · · xn . Thus a general n-qubit state has the form
|ψ = cx |x , where |cx |2 = 1. (2.57)
x x
The main distinguishing feature of quantum information theory is that the dimension
of a state space increases exponentially with the linear increasing of the number n of
subsystems of a compound system or more generally entangled degrees of freedom
(may be even of a single system, see Sect. 2.5.2). This is the main computational
resource. And it is huge. A relatively small physical system (say n = 100) can
process a gigantic information state (in our example, the superposition of 2n bits of
information). However, as in the single qubit case, this huge information resource is
unapproachable: a measurement can give us only one result.
Quantum computations are based on the physical realization of the Schrödinger
dynamics (unitary transformations) in the n-qubit spaces and the final measurement.
The last step has to be designed in an intelligent way, since the solution of a problem
has to be found in a single basis state obtained via measurement (Lüders projection).
This is one of the reasons why only a special class of problems can be solved by
using quantum computations. This class is very restricted. Ohya et al. proposed to
combine quantum algorithms with classical chaotic dynamics [25, 26]. This approach
12 We remark that information features of entanglement can be modeled (mimic) by using coarse
graining procedure for classical stochastic processes, even for classical Brownian motion [19]. In
the latter paper the entanglement is exhibited at the level of observables corresponding to coarse
graining.
36 2 Fundamentals of Classical Probability and Quantum Probability Theory
provides a possibility to solve new NP-problems13 , e.g., the SAT-problem, see Ohya
and Volovich [8, 27, 28].
However, in general the usage of unitary transformations makes physical realiza-
tion of quantum algorithms very difficult, since unitary dynamics is possible only in
isolation from the environment. The latter is practically impossible. The entangled
quantum n-qubit states are subjected to quick decoherence (destruction of superpo-
sition (2.57) and spontaneous reduction to one of the basis states). There is simply
not enough time to perform quantum calculations. Therefore we do not think that
bio-systems can perform real quantum calculations based on unitary transforma-
tions. Our conjecture is that bio-systems can process entangled states, but by using
non-unitary dynamics of the GKSL-type (Sect. 2.4), see Sect. 8.4.2 for more detail.
In Sect. 2.4 we considered the dynamics of the state of a system, say S, interacting
with an environment. To derive this dynamics, the Eq. (2.50), one proceeds in the
following framework. The states of the system under consideration are represented in
a Hilbert space H and the states of an environment (bath) are represented in another
Hilbert space K . Since an environment is a huge system, its state space has very high
dimension, in a mathematical model it can be considered as infinite-dimensional. The
states of the compound system, the system S and the environment, are represented in
the Hilbert space H ⊗ K . In the theory of open quantum systems the evolution of a
pure state of such a compound system is described (in the complete accordance with
the postulates of quantum mechanics) by the Schrödinger equation. (In quantum bio-
informatics we use its slight modification by using the dimensionless Hamiltonian
and the time scaling constant γ , instead of the Planck constant, see Sect. 2.3. In
this section we also proceed with such γ . In physics one can, as usual, proceed
with Hamiltonians having the dimension of physical energy and the Planck constant
instead of the scaling factor γ .) The Hamiltonian of a compound system is in general
very complicated. Therefore in real calculations its various approximations are in use.
Set Ut = e−it H/γ , the evolutionary operator of S interacting with the environment.
Then unitary dynamics of the pure state is given as
Ψ (t) = Ut Ψ0 , (2.58)
13 In computability theory, a decision problem, which has two possible answers, “yes” or “no”,
for an input of question is studied well, and its complexity is classified into several complexity
classes. NP (Nondeterministic Polynomial time) problem is a decision problem, whose solution is
not given in polynomial time on a non-deternimistic Turing machine, and NP-complete problem
is a NP problem reduced in polynomial time from any other NP problems. Whether NP-complete
problem can be reduced to Polynomial problem is one of the millennium prize problems, it has
been discussed for thirty yeas.
2.6 From Open Quantum System Dynamics to State-Observable Adaptive Dynamics 37
where Ψ0 is the initial state of S combined with the environment. For general states
given by density operators, the dynamics is given by the von Neumann equation and,
hence, it can be represented in the form, see (2.48),
where R0 is the initial state of the compound system. Typically one is interested
only in the dynamics of S (and not interested in what happens with an environment
around S). This dynamics is obtained by taking the trace of the density operator of
the compound system with respect to all environmental degrees of freedom:
where tr K denotes the partial trace with respect to the subspace K of the complete
state space H ⊗ K .
This dynamics is very complex because of the complexity of Hamiltonian of the
total system and, in particular, the complexity of interaction between S and environ-
ment. Then under a number of assumptions it is possible to prove that the dynamics
given by (2.60) can be reduced to the quantum master Eq. (2.50). One of the main
assumptions is Markovness of dynamics. Another important assumption is that inter-
action of S with the environment cannot change the state of the environment (or if
it changes then this change can be considered as negligible). (In Sect. 8.2.2 we shall
discuss conditions of derivation of (2.50) from (2.60) in more detail and in the biolog-
ical framework; in application to biological evolution.) We emphasize that in general
dynamics (2.60) is non-Markovian. In the latter case the system S “remembers” the
states of the compound system, S and the environment for sufficiently long period
and it makes its state update on the basis of this information.
We think that to match completely with biology, the theory of open quantum
systems has to be generalized. First of all we are critical to the approach assuming
a possibility to construct interaction Hamiltonian. This is very difficult in physics.
And in biology this problem is even more complicated. It is difficult or may be
even impossible to describe formally the character of the interaction of a bio-system
S with its environment. Such an environment has not only physical and chemical
counterparts, but also a complex information component. And the latter plays the
crucial role in many biological processes (e.g., signaling of surrounding cells is
crucial in the process of cell differentiation, Chap. 8. Starting with the Hamiltonian,
one postulates the Schrödinger dynamics for pure states and then the von Neumann
dynamics is its trivial consequence. However, the assumption that such a complex
system as an environment interacting with a system S can be in a pure state has
no justification, especially in biology. It is more natural to assume the presence of
mixture of states. Of course, one can easily avoid the last problem by postulating
directly the von Neumann equation. However, even for a density operator, the usage
of the unitary dynamics (2.59) is not justified.)
We think that, especially in quantum bio-informatics, the direct usage of the
quantum Hamiltonian approach to describe the dynamics of a system in some
38 2 Fundamentals of Classical Probability and Quantum Probability Theory
References
13. Zorn, C.H., Smith, Ch.: Pseudo-classical nonseparability and mass politics in two-party sys-
tems. In: Song, D., Melucci, M., Frommholz, I., Zhang, P., Wang, L., Arafat, S. (eds.) Quantum
Interaction-3. Lectures Notes in Computer Science, vol. 7052, pp. 83–94. Springer, Heidelberg
(2011)
14. Khrennikova, P., Khrennikov, A., Haven, E.: The quantum-like description of the dynamics of
party governance in the US political system (2012). arXiv:1206.2888 [physics.gen-ph]
15. Einstein, A., Podolsky, B., Rosen, N.: Can quantum-mechanical description of physical reality
be considered complete? Phys. Rev. 47(10), 777–780 (1933)
16. Fuchs, C.A.: Quantum mechanics as quantum information (and only a little more). In:
Khrennikov, A. (ed.) Quantum Theory: Reconsideration of Foundations. Series Mathemati-
cal Modeling 2, pp. 463–543. Växjö University Press, Växjö (2002)
17. Caves, C.M., Fuchs, Ch.A., Schack, R.: Quantum probabilities as Bayesian probabilities. Phys.
Rev. A 65, 022305 (2002)
18. Fuchs, Ch.A., Schack, R.: A quantum-Bayesian route to quantum-state space. Found. Phys.
41, 345–356 (2011)
19. Allahverdyan, A.E., Balian, R., Nieuwenhuizen, Th.M.: The quantum measurement process in
an exactly solvable model. In: Foundations of Probability and Physics-3. American Institute
of Physics. Series Conference Proceedings, vol. 750, pp. 16–24. Melville (2005)
20. Khrennikov, A.: Description of composite quantum systems by means of classical random
fields. Found. Phys. 40, 1051–1064 (2010)
21. Khrennikov, A., Ohya, M., Watanabe, N.: Classical signal model for quantum channels. J.
Russ. Laser Res. 31, 462–468 (2010)
22. Khrennikov, A.: Ubiquitous Quantum Structure: From Psychology to Finance. Springer,
Heidelberg-Berlin-New York (2010)
23. Khrennikov, A.: Prequantum classical statistical field theory: Schrödinger dynamics of entan-
gled systems as a classical stochastic process. Found. Phys. 41, 317–329 (2011)
24. Khrennikov, A., Ohya, M., Watanabe, N.: Quantum probability from classical signal theory.
Int. J. Quantum Inf. 9, 281–292 (2011)
25. Ohya, M.: Complexities and their applications to characterization of chaos. Int. J. Theor. Phys.
37, 495–505 (1998)
26. Ohya, M.: Some aspects of quantum information theory and their applications to irreversible
processes. Rep. Math. Phys. 27, 19–47 (1989)
27. Ohya, M., Volovich, I.V.: New quantum algorithm for studying NP-complete problems. Rep.
Math. Phys. 52, 25–33 (2003)
28. Ohya, M.: Adaptive dynamics and its applications to chaos and NPC problem. In: QP-PQ:
Quantum Probability and White Noise Analysis. Quantum Bio-Informatics, vol. 21, pp. 181–
216 (2007)
Chapter 3
Fundamentals of Molecular Biology
Abstract In this chapter, we briefly introduce the basic concept of molecular biol-
ogy and also biological information processing in general. As a detailed example
of the information processing phenomena (signal recognition, transformation, and
biological response) in a living system, we explain the diauxie (two phase) growth
of Escherichia coli in glucose and lactose mixed medium and show the mechanistic
simulation of the system according to the systems biology approach. In addition, we
introduce epigenetic mutation as another example for the information processing in
a living system.
Life is basically created on the flows of materials, energy, and information. In other
words, living organisms “live” in these flows at non-equilibrium steady state. Accord-
ingly, biological research is mainly targeted at the mechanisms of these major flows
(Fig. 3.1; [1]). We cite the introduction part from our publication [1] as below and
add discussion on our proposal to use “information biology” for the research field of
information flow.
During the World War II, a production method of (radio)-isotopes by nuclear
reactions was established. Biochemists used such isotopes as tracers to identify many
metabolic pathways, such as Calvin cycle [2] in photosynthesis in plants and Krebs
cycle [3]. Adenosine triphosphate (ATP) has been found by Fiske and Subbarow [4]
as a high energy substance in biochemical reactions. Muscle contraction occurs
depending on the consumption of the ATP hydrolysis energy [5], and Huxley and
Hanson proposed a sliding model of actomyosin fibers [6]. In muscles, the chemical
energy of ATP is converted to mechanical energy to exert force. There are many
kinds of energy transducing systems, such as photosynthesis, ion-pumping ATPases,
and so on. Then after the proposal of DNA structure by Watson and Crick [7],
Fig. 3.1 Research targets of 3 kinds of flows “Research field” (examples of researches)
life science (3 major flows) Flow of materials → “biochemistry” (metabolism, etc)
Flow of energy → “biophysics” (muscle contraction, etc)
Flow of information → “molecular biology” or “information biology”
(genetic code, signal transduction,
nerve/brain system, etc)
information flows and signal transduction in many biological systems have been
extensively elucidated, including highly complex information processing systems as
cognition and brain, which are significantly accelerated by the recent development
of large-scale and high-throughput research in biology, so called “-omic” studies,
such as genome, metabolome, transcriptome, proteome, and interactome, etc.
According to the DNA structure [7], genetic inheritance and information flows
from genes to proteins (genetic coding problem) have been studied, which is called
molecular biology in a narrow sense dealing with the information processing relat-
ing to genes. This information flow is now basically understood as a central dogma
in molecular biology [8]. Whereas signal transduction, which deals with the cellu-
lar information processing of environmental signals to cellular responses has been
studied biochemically; nerve/brain functions have been studied physiologically and
biochemically. They are now biochemically and biophysically studied at the mole-
cular level, i.e., molecular biologically. Thus the above named three research fields
cover mostly overlapping biological phenomena and the naming of them is rather
confusing and arbitrary. In Fig. 3.1, we proposed a classification of research fields
according to the kinds of research targets and also to the historical origin as explained
above. We included biochemical and biophysical research on information process-
ing (signal transduction and nerve/brain system, etc.) into molecular biology in a
wide sense, as dealing with any biological information processing. Or the naming of
molecular biology may mean any kind of biological research at the molecular level
in general, therefore may be taken to include both biochemistry and biophysics. If we
take this standpoint, a new name for the research field dealing with any information
processing in biology, such as “information biology” as proposed in a former (or
in the preface) chapter of this book, will be necessary: “Bioinformatics” seems to
be appropriate but many biologists may limit it as the research or analysis on any
kind of static biological data, such as DNA and amino acid sequences, with recent
development of simulation studies based on the structure biology data. Thus this
term may not be adequate for the general naming of researches on any biological
information processing. Therefore, we proposed the use of molecular biology for
the research on information processing in biology, by expanding the research fields
not only on the genetic coding problem, but also on any information processing in
biology (Fig. 3.1). Furthermore, in this book, we prefer using the information biology
as a new terminology for this research field because in the latter chapters we shall
show that our quantum-like formulation of any information processing in biology
gives the first holistic view of the information biology.
At the first conference of the quantum-bio-informatics research center (QBIC)
held in Japan in 2007, we presented basics of in silico biology to reproduce living
3.1 Research Fields in Life Science and Information Biology 43
systems in a computer based on the recent -omic data [9]. We expected to obtain
mechanistic parameters of whole biochemical reactions in a cell and, finally, to sim-
ulate the living system in a computer, which is the systems biology approach. In this
chapter, we introduce basics of classical molecular biology and information process-
ing in living organisms. As an example of the information processing, we present
diauxie effect of Escherichia coli (E. coli) growth explaining its molecular biolog-
ical basis and show an approach according to the systems biology for the analysis
of diauxie. In addition, we introduce epigenetic mutation as another example of
information processing.
At the era of the birth of the Earth, meteorites are thought to have fallen down
from universe with amino acids and bases in them, which are basic units of proteins
and nucleic acids, respectively. Somehow they formed ribonucleic acid (RNA), the
process of which is not well known. RNA is composed of ribose, bases, and phosphate
(Fig. 3.2). The existence of OH group at 2 position of ribose is important, that
makes RNA reactive. RNA can cut/form covalent bonds of itself or peptide bonds.
Furthermore, it can form double helix by the hydrogen bonds with bases in another
RNA strand having complementary base sequences, which enables the information
of the base sequence to be conserved and transferred to the other RNA. In other
words, this enables RNA to multiply with conserved base sequence. Thus RNA
is a kind of living things in the sense that it has the ability of assimilation and
information transmission to the next generation. The information transfer process
can be perturbed. This can generate mistakes in the base sequence conservation by
accidents. This is thought to be the first step of evolution (chemical evolution) of
living systems. Thus, at the beginning of the Earth and evolution of living things, the
RNA world [10] should have been prevailing.
Then RNA somehow created DNA with a similar property but without high reac-
tivity of RNA. They were thought to be covered with lipid bilayers (biomembranes)
to give birth to the first living organism, a primitive prokaryote.
In 1953, Watson and Crick reported a double helix model of DNA structure
[7]. DNA is now known to be responsible for genetic information transfer to the
next generation. As described above, the information of genes can be transmitted to
daughter cells by the conserved base sequences determined by hydrogen bonding
complementarity. After this discovery of DNA structure, the central dogma in mole-
cular biology was proposed, which suggested that the information of DNA sequence
as a gene is transcribed to a messenger RNA (mRNA) and it is translated into a
protein (Fig. 3.3) [8]. Proteins are major players in living systems. They work as
enzymes to catalyze chemical reactions or constituents in living bodies and create
living systems with other substances. Proteins are polymers consisting of amino
acids, whose sequences are determined by the base sequences with coding rule of
triplet (Table 3.1). DNA (or RNA) are composed of 4 kinds of bases, guanine (G),
44 3 Fundamentals of Molecular Biology
(a)
(b)
(c)
Fig. 3.2 Structures of ribose, deoxyribose, ribonucleotide and synthesizing DNA. a Chemical
structures of ribose and deoxyribose are shown. The OH at 2 site in 5 member ring of ribose is
important and responsible for high reactivity of RNA. b Bases of adenine (A), uridine (U) (thymine
(T) in case of DNA), guanine (G), and cytosine (C) are covalently linked to the OH group at the 1
site of ribose or deoxyribose. They are called ribonucleoside or deoxyribonucleoside. Phosphate is
linked to the OH at 5 site forming a ribonucleoside monophosphate. This phosphate forms bridge
with the 3 OH of the next nucleotide forming RNA in case of ribonucleotides and DNA in case of
deoxyribonucleotides. c Bases in DNA or RNA can form hydrogen bonding (represented by 2 dots
or 3 dots between adjacent two bases on different strands depending on the hydrogen bond numbers)
with complementary bases (A:T(U) and G:C) which enables the long DNA or RNA sequences to
form double helix with complementary strands, which is the basic mechanism of genetic information
conservation into next generation. In this specific figure, replicating DNA is drawn schematically.
The deoxyribose rings are shown as circles and the phosphate ester bridges are represented by bars
connecting the circles (deoxyribose), resulting in a long chain of nucleotides, which are called as
DNA (in case of deoxyribose). The replication apparatus unwinds the helix of parental DNA (upper
part) and synthesizes new daughter DNA strands complementary to the parental DNA strands at
the replication fork, which is the mechanism of genetic information conservation
3.2 Molecular Biology and Genome 45
Fig. 3.3 Genetic information flow from DNA (RNA) to proteins to living systems [1]. The base
sequences on DNA (a) are transcribed into the sequences on RNA, which are translated into amino
acid sequences in proteins (b). Proteins fold by themselves to form well-defined three dimensional
structures (c). The assembly of proteins and other components create a living system, such as an
Enterococcus hirae (E. hirae) prokaryotic cell (d), having a representative metabolic system shown
in (e). c A schematic representation of the three dimensional structure of a protein with helices and
sheets. d Electronmicrograph of a bacterium, E. hirae. e A scheme of the metabolic pathway in E.
hirae, showing uptake of glucose outside, glycolytic digestion of glucose producing lactate which
is exported outside with production of ATP, which then is utilized to pump out inside H+ and Na+
by membrane bound ion-translocating ATPases
adenine (A), thymine (T) (uridine, U), and cytosine (C), which are responsible for
the information transfer by their hydrogen bonding complementarity (see Fig. 3.2).
The combination of 3 bases in sequence makes up one word, thus, in principle, there
can be 64 (=4 × 4 × 4) different words, where each word corresponds to a cer-
tain amino acid in proteins. However, proteins are composed of only 20 kinds of
amino acids. Therefore, there are several redundancies in the usage of triplets (the
table, such as Table 3.1, is called codon usage table). Proteins fold into the well-
defined three dimensional structures according to their amino acid sequences in a
46 3 Fundamentals of Molecular Biology
As we have described above, a protein spontaneously folds into its certain three
dimensional structure determined by its amino acid sequence. The folding process
of proteins typically takes place on the temporal scale from millisecond to second.
The folding involves many kinds of interactions between atoms in proteins as well as
surrounding solvent molecules/atoms, such as van der Waals, electrostatic, hydrogen
bonding, and hydrophobic interaction in water. A protein with 100 amino acids, for
example, may consist of about 2,000 atoms, which are surrounded by many water
molecules. In 1969, Levinthal showed a paradox in a theory of protein folding [13]:
3.2 Molecular Biology and Genome 47
Let us assume that proteins search their most stable structures by checking the sta-
bility one by one through changing every dihedral angle in the protein backbone, in
which each amino acid residue has been assumed to take only 9 stable conforma-
tions in dihedral angles. For a 100 amino acid protein (since in chemistry transition
time of one dihedral angle is about 100 ps), the search would take 10−10 s × (9)100
dihedral bonds, corresponding to ∼8 × 1077 years, which is far beyond the time of
universe! Therefore, he proposed that some secret mechanisms should be adopted
in the protein folding process. The protein folding mechanism has been suggested
to follow a glass transition theory in statistical mechanics by Bryngelson et al. as a
funnel model [14, 15]. Currently, the whole processes of the folding of several small
proteins have been successfully simulated by an atomic level molecular dynamics
simulation using a specially designed CPU, ANTON [16].
In this book, we formulate an adaptive dynamics for phenomenological descrip-
tions of biological events as a whole. We think that it is important for each composite
element to have interactions with other members forming an interaction network.
Even in the protein folding process, each composite atom is interacting with other
members by various kinds of forces. Accordingly, we can imagine that to reach its
final structure, the folding process must be as quick as a quantum computer: Each step
cannot be described according to classical total probability conservation law. There-
fore, this system may be a good example of such adaptive dynamics. An example of
such system has been described in our previous publication [17].
As described in Fig. 3.1, there are many kinds of information transducing processes
in biology. Above we described the central dogma underlying the genetic coding
mechanism. In biology, we study many other kinds of information transducing
processes, such as signal transduction, cognition (brain system), evolution, ecologi-
cal system, etc. Signal transduction plays major roles in many biological phenomena
such as adaptation, reception/response of environmental signals such as hormones,
development, differentiation, embryogenesis, etc. A representative example is lac-
tose/glucose diauxie phenomena of E. coli growth. It contains several information
processing steps, such as signal recognition, interference by signals of other cellu-
lar processes, and gene expression regulations. The lactose operon is the first gene
expression regulatory system described by Jacob and Monod [18], who were the
Nobel Laureates for the discovery and proposal (operon theory). We explain in detail
the lactose operon and the diauxie phenomena of E. coli in terms of molecular biology
in the next section. Other signal transduction systems, such as differentiation, devel-
opment, and so on, can be described analogously in the sense that their information
processing system is composed of network systems with many elements interacting
with each other. Adaptive dynamics is powerful in dealing with such complex sys-
tems in a phenomenological way. In further chapters we shall show that every such
48 3 Fundamentals of Molecular Biology
system violates classical total probability conservation law. Some of the examples
have been presented previously [17, 19].
Other biological systems of brain/cognition, evolution, ecological systems, etc.,
even including economical activities, can be similarly formulated by the quantum-
like formalism. One of the authors of this book recognized that the total proba-
bility conservation law was violated in biological phenomena, such as cognition,
psychology, and so on, from the viewpoint of the analogy to quantum mechanical
formalism [20]. Another author invented and formulated adaptive dynamics during
his study on quantum information theory by taking into account that the observation
or interaction with the environment is essential in the real world to give the quantum
mechanical behavior in physics [21]. By our collaboration [17, 19], we found that
this adaptive dynamics is powerful to describe and formulate the quantum-like fea-
tures in biological phenomena. Such topics on the concept and tools of quantum-like
formulation of biological phenomena are discussed in the latter chapters of this book.
They have many composite elements interacting with each other, forming networks,
thus one element influences every other element and vice versa. Furthermore, they
interact at the same time, summing up all effects at instance. This is the same working
principle as quantum computer, and the response should be very quick and already
optimized. Our adaptive dynamics can lead to phenomenological formulation of
these phenomena, which is the main proposal of our book.
Fig. 3.4 E. coli diauxie phenomenon. a The growth of E. coli was monitored in time dependent
manner. The cells grown in the presence of low concentration of glucose and usual concentration of
lactose shows biphasic growth character. This classical observation has been known for long [24]
and the figure was drawn schematically, showing the biphasic growth curve and the expression of
β-galactosidase upon consumption of glucose in the medium. b The schematic molecular mech-
anism of the diauxie [19]. The details are described in the text. Big ovals represent cells. Boxes
with I, P, O, Z, Y and A inside represent genes of lactose repressor and lactose operon on E. coli
genome DNA. Black ovals on the cell membrane represent lactose permease (LacY) responsible for
the lactose transport into cells. Grey ovals represent EnzymeIICBGlc in glucose phosphotransferase
system (PTS) working for transport and phosphorylation of glucose outside forming glucose-6-P
inside cells
is produced from ATP by adenylate cyclase in cell membrane when the protein is
activated by EnzymeIIAGlc (Figs. 3.4 and 3.5).
The mechanism of the diauxie is explained by catabolite repression and inducer
exclusion [22, 23]. Catabolite repression means repression of the gene expression.
Glucose is taken up by E. coli through glucose phosphotransferase system (PTS)
(Fig. 3.5; [26]). Phosphoenolpyruvate inside cells is used to form phospholyrated HPr
(heat stable protein or histidine containing protein), HPr-P, by EnzymeIGlc . Next,
phosphate in HPr-P is transferred to EnzymeIIAGlc to form the phosphorylated form
of EnzymeIIA. The phosphate in EnzymeIIA is used to phosphorylate and trans-
port outside glucose to make inside glucose-6-P by EnzymeIICBGlc in membrane.
The phosphorylated form of EnzymeIIAGlc , which is produced and maintained in
large amount in the absence of glucose in medium, activates the membrane bound
adenylate cyclase to produce large amount of cAMP, thus advantageous for lactose
operon gene expression. The non-phosphorylated EnzymeIIAGlc , which is produced
in large amount when glucose outside is transported, inhibits the LacY, lactose per-
mease in membrane. As a consequence of the inhibition, lactose (inducer) cannot
be transported into the cell. Therefore, this mechanism is called inducer exclusion.
50 3 Fundamentals of Molecular Biology
Fig. 3.5 Glucose phosphotranspherase system (PTS) in E. coli [26]. The details are described in
the text. The first reaction (I) in the PTS system is the phosphate transfer from phosphoenolpyruvate
(PEP) to His residue in the Enzyme I producing pyruvate. The second transfer process of phos-
phate (II) is the phosphorylation of His residue in HPr (His containing or heat stable protein) by
Enzyme I phosphate. The third reaction (III) is the phosphate transfer from HPr-P to Enzyme IIAGlc
(or Crr = catabolite repression resistant), the phosphate of which in turn is transferred (IV ) to
Enzyme IIB in membrane bound Enzyme IICB (PtsG) complex, which then is transferred to glu-
cose outside transported through Enzyme IICB complex forming glucose-6-phosphate inside the
cell (V ). IIA is the key component in the regulation of diauxie. The phosphorylated form of IIA is
the signal of the absence of glucose outside because without glucose transport, the phosphate group
in IIA can be kept for a long time without consumption to form Glc-6-P, thus phosphorylated form
of IIA activates membrane bound adenylate cyclase, which produces cAMP, a signal substance of
catabolite repression release. cAMP in turn is bound to CRP or CAP to form complex, which binds
to the promoter region of lactose operon DNA, which then stimulates the RNA polymerase binding
to the promoter region of the lactose operon, thus activating the lactose operon gene expression. On
the contrary, the non-phosphorylated form of IIA is the signal of the presence of glucose outside
because during glucose transport the phosphate group in IIA is constantly consumed to be trans-
ferred to glucose, thus IIA non-phosphorylated form accumulates and directly inhibits the lactose
permease activity inhibiting lactose transport, which is called inducer exclusion
As a result, E. coli establishes its system to utilize glucose more preferentially than
lactose, which is convenient for the cell, because glucose is the easier starting sub-
strate for energy metabolism in glycolysis pathway than lactose. (In order to be uti-
lized, lactose should be first hydrolyzed to galactose and glucose by β-galactosidase.)
Summing up the aforesaid (Fig. 3.4), when glucose and lactose are present in
medium, E. coli cells tend to utilize glucose first by inhibiting lactose utilization
by gene expression repression owing to the low cAMP (catabolite repression) and
LacY inhibition (inducer exclusion). After glucose consumption, cells recognize
that there is no glucose in the medium and start to change the cellular system to
be able to utilize lactose, which is the adaptation to environment by raising cAMP
concentration to start gene expression and by stopping inducer exclusion to take up
the inducer, lactose. During this changing period of the cellular system, cells start to
produce β-galactosidase and stop growth, which brings about the two step growth
phenomenon.
As understood by the above description, this process includes the reception of a
signal outside, signal transduction through cAMP production and LacY inhibition,
and regulation of the gene expression. These kinds of players similarly play important
roles in the development, differentiation, and embryogenesis. Thus, their features
should be similarly described by our adaptive dynamics.
3.3 Various Information Transductions in Biology 51
Cognition (brain system), evolution, and ecological systems have similar mutual
network interactions between member elements. Therefore, we can expect that they
can be described essentially in a similar manner. We would like to describe our
systems biology approach to the diauxie phenomenon in the next section to compare
it with our theoretical, phenomenological, and holistic approach described in the
latter chapters.
In this sense, the adaptive dynamics is very useful to describe complex systems
from micro (atomic level interaction leading to protein self-organization, folding) to
macro (species or individual interaction leading to ecological state on Earth). The
“adaptive” is a good term because it is suitable to describe such events as adap-
tation in biology: This adaptation is also the main reaction in the differentiation,
development, and embryogenesis. We hope that the readers of this book realize that
the biological systems can be modeled as such quantum-like systems finding out
optimized responses (answers) very quickly and efficiently.
Fig. 3.6 System simulation of the diauxie growth of E. coli. a Schematic presentation of the system
we used as shown on a PC monitor. This system consisted of glycolytic network, gene expres-
sion/regulation, and signal transduction (inducer exclusion and catabolite repression) as described
in the text. Circles represent proteins and metabolites, boxes represent reactions, and edges repre-
sent the connection between proteins, metabolites, and their reactions. b Simulation result showing
the ATP concentration in the cell during growth in the presence of glucose and lactose mixture.
ATP increased biphasically mimicking the biphasic growth of cell, diauxie. During the first growth
phase, glucose in the medium was consumed and lactose retained. During the second growth phase,
after a short period of cease of growth, lactose was consumed. The β-galactosidase production (lacZ
gene expression) was induced during the intermittent growth stop phase after the consumption of
glucose and the products accumulated to be used for lactose hydrolysis and utilization
Our living organisms utilize three typical regulation mechanisms for regulating flows
of matter (metabolism etc.), energy, and information (such as adaptation that we
described above, etc.). The first one is the allosteric effect or ligand induced con-
formational change causing functional change of a protein, an example of which in
the information flow is the allo-lactose induced LacI repressor dissociation from the
operator region of lactose regulatory region or the inducer exclusion caused by the
non-phosphorylated EnzymeIIAGlc inhibiting directly LacY permease. The second
one is the chemical modification of a protein changing its functional activity, an exam-
ple ofwhich in the information flow is the phosphorylation or de-phosphorylation of
EnzymeIIAGlc changing its activity, on the one hand activating adenylate cyclase
producing cAMP for gene expression stimulation and on the other hand inhibiting
the LacY permease directly bringing about the inducer exclusion. The above two
mechanisms are suitable for quick response to an environmental change (in a few
seconds or in a few minutes). The third one involves the gene expression regulation
leading to new protein production, which usually takes more time (a few hours) but
still within the individual own generation, an example of which in the information
flow is the gene expression regulation of lactose operon depending on the presence
of lactose or glucose. We described these mechanisms in detail above (Sect. 3.3.1
and Fig. 3.4). But these three mechanisms do not utilize mutations (base sequence
changes in genome) in any sense.
3.3 Various Information Transductions in Biology 53
In the case of diauxie (Sect. 3.3.1), the adaptation process is not accompanied by
any base sequence changes (mutation) or modification of the DNA. Still, the adapted
phenotype of the parental cell can be transmitted to the daughter cells for several
generations even in the absence of the specific environmental pressure, which is an
apparent indication of epigenetic mutation. But for such single cell organisms, we do
not call this phenomenon epigenetic mutation, and we rather classify as adaptation.
In multicellular organisms, adaptation takes place in their bodies (usually in
somatic cells) by mechanisms with or without modifications of their chromatin struc-
tures. The phenotypic change caused by the chromatin structure modifications with-
out base sequence changes is usually called epigenetics. The mechanism seems quite
similar to that adopted in the above adaptation process in E. coli (Sect. 3.3.1); the
environmental pressure stimulates or inhibits the chromatin modification system to
induce or repress the relevant gene expression. The only difference is that epige-
netics utilize chromatin structure modification such as DNA methylation or histone
acetylation, etc., without base sequence changes. Recently it has been known that
such epigenetic effect can be transmitted to the next generation called epigenetic
mutation or epimutation in order to distinguish it from the usual mutation caused by
base sequence changes. (As understood from the above explanation, the definitions
of epigenetics and epimutation are not unequivocal.)
The epigenetics is a kind of adaptation and the mechanism of the appearance
of epigenetics should be similar to the adaptation processes. But in the case of
epimutation, the phenotype is transmitted to the next generation even in the absence
of the environmental pressure. Therefore, epimutation is thought to be the underlying
mechanism leading to a Lamarckian evolution. Thus it is easy to understand that such
mutation and evolution are fast as compared with the usual evolution taken place
by the Darwinian mechanism with random mutations and selection under certain
environmental pressure. We formulated such epimutation and evolution processes
using our quantum-like formalism in the latter chapters [31].
However the final answer or response obtained by epimutations or random muta-
tions/selection processes should be the same since both evolution processes occur
under the condition with every composite element or member interacting with each
other and also with the environment forming a big interacting network similar to the
system of adaptation. The only difference is the speed of the appearance. In the latter
chapters of this book, we present the application of our quantum-like formalism or
adaptive dynamics to the epimutation phenomena.
References
1. Yamato, I.: From structure and function of proteins toward in silico biology. Quantum Bio-
Informatics V. (From Quantum Information to Bio-Informatics). In: Accardi, L., Freudenberg,
W., Ohya, M. (eds.) Proceedings of the International Symposium of Quantum Bio-Informatics
Research Center 2011, Chiba, pp. 473–485. World Scientific, Singapore (2013)
2. Bassham, J., Benson, A., Calvin, M.: The path of carbon in photosynthesis. J. Biol. Chem. 185,
781–787 (1950)
54 3 Fundamentals of Molecular Biology
3. Krebs, H.A., Johnson, W.A.: Metabolism of ketonic acids in animal tissues. Biochem. J. 31,
645–660 (1937)
4. Fiske, C.H., Subbarow, Y.: Phosphorus compounds of muscle and liver. Science 70, 381–382
(1929)
5. Szent-Györgyi, A., Banga, I.: Adenosinetriphosphatase. Science 93, 158 (1941)
6. Huxley, H., Hanson, J.: Changes in the cross-striations of muscle during contraction and stretch
and their structural interpretation. Nature 173, 973–976 (1954)
7. Watson, J.D., Crick, F.H.C.: Molecular structure of nucleic acids: a structure for deoxyribose
nucleic acid. Nature 171, 737–738 (1953)
8. Crick, F.: Central dogma of molecular biology. Nature 227, 561–563 (1970)
9. Yamato, I., Ando, T., Suzuki, A., Harada, K., Itoh, S., Miyazaki, S., Kobayashi, N., Takeda,
M.: Toward in silico biology (from sequences to systems). Quantum Bio-Informatics (From
Quantum Information to Bio-Informatics). In: Accardi, L., Freudenberg, W., Ohya, M. (eds.)
Proceedings of the International Symposium of Quantum Bio-Informatics Research Center
2007, Chiba, pp. 440–455. World Scientific, Singapore (2008)
10. Gilbert, W.: The RNA world. Nature 319, 618 (1986)
11. Anfinsen, C.B.: Principles that govern the folding of protein chains. Science 181, 223–230
(1973)
12. Gibson, D.G., Benders, G.A., Andrews-Pfannkoch, C., Denisove, E.A., Baden-Tillson, H.,
Zaveri, J., Stockwell, T.B., Brownley, A., Thomas, D.W., Algire, M.A., Merryman, C., Young,
L., Noskov, V.N., Glass, J.I., Venter, J.C., Hutchison 3rd, C.A., Smith, H.O.: Complete chemical
synthesis, assembly, and cloning of a Mycoplasma genitalium genome. Science 319, 1215–
1220 (2010)
13. Levinthal, C.: How to fold graciously. Mossbauer Spectroscopy in Biological Systems: Pro-
ceedings of a Meeting Held at Allerton House, pp. 22–24. Monticello, Illinois (1969)
14. Bryngelson, J.D., Wolynes, P.G.: Spin glasses and the statistical mechanics of protein folding.
Proc. Natl. Acad. Sci. USA 84, 7524–7528 (1987)
15. Bryngelson, J.D., Onuchic, J.N., Socci, N.D., Wolynes, P.G.: Funnels, pathways, and the energy
landscape of protein folding. Proteins: Struct., Funct., Bioinform. 21, 167–195 (1995)
16. Shaw, D.E., Maragakis, P., Lindorff-Larsen, K., Piana, S., Dror, R.O., Eastwood, M.P., Bank,
J.A., Jumper, J.M., Salmon, J.K., Shan, Y., Wriggers, W.: Atomic-level characterization of the
structural dynamics of proteins. Science 15, 341–346 (2010)
17. Basieva, I., Khrennikov, A., Ohya, M., Yamato, I.: Quantum-like interference effect in gene
expression: glucose-lactose destructive interference. Syst. Synth. Biol. 5, 59–68 (2011)
18. Jacob, F., Monod, J.: Genetic regulatory mechanism in the synthesis of proteins. J. Mol. Biol.
3, 318–356 (1961)
19. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Tanaka, Y., Yamato, I.: Quantum-like model
of diauxie in Escherichia coli: operational description of precultivation effect. J. Theor. Biol.
314, 130–137 (2012)
20. Khrennikov, A.: Classical and quantum mechanics on information spaces with applications
to cognitive, psychological, social and anomalous phenomena. Found. Phys. 29, 1065–1098
(1999)
21. Ohya, M.: Adaptive dynamics and its applications to chaos and NPC problem. QP-PQ: quantum
probability and white noise analysis. Quantum Bio-Inform. 21, 181–216 (2007)
22. Beckwith, J.R., Zipser, D.: The Lactose Operon. University of Tokyo Press, Tokyo (1971)
23. Beckwith, J.R.: The lactose operon in Escherichia coli and Salmonella typhimurium. In: Neid-
hardt, F.C., Ingraham, J.L., Low, K.B., Magasanik, B., Schaechter, M., Umbarger, H.E. (eds.),
pp. 1444–1452. ASM, Washington (1987)
24. Loomis Jr., W.F., Magasanik, B.: Glucose-lactose diauxie in Escherichia coli. J. Bacteriol. 93,
1397–1401 (1967)
25. Inada, T., Kimata, K., Aiba, H.: Mechanism responsible for glucose-lactose diauxie in
Escherichia coli challenge to the cAMP model. Genes Cells 1, 293–301 (1996)
26. Postma, P.W.: Phosphotransferase system for glucose and other sugars in Escherichia coli
and Salmonella typhimurium. In: Neidhardt, F.C., Ingraham, J.L., Low, K.B., Magasanik, B.,
Schaechter, M., Umbarger, H.E. (eds.), pp. 127–141. ASM, Washington (1987)
References 55
27. Kitano, H.: Systems biology: a brief overview. Science 295, 1662–1664 (2002)
28. Tomita, M., Hashimoto, K., Takahashi, K., Shimizu, T.S., Matsuzaki, Y., Miyoshi, F., Saito,
K., Tanida, S., Yugi, K., Venter, J.C., Hutchison 3rd, C.A.: E-CELL: software environment for
whole-cell simulation. Bioinformatics 15, 72–84 (1999)
29. Mendes, P.: GEPASI: a software for modeling the dynamics, steady states and control of
biochemical and other systems. Comput. Appl. Biosci. 9, 563–571 (1993)
30. Albert, J., Rooman, M.: Dynamic modeling of gene expression in prokaryotes: application to
glucose-lactose diauxie in Escherichia coli. Syst. Synth. Biol. 5, 33–43 (2011)
31. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Yamato, I.: A model of epigenetic evolution
based on theory of open quantum systems. Syst. Synth. Biol. (2013). doi:10.1007/s11693-013-
9109-2
Chapter 4
Adaptive Dynamics and General Approach
to Non-Kolmogorov Probability Theory
In this chapter, we will discuss the non-Kolmogorov probability theory. Our aim is
to explain where and why the usual probability theory is broken. Further, we discuss
the mathematical foundation of lifting theory and the concept of adaptive dynamics
with its mathematical description. Then we apply these mathematical basis to the
study of the non-Kolmogorov probability theory.
The two slit experiment is the basic example demonstrating that QM describes statis-
tical properties in microscopic phenomena, to which the classical probability theory
seems to be not applicable, see, e.g., Feynman et al. [1]. In this section, we consider
the experiment with the symmetric setting: the source of photons is located symmet-
rically with respect to two slits, Fig. 4.1. Consider a pair of random variables a and
b. We select a as the “slit passing variable,” i.e., a = 0, 1, see Fig. 4.1, and b as the
position on the photo-sensitive plate, see Fig. 4.2. Remark that the b-variable has the
continuous range of values, the position x on the photo-sensitive plate. We denote
P(a = i) by P(i) (i = 1, 2), and P(b = x) by P(x).
where ψ0 and ψ1 are two wave functions, whose absolute values |ψi (x)|2 give the
distributions of photons passing through the slit i = 0, 1, see Fig. 4.2.
The term
|ψ0 (x)| |ψ1 (x)| cos θ
implies the interference effect of two wave functions. Let us denote |ψi (x)|2 by
P(x|i), then the Eq. (4.1) is represented as
P(x) = P(0)P(x|0) + P(1)P(x|1) + 2 P(0)P(x|0)P(1)P(x|1) cos θ. (4.2)
Here the values of probabilities P(0) and P(1) are equal to 1/2 since we consider the
symmetric settings. For general experimental settings, P(0) and P(1) can be taken
as the arbitrary nonnegative values satisfying P(0) + P(1) = 1. In the above form,
the classical probability law, the formula of total probability (FTP),
In this section, we shall repeat the previous considerations on the violation of FTP in
the two slit experiment. Now we proceed in the abstract contextual framework, see
the monograph [2] for details. We state again that the “two slit experiment” consists
of a few experiments with essentially different experimental setups. And, by violating
FTP, one operates with statistical data collected in these incompatible experiments.
Operating with abstract contexts may be difficult for biologists. Therefore one can
read only the beginning of this section devoted to the contextual reformulation of the
notion of conditional probability and then jump to formula (4.10).
In classical probability theory we assume that it is possible to form the event
B ∩ A, “both events A and B take place.” Suppose that these events correspond to the
measurements of two (discrete) random variables a and b. That is, we can consider
the sets of the events Aα = {a = α} and Bβ = {b = β}. If these random variables are
incompatible, i.e., it is impossible to construct an experimental context for their joint
measurement, then the event Cα,β = Aα ∩ Bβ is meaningless. This happens, e.g.,
in quantum physics with incompatible observables represented by noncommutative
4.1 Violation of Formula of Total Probability and Non-Kolmogorov Probability Theory 61
operators. (In Sect. 4.1.2, we discuss joint measurements in a quantum system, and
its relation with noncommutativity.) Therefore conditioning with respect to an event
has a restricted domain of application. In the general situation conditioning has to
be treated as conditioning with respect to the context of measurement [3].
Hence, in the expression P(B|C) the symbols B and C have to be treated in very
different ways.1 The first one, e.g., B = Bβ , still denotes an event corresponding to
the measurement for the value β of the random variable b. However, the symbol C
is used to denote context for the measurement of the b-observable. We now turn to
FTP. Can we generalize FTP by using context conditioning instead of Kolmogorovian
event conditioning?
As was shown, see (4.2), in quantum physics the standard FTP can be violated.
To simplify the introduction to the two slit experiment, in Sect. 4.1.1 we studied
this experiment in the symmetric setting, the source was located symmetrically with
respect to the slits. In such a setting we can use equal probabilities for passing through
slits without further explanation. To formalize the general situation, we introduce
two (incompatible!) observables: b gives the point x of detection on the registration
screen and a = 0, 1 gives information about which slit is passed. Both observables
are measured under the context C, both slits are open, (Figs. 4.1 and 4.5 ) and the
following contextual probabilities are obtained: P(b = x|C) and P(a = i|C), i =
0, 1. (In Sect. 4.1.1 these probabilities were denoted as P(x) and P(i), i = 0, 1).
Besides the context C, two other contexts are involved: Ci , only i-th slit is open,
i = 0, 1, Figs. 4.3 and 4.4. We also measure the b-observable with respect to these
two contexts and obtain the contextual probabilities P(b = x|C0 ) and P(b = x|C1 ).
(In Sect. 4.1.1 these were simply P(x|0), P(x|1).) We remark that, for these contexts,
we can measure even the a-observable and we have:
1 We started to use the symbol C to denote experimental context; symbols A, B are reserved
for events.
62 4 Adaptive Dynamics and General Approach to Non-Kolmogorov Probability Theory
As in Sect. 4.1.1, we obtain the generalized FTP as the following formula with
the interference term:
Thus the classical FTP is violated. As was already pointed in Sect. 4.1.1, the main
source of surprise by violation of classical FTP (and, hence, classical probability
theory), see, e.g., Feynman et al. [1], is the application of event conditioning and the
set-theoretical algebra for events. Motivated by (4.4) and (4.5) people (even experts
in quantum foundations) typically identify the contexts Ci , i = 0, 1, with the events
Ai = {a = i} and the context C with the event A0 ∪ A1 . One also expects that FTP
holds true:
(with the natural assumptions: P(Ai ) = P(Ai |A0 ∪ A1 ) and P(b = x) = P(b =
x|A0 ∪ A1 )).
Now we can repeat the previous contextual analysis in the abstract framework, i.e.,
without coupling to the two slit experiment. There are two in general incompatible
observables, say a and b. For our applications, it is sufficient to consider dichotomous
observables, a = α1 , α2 and b = β1 , β2 . There are given the experimental context
C and probabilities with respect to it, P(a = αi |C) and P(b = βi |C). There are also
given two other contexts denoted by Cαi , i = 1, 2, such that
|λi | ≤ 1, i = 1, 2, (4.11)
4.1 Violation of Formula of Total Probability and Non-Kolmogorov Probability Theory 63
However, in general the interference coefficients can exceed one! This can happen
even in quantum theory—for generalized observables a and b, which are represented
mathematically not by Hermitian operators, but by positive operator valued mea-
sures (POVMs), see Sect. 10.1 for details. In quantum bio-informatics it is common
that interference is so strong that |λ| > 1, Sect. 5.1.2, Eq. 5.3. Thus here the stan-
dard quantum formalism based on the representation of observables by Hermitian
operators is not applicable. One has to use its generalizations, which would describe
adequately the biological situation. One possibility is to use POVMs (as people do in
quantum information theory). And we test this apparatus in Chap. 10. However, we
prefer to apply a novel quantum probabilistic model [4] based on the entanglement
between a system and an environment. This model is based on such an advanced tool
of quantum information theory as a lifting map, see Sect. 4.2 for the detailed pre-
sentation. This model matches very well with the probabilistic structure of quantum
bio-informatics, from cells to brains.
The derivation of FTP with the interference term is straightforward [3]. Simply
define the interference coefficients as
P(b = βi |C) − P(a = α1 |C)P(b = βi |Cα1 ) − P(a = α2 |C)P(b = βi |Cα2 )
λi =
2 P(a = α1 |C)P(b = βi |Cα1 )P(a = α2 |C)P(b = βi |Cα2 )
(4.14)
and then solve this equality with respect to the probability P(b = βi |C).
The meaning of the nominator is clear: it is the difference between two types
of contextual probabilities, one can say “interference” between these contexts. The
denominator gives a proper normalization, cf. Sect. 4.1.1.
In Sect. 4.1.1, the FTP with the interference term was derived by using the quan-
tum wave function, in particular, the phase θ was related to the phase of the wave
function. Surprisingly, one can proceed the other way around and reconstruct the cor-
responding quantum state ψ, the “wave function”, from the FTP with the interference
term, see [3].
64 4 Adaptive Dynamics and General Approach to Non-Kolmogorov Probability Theory
As we have seen, the coefficient of interference given by (4.13) need not be bounded
by one. This is one of the reasons to use generalizations of quantum formalism based
on lifting maps and POVMs. In this section we present another reason for departure
from the usage of the standard quantum observables given by Hermitian operators.
Consider the quantum formalism in the two dimensional space. There are given two
quantum observables a and b represented by Hermitian matrices (2 × 2), a and
b with
eigenvalues α1 , α2 and β1 , β2 , respectively. These are values which can be obtained
in measurements of these observables. Let Cib = {b = βi }. Consider contextual
probabilities pji = P(a = αj |Cib ). Thus we measure the a-observable under the con-
dition that the b-observable was measured and the result b = βi was obtained. These
probabilities are also known as transition probabilities (Sect. 4.1.1). We remark that,
for any context C, P(a = α1 |C) + P(a = α2 |C) = 1. (The total probability to get
one of the values of a under the fixed context C equals one.) Hence,
The combination of the constraints (4.15) and (4.16) is known as double stochastic-
ity. (The matrix of transition probabilities (pji ) is called doubly stochastic.) This is
essentially a quantum property. In classical probability theory double stochasticity
can be violated. Here typically we have only stochasticity of transition probabilities.
However, biological systems can produce contextual (transition) probabilities,
which violate the condition of double stochasticity, see, e.g., Sect. 5.1.2 and also
[5–9] it was pointed out that the well known statistical data on the disjunction effect
collected by Shafir and Tversky [10–12] violate this condition. Hence, sometimes the
conventional quantum observables given by Hermitian matrices cannot be applied.
More general quantum observables, see Sect. 4.1.1 for discussion, have to be used.
At the end of this section, we explain the origin of the term transition probabilities
and derive double stochasticity of quantum transition probabilities.
Let (eaj ) and (ebi ) be two orthonormal bases consisting of eigenvectors of the
operators a and b. Thus aeaj = αj eaj and
bebi = βi ebi . By the projection postulate
4.1 Violation of Formula of Total Probability and Non-Kolmogorov Probability Theory 65
(Sect. 2.2.2) after measurement of b with the result b = βi the initial state ψ of
a quantum system is projected onto the eigenstate ebi of b corresponding to the
eigenvalue βi . Now consider this eigenstate as the initial state and perform the a-
measurement. Suppose that a = αj . Again by the projection postulate it is projected
onto the eigenstate eaj of
a corresponding to its eigenvalue αj . This process can be
formally considered as transition from the input-state ebi to the output-state eaj . The
probability of such transition is equal to pji = P(a = αj |b = βi ).
We can express transition probabilities via eigenvectors of a and
b. Since for
the successive a-measurement the output of the previous b-measurement, namely,
ψout = ebi , is considered as the input state, we can apply the basic probabilistic
rule of quantum mechanics, Born’s rule, see the Eq. (2.33), in the following form:
pji = |ψout , eaj |2 = |ebi , eaj |2 . Hence, pj1 + pj2 = |eb1 , eaj |2 + |eb2 , eaj |2 =
eaj , eaj = 1.
The same terminology is used in the theory of Markov chains. However, in this
theory states are not given by normalized vectors of complex Hilbert space and
probabilities are not defined with the aid of Born’s rule (thus there is no reason for
their double stochasticity).
μ(A ∩ B)
.
μ(B)
FρF
ρF = .
trρF
66 4 Adaptive Dynamics and General Approach to Non-Kolmogorov Probability Theory
trFρFE trρFEF
trρF E = = . (4.18)
trρF trρF
ϕ(E ∧ F)
ϕ(E ∧ F) and , (4.19)
ϕ(F)
where ϕ is a state (a measure) and ∧ is the meet of two events (projections) corre-
sponding to ∩ in CP, and for the state describing by a density operator, we have
ϕ(·) = trρ(·).
We ask when the above two expressions (4.18) and (4.19) in QP are equivalent.
From the next proposition, ϕ(· ∧ F)/ϕ(F) is not a probability measure on P(H ).
Proposition 4.1 (1) When E commutes with F, i.e. EF = FE, the above two expres-
sions are equivalent, namely,
ϕ(FEF) ϕ(E ∧ F)
= .
ϕ(F) ϕ(F)
ϕ (E ∧ F) ϕ (FEF) ϕ (EF)
= = .
ϕ (F) ϕ (F) ϕ (F)
Therefore,
Kϕ Px ∨ Py | Pz = Kϕ (Px | Pz ) + Kϕ Py | Pz
In CP, the joint distribution for two random variables f and g is expressed as
μf ,g (Δ1 , Δ2 ) = μ f −1 (Δ1 ) ∩ g−1 (Δ2 )
for any Borel sets Δ1 , Δ2 ∈ B (R). The corresponding quantum expression is either
for two observables A, B and their spectral measures EA (·), EB (·) such that
It is easily checked that the above expressions do not satisfy either the condition of
probability measure or the marginal condition unless AB = BA, so that they can not
be the joint quantum probability in the classical sense.
Let us explain the above situation, as an example, in a physical measurement
process. When an observable A has a discrete decomposition like
A= ak Fk , Fi ⊥Fj (i = j),
k
pk = trρFk
Fk ρFk
ρk = .
trρFk
but not
pjk = trρEj
k
68 4 Adaptive Dynamics and General Approach to Non-Kolmogorov Probability Theory
Therefore we conclude that in quantum system the above two candidates cannot sat-
isfy the properties of both conditional and joint probabilities in the sense of classical
system.
The above discussion shows that the order of the measurements of two observ-
ables A and B is essential and it gives us a different expectation value, hence the
state changes.
As we have seen, in the two slit experiment the basic law of classical probability
theory, FTP, is violated. Thus we have to apply another probability model, different
from the conventional Kolmogorov model.
As was pointed out in the preface, the situation in probability theory can be
compared with the situation in geometry. We know well that, besides the Euclidean
geometry, there exist various non-Euclidean geometries. The first non-Euclidean
model was proposed by Lobachevsky; it is also known as the hyperbolic geometry.
We recall the parallel postulate of the Euclidean geometry. By this postulate in two-
dimensional Euclidean space, for any line L and point P which does not belong to
L , there exists precisely one line going through the point P that does not intersect
L.; i.e., that is parallel to R. In hyperbolic geometry there exist at least two distinct
lines going through the point P which do not intersect L. Thus the parallel postulate
is violated. We also remark that in general we can make the sum of inner angles
of triangle as we like. For instance, this sum can be less than 180◦ or more than
180. Therefore we have many different non-Euclidean geometries. This entails the
non-Kolmogorov probability theory versus the Kolmogorov probability theory, that
is, we have many different non-Kolmogorov probability theories.
Such non-Kolmogorov probability models will be discussed in the following chap-
ters. The dynamics in non-Kolmogorov probability model can be mathematically
described by the adaptive dynamics discussed in the Sect. 4.4.
In this section, we present the notions of lifting [13] and channel which are advanced
tools of quantum information [14]. They will play a basic role in AD, see Sect. 2.6 for
the preliminary discussion. For simplicity, we proceed in the finite dimensional case.
4.2 Lifting and Channel 69
Let H be a complex Hilbert space. Denote the spaces of linear operators acting
in this space by the symbol O(H ). (Thus A : H → H , linearity means that
A transforms linear combinations of vectors into linear combinations. By fixing an
orthonormal basis in H we can represent it as Cn and O(H ) as the space of all n×n
complex matrices. However, we prefer to use the operator terminology, which can
be applied even to infinite dimensional Hilbert spaces.)2 The space of all quantum
states, i.e., density operators in H , see Sect. 2.2.1 for definition, is denoted by the
symbol S (H ).
Let H and K be two complex Hilbert spaces and let H ⊗ K be their tensor
product (Sect. 2.5). In many applications the first Hilbert space is the state space of
a system under investigation, the second represents states of an environment (bath)
and the third is the state space of the compound system. However, for a moment, we
do not make such a specification and just consider these spaces as state spaces of two
systems (physical, biological, social,…) and the compound system formed of them.
We shall also use O(H ⊗ K ) as the space of observables (see the above footnote)
for the compound system.
Definition Quantum-like lifting is a map
E ∗ : S (H ) → S (H ⊗ K ).
where I ≡ IH is the unit map in H and V ∗ denotes the adjoint operator (Sect. 2.3,
see (2.49)). Then the map
E ∗ ρ = V ρV ∗
is a lifting. Liftings of this type are called isometric. Every isometric lifting is a
pure lifting.
θt ≡ Ut (ρ1 ⊗ ρ2 )Ut∗ ,
Example 4.3 The projection quantum measurement process is written, in the termi-
nology of lifting, as follows. Any observable can be represented as A = n an πn ,
where πn , n = 1, . . . , m, are the projectors on the eigensubspaces corresponding to
the eigenvalues an (we state again that we work in the finite dimensional case). In
accordance with the projection postulate (in the Lüders form, see Sect. 2.2.2, (2.36))
the postmeasurement state (if the result of measurement is not specified) will be
∗
ρ= πn ρπn ,
n
where ρ is the input state. This is a channel. Now we fix a discrete probability μ. It is
given by a sequence of weights, say μn ≥ 0, n μn = 1. The lifting E ∗ associated
with this channel and with a fixed decomposition of ρ as
ρ= μn ρn
n
4.2 Lifting and Channel 71
where ρn ∈ S (H ) is given by
E ∗ρ = μn ρn ⊗ ∗
ρn .
n
The adaptive dynamics has two aspects, one of which is the “observable-adaptivity”
and another is the “state-adaptivity”.
The idea of observable-adaptivity comes from the papers [15–17] studying chaos.
We claimed that the description of real observations does not match completely with
mathematical universalities such as taking limits, sup, inf,etc. Observation of chaos
is the result of taking into account suitable scales of, for example, time and space. In
the limiting case such an observation is impossible, i.e., chaos will disappear.
The idea of the state-adaptivity is implicitly started in the construction of a com-
pound state for quantum communication [14, 18–20] and in Accardi’s Chameleon
dynamics [21].3 This adaptivity can be used to solve a problem pending for more
than 30 years: whether there exists an algorithm solving a NP complete problem
in polynomial time. We found such algorithms firstly by quantum chaos algorithm
[22, 23] and secondly by the adaptive dynamics [13] based on quantum algorithm
of the SAT [24, 25].
3 This dynamics provides a possibility to express the essentials of quantum measurements to adap-
tivity. In particular, Accardi was able to model violation of the famous Bell inequality with the aid
of this dynamics.
72 4 Adaptive Dynamics and General Approach to Non-Kolmogorov Probability Theory
In the Chap. 1 we discussed the intuitive meaning of the adaptive dynamics, and we
pointed out that there exist two types of adaptive dynamics; the observable-adaptive
dynamics and the state-adaptive dynamics.
The observable-adaptive dynamics is a dynamics characterized as follows:
(1) Measurement depends on how to see an observable to be measured.
(2) The interaction between two systems depends on how a fixed observable exists,
that is, the interaction is related to some aspects of observables to be measured
or prepared.
The state-adaptive dynamics is a dynamics characterized as follows:
(1) Measurement depends on how the state and observable to be used exist.
(2) The correlation between two systems depends on how the state of at least one of
the systems exists, e.g., the interaction Hamiltonian depends on the state.
In Sect. 1.4, we explained the adaptivity of biological systems, and introduced the
concept of the adaptive dynamics. In this section, we will discuss how to use the
concept of lifting to explain phenomena breaking the usual probability law.
The adaptive dynamics implies that the dynamics of a state or an observable
after an instant (say the time t0 ) attached to a system of interest is affected by the
existence of some other observable and state at that instant. Let ρ ∈ S (H ) and
A ∈ O (H ) be a state and an observable before t0 , and let σ ∈ S (H ⊗ K ) and
Q ∈ O (H ⊗ K ) be a state and an observable to give an effect to the state ρ and the
observable A. In many cases, the effect to the state is dual to that to the observable,
so that we will only discuss the effect to the state. This effect is described by a lifting
Eσ∗Q , so that the state ρ becomes Eσ∗Q ρ first, then it will be trK Eσ∗Q ρ ≡ ρσ Q . The
adaptive dynamics here is the whole process such as
That is, what we need is how to construct the lifting for each problem to be studied,
that is, we properly construct the lifting Eσ∗Q by choosing σ and Q properly. The
expectation value of another observable B ∈ O (H ) in the adaptive state ρσ Q is
trρσ Q B = tr H tr K B ⊗ IEσ∗Q ρ.
A = {ak ∈ R, Fk ∈ P (H )} ,
B = bj ∈ R, Ej ∈ P (K ) ,
where we do not assume Fk , Ej are projections, but they satisfy the conditions
k Fk = I, j Ej = I as positive operator valued measure (POVM) corresponding
4.4 State and Observable Adaptive Dynamics 73
where is a certain operation (relation) between A and B, more generally, one can
take a certain operator function f (Ej , Fk ) instead of Ej Fk . If σ, Q are independent
from any Fk , Ej and the operation is the usual tensor product ⊗ so that A and
B can be considered in two independent systems or to be commutative, then the
above “joint-like” probability becomes the joint probability. However, if this is not
the case, e.g., Q is related to A and B, the situation will be more subtle. Therefore, the
problem is how to set the operation and how to construct the lifting Eσ∗Q in order
to describe the particular problems associated to systems of interest. In the sequel,
we will discuss this problem in the context dependent systems like bio-systems and
psyco-systems mentioned in Introduction. That is, we discuss how to apply formula
(4.22) to the following three problems breaking the usual probability law: (1) State
change of tongue for sweetness, (2) Lactose-glucose interference in E. coli growth,
(3) Updating the Bayesian law.
References
1. Feynman, R., Hibbs, A.: Quantum Mechanics and Path Integrals. McGraw-Hill, New York
(1965)
2. Khrennikov, A.: Interpretations of Probability, 2nd edn. De Gruyter, Berlin (2009)
3. Khrennikov, A.: Contextual Approach to Quantum Formalism. Springer, Berlin (2009)
4. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Yamato, I.: Non-Kolmogorovian approach
to the context-dependent systems breaking the classical probability law. Found. Phys. 43(7),
895–911 (2013)
5. Conte, E., Todarello, O., Federici, A., Vitiello, F., Lopane, M. and Khrennikov, A.: A pre-
liminary evidence of quantum-like behaviour in measurements of mental states. In: Quantum
Theory: Reconsideration of Foundations. Series Mathematical Modelling in Physics, Engi-
neering and Cognitive Sciences, vol. 10, pp. 679–702. Växjö University Press, Växjö (2004)
6. Conte, E., Todarello, O., Federici, A., Vitiello, F., Lopane, M., Khrennikov, A., Zbilut, J.P.:
Some remarks on an experiment suggesting quantum-like behavior of cognitive entities and
formulation of an abstract quantum mechanical formalism to describe cognitive entity and its
dynamics. Chaos, Solitons Fractals 31(5), 1076–1088 (2007)
7. Conte, E., Khrennikov, A., Todarello, O., Federici, A., Zbilut, J.P.: A preliminary experimental
verification on the possibility of Bell inequality violation in mental states. Neuroquantology 6,
214–221 (2008)
8. Conte, E., Khrennikov, A., Todarello, O., Federici, A., Zbilut, J.P.: Mental states follow quantum
mechanics during perception and cognition of ambiguous figures. Open Syst. Inform. Dyn. 16,
1–17 (2009)
9. Khrennikov, A.: Ubiquitous Quantum Structure: From Psychology to Finance. Springer, Hei-
delberg (2010)
4 See Chap. 10 for the detailed presentation of theory of POVMs and their applications in cognitive
science and decision making.
74 4 Adaptive Dynamics and General Approach to Non-Kolmogorov Probability Theory
10. Tversky, A., Shafir, E.: The disjunction effect in choice under uncertainty. Psychol. Sci. 3,
305–309 (1992)
11. Shafir, E., Tversky, A.: Thinking through uncertainty: nonconsequential reasoning and choice.
Cogn. Psychol. 24, 449–474 (1992)
12. Croson, R.: The disjunction effect and reasoning-based choice in games. Org. Behav. Hum.
Decis. Process. 80, 118–133 (1999)
13. Accardi, L., Ohya, M.: A Stochastic limit approach to the SAT problem. Open Syst. Inf. Dyn.
11, 1–16 (2004)
14. Accardi, L., Ohya, M.: Compound channels, transition expectations, and liftings. Appl. Math.
Optim. 39, 33–59 (1999)
15. Kossakowski, A., Ohya, M., Togawa, Y.: How can we observe and describe chaos? Open Syst.
Inf. Dyn. 10(3), 221–233 (2003)
16. Ohya, M.: Foundation of chaos through observation. In: Hida, T., Saito, K., Si, S. (eds.) Quantum
Information and Complexity, pp. 391–410 (2004)
17. Ohya, M.: Adaptive dynamics in quantum information and chaos. In: Hida, T. (ed.) Stochastic
Analysis: Classical and Quantum, pp. 127–142 (2005)
18. Ohya, M.: On compound state and mutual information in quantum information theory. IEEE
Trans. Inf. Theory 29, 770–777 (1983)
19. Ohya, M.: Some aspects of quantum information theory and their applications to irreversible
processes. Rep. Math. Phys. 27, 19–47 (1989)
20. Ohya, M.: Note on quantum probability. Il Nuovo Cimento 38, 203–206 (1983)
21. Accardi, L.: Urne e camaleonti: Dialogo sulla realta, le leggi del caso e la teoria quantistica.Il
Saggiatore, Rome (1997); English edn., World Scientific, Singapore (2002); Japanese edition,
Makino, Tokyo (2002); Russian edn., Regular and Chaotic dynamics, Ijevsk (2002)
22. Ohya, M., Volovich, I.V.: New quantum algorithm for studying NP-complete problems. Rep.
Math. Phys. 52, 25–33 (2003)
23. Ohya, M., Volovich, I.V.: Mathematical Foundations of Quantum Information and Computation
and Its Applications to Nano- and Bio-systems. Springer, Berlin (2011)
24. Ohya, M., Masuda, N.: NP problem in quantum algorithm. Open Syst. Inf. Dyn. 7(1), 33–39
(2000)
25. Accardi, L., Sabbadini, R.: On the Ohya-Masuda quantum SAT Algorithm. In: Proceedings
International Conference UMC’01. Springer (2001)
Chapter 5
Application of Adaptive Dynamics to Biology
In this chapter, we illustrate by a few examples of cell and protein behavior that
microscopic bio-systems can exhibit complex probabilistic behavior, which can-
not be described by classical probabilistic laws. These examples support authors’
conjecture that the behavior of microscopic bio-systems have to be described by QL-
models, i.e., models based on the formalism of quantum-mechanics, see Chap. 1. As
is emphasized in this chapter, we do not couple QL-behavior with quantum physical
processes in bio-systems. We present arguments that such a behavior can be induced
by the information complexity of even smallest bio-systems, by their adaptivity
to context changes.1 Although our examples of the QL-behavior are rather simple,
lactose-glucose interference in E. coli growth, the interference effect for the differen-
tiation of a tooth stem cell induced by the presence of mesenchymal cells, interference
in behavior of PrPC and PrPSc prions, these examples can stimulate the interest to
QL-models of adaptive dynamicsand lead to more complex examples of nonclassical
probabilistic behavior in cellular biology.
We concentrate our modeling mainly on the glucose effect on E. coli growth. This
is a basic example of destructive interference in cell’s activity, the interference of
1 It is well known that the behavior of each cell is characterized by huge complexity. It can be
compared with an actor who is playing simultaneously at a few different scenes participating in
different (sometimes mutually incompatible) performances. Even smaller bio-systems, such as
viruses and even proteins, exhibit complex behavior, which is characterized by nontrivial context-
dependence and adaptivity to context variations.
© Springer Science+Business Media Dordrecht 2015 75
M. Asano et al., Quantum Adaptivity in Biology: From Genetics to Cognition,
DOI 10.1007/978-94-017-9819-8_5
76 5 Application of Adaptive Dynamics to Biology
two factors: the presence of lactose and glucose in a E. coli cell.2 One of the reasons
for this is the possibility to find quantitative statistical data for varying experimental
contexts, the presence of only lactose, only glucose, and both lactose and glucose,3
see, e.g., Inada et al. [2].
We shall use the contextual viewpoint to conditioning and FTP (the formula of total
probability) presented in Sect. 4.1.1.
The first problem under investigation is not so sophisticated, but quite common. It
has a heuristically simple contextual structure. We consider the following cognitive
experiment.
One takes sugar S or (and) chocolate C and he is asked whether it is sweet or not.
The answers “yes” and “no” are numerically encoded by 1 and 2. Then the basic
classical probability law need not be satisfied, that is,
because the LHS P(C = 1) will be very close to 1, but the RHS will be less than 21 .
Note that the LHS P(C = 1) is obtained in the context that subjects do not taste sugar;
they start directly with chocolate. The contexts “a tongue tasted sugar”, say Ssug , and
“a tongue did not taste sugar”, say S¬sug , are different. The probabilities in LHS and
RHS in the above equation should be replaced by PS¬sug (C = 1) and PSsug (C = 1).
The problem to be discussed is how to obtain these probabilities mathematically.
2 Since the operon theory was proposed in 1956–1961 by Jacob and Monod [1], the regulatory system
of gene expression of lactose operon has been extensively studied and the molecular mechanism
of it has been mostly elucidated including the catabolite repression phenomenon. The description
of the molecular biological mechanism so far obtained and also the systems biology approach
(reductionism) are presented in Sects. 3.3.1 and 3.3.2.
3 It is clear that this contextual structure is isomorphic to the contextual structure of the two slit
Our considerations are based on an article reporting the glucose effect on E. coli
(Escherichia coli) growth [2], where the β-galactosidase activity at certain growth
phase was measured under the following experimental contexts:
C : 0.4 % lactose + 0.1 % glucose,
CL : 0.4 % lactose (and no glucose),
CG : 0.4 % glucose (and no lactose).
The activity is represented in Miller units (enzyme activity measurement condi-
tion). Then one can obtain the probabilistic data: 0.4 % lactose +0.1 % glucose, 43
units; 0.4 % lactose, 2920 units; 0.4 % glucose, 33 units.
We recall that by full induction, the activity reaches to 3000 units. We want to
represent these data in the form of contextual probabilities and put them into the
formula of total probability. We introduce the observable b = ±1, which describes
the β-galactosidase activity. We also introduce the observable a = L, G describing
the detection of molecules of lactose and glucose. The experimental data provide the
contextual (conditional) probabilities:
43
P(b = +1|C) = ≈ 0.014,
3000
2920 33
P(b = +1|CL ) = ≈ 0.973, P(b = +1|CG ) = ≈ 0.011.
3000 3000
We also determine the probabilities of “realizations of contexts CL and CG in the
context C”:
0.4 0.1
P(a = L|C) = = 0.8, P(a = G|C) = = 0.2.
0.5 0.5
In the classical probabilistic framework we would obtain the equality (FTP):
Thus classical FTP is violated. A cell does not split a context C into two disjoint
and not interfering contexts CL and CG . It treats C (which is physically just the union
of two ensembles of molecules, lactose and glucose)4 as a new environment.
4 We ignore the presence of other types of molecules, which are not essential for this type of gene
expressions.
78 5 Application of Adaptive Dynamics to Biology
i.e.,
λ+ = −1.74. (5.3)
This coefficient is larger than one. Since in any way it is nonzero, we cannot proceed
with Kolmogorov’s model. Since it is larger than one, we cannot proceed with the
standard quantum model (see discussion in Sect. 4.1.1). We shall come back to the
problem of interference in E. coli metabolism in Sect. 5.2.2, where we shall use the
generalized quantum calculus generated by QL adaptive dynamics.
Another source of impossibility to apply the standard quantum formalism based
on the representation of observables by Hermitian operators is violation of the
condition of double stochasticity for contextual probabilities. Contrary to (4.16),
P(b = +1|CL ) + P(a = G|C)P(b = +1|CG ) = 1.
Remark We recall that lactose induces the enzyme, but without induction certain
percentage would be expressed by fluctuation of gene expression. The concentration
of glucose is not important, but the following should be taken into account:
If we add 0.2 % glucose in the medium, cells can grow to its stationary phase only
on glucose and they do not try to utilize lactose. Thus, if we want to see the enzyme
induction during the growth, we have to limit the glucose concentration, usually
0.02 %. That amount is insufficient for the support of cell growth and cells try to
utilize lactose after the consumption of glucose. If we add only 0.02 % glucose in
the medium without any other carbon (energy) source, then the enzyme level would
be similar as in the presence of 0.2 % glucose and cells stop growing. If there is any
other carbon source than 0.02 % glucose, then cells continue to grow and the enzyme
level changes depending on the kind of carbon source (for lactose, the level is quite
high; for maltose, the level would be low, but significant; for pepton (amino acid
mixture), the level would be a little bit more).
The mathematically oriented reader may be disappointed that the concentration
of glucose under the context C, namely, 0.1 %, does not match with its concentration
under context the CG , namely, 0.4 %. However, by taking into account the statement
in the second sentence of the above remark we can assume that the level of activation
of enzyme would be approximately the same for the experimental context
C : 0.4 % lactose + 0.4 % glucose.
For this context, we can also check FTP under the assumption
43
P(b = +1|C ) ≈ P(b = +1|C) = ≈ 0.014.
3000
5.1 Violation of the Formula of Total Probability in Molecular Biology 79
Here,
P(a = L|C ) = P(a = G|C ) = 0.5.
Hence,
λ+ = −2.282.
It is again larger than one and we have to use a generalization of the standard quantum
formalism.
It is well known that a dental epithelial cell grows in a medium as it is (no differ-
entiation). A dental mesenchymal cell grows similarly. However, if they are grown
together (mix), then a dental epithelial cell differentiates to a tooth cell [3]. The
growth condition is the same. Hence, growing independently, they grow as they are
(no differentiation). If they grow mixed, dental epithelial cells differentiate.
We introduce a random variable b = +1, in the case of differentiation of a cell,
and otherwise b = −1. We also introduce three contexts:
C : the presence of mesenchymal cells and tooth stem cells;
CM : the presence of only mesenchymal cells;
CST : the presence of only tooth stem cells.
Although we do not have the complete statistical data (and this is a general prob-
lem in cellular biology), we can assume that P(b = +1|CM ) = εM 1 (the
probability of differentiation in the ensemble consisting of only mesenchymal cells)
and P(b = +1|CST ) = εST 1 (the probability of differentiation in the ensemble
consisting of only tooth stem cells) and also that P(b = +1|C) ≈ 1 (the probability
of differentiation in a mixture of mesenchymal cells and tooth stem cells). We now
introduce another variable a describing the detection of mesenchymal cell, a = M,
and tooth stem cell, a = ST . Now we can take, e.g., a sample of cells with equal
concentrations of mesenchymal cells and tooth stem cells, i.e.,
Then the classical FTP is violated and the interference coefficient λ+ is very large.
We can also mention an experiment on the differentiation of PL1+ cells [4]. This
experiment was performed in the study of new methods of identification of cancer
cells. A population of PL1+ cells was identified by exposing it under two comple-
mentary contexts: (a) differentiation context (serum rich media) and (b) stem cells
culture context. We can use this experiment as an additional example of destructive
interference.
80 5 Application of Adaptive Dynamics to Biology
In spite of its small size, a cell is still an extremely complex biological system. Now
we are looking for interference effects and violation of laws of classical probability
theory for essentially simpler systems, namely, proteins (which are, although very
small and less complex as compared with cells, are still very large and complex at
the molecular scale).
Let us consider prion protein (cause of CJD, mad cow disease , see, e.g., [5]).
Usually it stays as globular soluble form in mono-dispersed state; this is safe confor-
mation. On heating, it may denature and aggregate taking random conformations. If
there exit seeds for β-form conformation, a mutant prion protein, or wild type prion
protein (but after prolonged incubation), they tend to make filamentous aggregate
(this is the cause of CJD) as a whole, which is resistant to solubilization with deter-
gent and to denaturation by heat. Thus, wild type prion from a normal animal can be
incubated for a long time as a clear solution. But by adding mutant prion or rather
β-form wild type prion obtained by incubating the wild type prion solution for a very
long time, most of the prion proteins become insoluble as filamentous aggregates.
After this, the prion proteins can no more resume the initial soluble conformation.
We introduce a random variable b = +1, if protein can be incubated for a long
time as a clear solution, and otherwise b = −1. We also introduce three contexts:
C : the presence of both types of proteins;
CM : the presence of only mutant prion proteins PrPSc ;
CPP : the presence of only wild type prion proteins PrPC .
Although we again have no sufficiently statistical experimental data, we can
assume that P(b = +1|CPP ) ≈ 1 − δPP and P(b = +1|CM ) ≈ δM and also
that P(b = +1|C) = δ, where all δs are very small.
We also introduce the observable a of detection of protein type, a = M, PP. Now
we can take, e.g., a sample of proteins of both types with equal concentrations, i.e.,
We presented some examples of biological phenomena which violate the usual prob-
ability law. These violations are of the contextual nature. In this section, we discuss
these context dependences from the point of view of adaptive dynamics.
5.2 Interpretation of the Above Violations by Adaptive Dynamics 81
where x0 = √1 (|e1 + |e2 ) . Here the neutral pure state ρ describes the state of
2
tongue before the experiment, and we start from this state.
When one tastes sugar, the state of tongue will be changed. Such change can be
written as the operator (in this case, two-dimensional matrix) on the Hilbert space.
The operator corresponding to tasting sugar is mathematically (and operationally)
represented as
λ1 0
S= ,
0 λ2
where λi are complex numbers satisfying |λ1 |2 + |λ2 |2 = 1. This operator can be
regarded as the square root of the sugar state σS :
Taking sugar, he will taste that it is sweet with the probability |λ1 |2 and non-sweet
with the probability |λ2 |2 , so |λ1 |2 should be much higher than |λ2 |2 for usual sugar.
This comes from the following change of the neutral initial (i.e., non-adaptive) state
of a tongue:
S ∗ ρS
ρ → ρS = ΦS (ρ) ≡ . (5.4)
tr |S|2 ρ
E1 ρS E1 + E2 ρS E2 .
This is the unread state (as in quantum measurement, see Sect. 2.2.2). We can use
the above two expressions, which give us the same result for computation of the
probabilities for the S-variable.
However, for a while the tongue becomes dull to sweetness (and this is the crucial
point of our approach for this example), so the tongue state can be written by means
82 5 Application of Adaptive Dynamics to Biology
01
of a certain “exchanging” operator X = such that
10
ρSX = XρS X.
C ∗ ρSX C
ρSXC = ΦC (ρSX ) ≡ ,
tr |C|2 ρSX
with |μ1 |2 + |μ2 |2 = 1. Common experience tells us that |λ1 |2 ≥ |μ1 |2 ≥ |μ2 |2 ≥
|λ2 |2 and the first two quantities are much larger than the last two quantities.
As can be seen from the preceding consideration, in this example the “adaptive
set” {σ, Q} is the set {S, X, C}. Now we introduce the following (nonlinear) lifting:
P(S = j, C = k) = tr Ej ⊗ Ek Lσ Q (ρ).
The probability that one tastes sweetness of the chocolate immediately after tasting
sugar is
|λ2 |2 |μ1 |2
P(S = 1, C = 1) + P(S = 2, C = 1) = ,
|λ2 |2 |μ1 |2 + |λ1 |2 |μ2 |2
which is PSsug (C = 1). Note that this probability is much less than
which is the probability of sweetness tasted by the neutral tongue ρ. This means that
the usual formula of total probability should be replaced by the adaptive (context
dependent) probability law.
5.2 Interpretation of the Above Violations by Adaptive Dynamics 83
2920
CL : P(+|L) = ,
3000
33
CG : P(+|G) = . (5.5)
3000
We set P(+) ≡ P(+|C); thus, see Sect. 5.1.2, from the experimental data we obtain
43
P(+) = ,
3000
We now use our mathematical model for computation of the above probabilities,
by using the concept of lifting. First, we introduce the initial state ρ = |x0 x0 | in
the Hilbert space H = C2 . The state vector x0 is written as
1 1
|x0 = √ |e1 + √ |e2 .
2 2
The basis {e1 , e2 } denote the detection of lactose or glucose by E. coli, i.e., the events
L and G. In the initial state ρ, the E. coli has not recognized the existence of lactose
and glucose yet. When E. coli recognizes them, the following state change occurs:
DρD∗
ρ → ρD = ΦD (ρ) ≡ ,
tr (|D|2 ρ)
where
α 0
D=
0β
84 5 Application of Adaptive Dynamics to Biology
with |α|2 + |β|2 = 1. Note that |α|2 and |β|2 give the probabilities of the events L
and G: P(L) and P(G). The state σD ≡ DD∗ encodes the probability distribution
P(L), P(G):
σD = P(L) |e1 e1 | + P(G) |e2 e2 | .
In this sense, the state σD represents the chemical solution of lactose and glucose.
We call D and ρD the detection operator and detection state, respectively. The state
determining the activation of the operon in E. coli depends on the detection state
ρD . In our operational model, this state is obtained as the result of the following
transformation:
QρD Q∗
ρDQ = ΦQ (ρD ) ≡ ,
tr (QρQ∗ )
We call ρDQ and Q the activation state for the operon and activation operator,
respectively. (The components a, b, c, and d will be discussed later.)
We introduce lifting
by which we can describe the correlation between the activity of lactose operon and
the ratio of concentration of lactose and glucose. From the discussion in Sect. 4.4,
the joint probabilities PDQ (L, +) and PDQ (G, +) are given by
The probability PDQ (±) is obtained as PDQ (L, ±) + PDQ (G, ±).
Let us consider a context CL such that the detection operator D satisfies the
condition P(L) = |α|2 = 1. We denote such D by the symbol DL . The probabilities
PDL Q (±) = PCL (±)(≡ P(±|CL )) are calculated as
|a|2 |c|2
PCL (+) = , PC (−) = .
|a|2 + |c|2 L
|a|2 + |c|2
Here, kL is a certain real number. In a similar way, we consider the context CG and
obtain
33 iθ+G 2967 iθ−G
b= e kG , d = e kG
3000 3000
In general, the value of this probability is different from that of PCL (+)|α|2 +
PCG (+)|β|2 . The rate |k̃L |/|k̃G | essentially determines the degree of the difference.
Recall the experimental data context C. In this case, P(L) = |α|2 = 0.8 > P(G) =
|β|2 = 0.2, but PC (+) is very small. According to our interpretation, this
implies that
k̃L 0
the rate |k̃L |/|k̃G | is very small. In this sense, the operator F = in Eq. (5.7)
0 k̃G
specifies the preference in E. coli’s metabolism. We call F the preference operator.
Finally note that if α, β are real and k̃L = k̃G ∗ , the formula of total probability holds.
In the previous section, we introduced the E. coli’s preference {kL /kG ,θ } based on
the adaptive dynamics. In this section, we estimate the preference for a different
type of strains. The calculation of preferences is similar to the calculation in the
previous section.
Let us explain our experiment to obtain the probabilities P(+). By the method of β-
galactosidase assay [6], we measured the values of Miller unit (MU) in five different
situations, see Table 5.1. E. coli is grown in the media containing (1) 0.4 % lactose; (2)
0.4 % glucose;(3) 0.2 mM IPTG (Isopropyl β-D-1-thiogalactopyranoside); (4) 0.1 %
glucose and 0.4 % lactose; (5) 0.4 % glucoseand 0.4 % lactose. In this experiment,
86 5 Application of Adaptive Dynamics to Biology
we used four types of E. coli [7]; W3110, ML30, ML308 (lacI), ML308-2 (lacI,
lacY ). We also used two types of preculture condition; grown in minimal medium
with 0.4 % glycerol(Gly) or in Luria broth(LB).
The strains of W3110 and ML30 are wild type E. coli, which show stronger
preference for glucose than for lactose. In fact, as shown in data A and C, both
MU(4) and MU(5) have small values similar to MU(2). On the other hand, as seen
in data D, the strain ML308 has weak preference for glucose since MU(2), MU(4),
and MU(5) are large. ML308 is a mutant of repressor minus (lacI − ), so that the
β-galactosidase is produced in the cell even in the presence of high concentration of
glucose. As seen in data E, ML308-2 shows different behavior from other strains;
MU(4) and MU(5) are smaller than MU(2). ML308-2 is a mutant, which is not
only repressor minus but also lactose transport system minus (lacY − ). Table 5.1 also
shows the result for two different preculture conditions as seen in data A and B. We
clearly see that the value of MU(4) and MU(5) in data B is very high in comparison
to those in data A. This result shows that E. coli’s preferences are affected by the
preculture condition.
Also we can calculate the conditional probabilities P(±|L), P(±|G), P(4) (+) and
P(5) (+) from the obtained MU values (3)–(5). For data C, we can calculate them
as follows.
1763 14
P(+|L) = , P(+|G) = ,
3133 3133
184 78
P(4) (+) = , P(5) (+) = .
3133 3133
The Table 5.2 shows the probabilities for data (A)–(E).
One can confirm that the probabilities in data A and C do not satisfy FTP;
On the other hand, one can see that the probabilities in data (D) satisfy FTP approxi-
mately. The strain of ML308 is a mutant, which can not produce the repressor protein.
Therefore, β-galactosidase is always produced in the cell of ML308. ML308-2 is
not only lacI − but also lacY − , so that the lactose transporting system of ML308-2 is
defective and different from that of other strains. Hence the violation of FTP is also
seen in data (E).
Here, let us recall Eq. (5.8) in order to calculate lactose/glucose preferences
{kL , kG , θ }. By assigning the values of probabilities P(+|L), P(+|G), P(4) (L),
P(5) (L), P(4) (+) and P(5) (+) to Eq. (5.8), we have two equations:
kL kL
P(+|L) P(5) (L)Δ − P(4) (L)Δ + P(4) (+)Δ − P(5) (+)Δ
kG kG
+ P(+|G) P(5) (G)Δ − P(4) (G)Δ = 0, (5.9)
and
kL kG
P(4) (+)− P(+|L)P(4) (L) kG + P(+|G)P(4) (G) kL
cosθ = , (5.10)
2Δ
where
Δ =P(4) (−) P(+|L)P(+|G)P(4) (L)P(4) (G) + P(4) (+) P(−|L)P(−|G)P(4) (L)P(4) (G),
Δ =P(5) (−) P(+|L)P(+|G)P(5) (L)P(5) (G) + P(5) (+) P(−|L)P(−|G)P(5) (L)P(5) (G).
√
From Eqs. (5.9) and (5.10), we can obtain√ kL /kG and cos θ , and Table 5.3 shows
these values. We can see that the value of √kL /kG is small for data A and C, which
violate FTP. On the other hand, the value of kL /kG is nearly one, and cos θ is nearly
zero in data D, which hold FTP approximately.
88 5 Application of Adaptive Dynamics to Biology
√
We remark that the values of kL /kG , cos θ in data A are different from those in
data B. This result means that
√ the preculture
condition affects E. coli’s preference. By
assigning the preferences kL /kG , θ to Eq. (5.8), we compute P(+) with respect to
P(L), see Fig. 5.1a. One can see that the behavior of lactose operon depends not only
on the strain but on the preculture condition as well; The P(+) of W3110(Gly) in
Fig. 5.1a is increasing more linearly than that of W3110(LB). Therefore, the degree
of violation of FTP for W3110 precultured on glycerol is less than that for W3110
precultured in LB.
For the data C, D, and E, we plot the values of P(+) with respect to P(L) in
Fig. 5.1b. It shows the influence of lacI gene or lacY gene to lactose operon. The
P(+) of ML308, which does not violate FTP so much, is linearly-increasing with
respect to P(L).
(b)
References 89
References
Cognitive psychologists and game theorists have long been aware that it is impossible
to use classical (Kolmogorov) probability to describe statistical data collected, e.g.,
in games of the Prisoners Dilemma type. However, they selected another pathway
for the departure from Kolmogorovness, not through probabilistic manifolds with
charts of the Kolmogorov type, but through violating the basic axiom of the Kol-
mogorov’s axiomatics of probability theory—additivity of probability. They used a
subadditive generalization of a probability measure, see, e.g., [1–4] (see also [5–7]
for comparison of nonadditivity and QL approaches). This helped to solve a num-
ber of problems of cognitive psychology and game theory. However, the subaddi-
tive approach does not have such a degree of generality as the quantum formalism.
1 Some of these problems were also studied by Cheon and Takahashi [20, 21], Khrennikov [22, 23],
Fichtner et al. [24], Franco [25], Khrennikov and Haven [26], Pothos and Busemeyer [27, 28],
Lambert-Mogiliansky et al. [29], Dzhafarov and Kujala [30, 31], and Wang et al. [32, 33], see also
the special section on cognition, decision making and finances in the conference proceedings [34].
6.2 Decision Making Process in Game 93
to be chosen and obtain payoffs, which are assigned for the results of their choices.
In the player’s decision making, the following three assumptions are required:
• Players know the rule of the game; each player knows all selectable strategies and
payoffs
• A player behaves rationally so as to maximize his own payoff
• Each rational player recognizes the rationalities of other players.
The goal of game theory is to explain various interdependencies in the real and mental
world, and it is believed that the concept of Nash equilibrium provides a “rational”
solution. However, there are some experiments, in which real players frequently do
not achieve a rational solution, and it seems to be “irrational” [16, 17]. There are
cases where a real player’s decision making process is essentially different from a
“rational” player’s one in conventional game theory.
We study this problem by using the game given by the payoff table of Table 6.1
as the basic example. This is a well-known two-player game called a prisoner’s
dilemma (PD) game. The players A and B have two strategies denoted by 0 A,B and
1 A,B , and the values of payoffs are assigned for the results of (0 A , 0 B ), (0 A , 1 B ),
(1 A , 0 B ) and (1 A , 1 B ). Generally, PD game has a unique Nash equilibrium: Since the
dominant strategies 1 A and 1 B are the best for the rational players A and B, the result
(1 A , 1 B ) is the Nash equilibrium. A player’s decision making process is described as
a combination of rational reasoning of one player and her/his estimation of another
player as rationally reasoning. The diagram in Fig. 6.1 represent the decision making
process of the A-player. The two solid arrows in this diagram express that the player
A prefers the result (1 A , 0 B ) to (0 A , 0 B ) and prefers (1 A , 1 B ) to (0 A , 1 B ). The two
dotted arrows explains that the player A reasons that the player B will prefer (0 A , 1 B )
to (0 A , 0 B ) and (1 A , 1 B ) to (1 A , 0 B ). These arrows represent A’s preference. One
can see that A’s reasoning always leads to the result (1 A , 1 B ), namely, the player A
chooses 1 A believing that the player B will choose 1 B . However, we wonder whether
such a description of the decision making process provides the complete picture of
reasoning of a real player.
Let us consider another PD game represented by the payoff matrix of Table 6.2; In
this game, a real player will choose the strategy 0, because such a player may assume
that the payoff 100000 for the selection (0 A , 0 B ) of the strategies is very attractive
compared with the selection (1 A , 1 B ), and this choice is also attractive for the other
player. Moreover, she/he may reason “that the player B will think in the same way.”
She/He cannot neglect this possibility in her/his decision making, and this opens the
possibility to shift the expectation from (1 A , 1 B ) to (0 A , 0 B ). We express such a shift
as the diagonal arrows in the diagram of Fig. 6.2.
In this situation, player A’s reasoning would be cyclic such as
(0 A , 0 B ) −→ (1 A , 0 B ) −→ (1 A , 1 B ) −→ (0 A , 0 B ) −→ · · ·
Then her/his choices would oscillate between the strategies 1 A and 0 A . However, it
is clear that the intention to choose 0 A is dominant since it comes from the payoff-
difference between 100000 of (0 A , 0 B ) and 3 of (1 A , 1 B ).
The additional shift might seem to be strange. In the framework of game theory,
the player A’s preferences are explained in the following way. “If the player B chosen
x B ∈ {0 B , 1 B }, then I would prefer to choose y A ∈ {0 A , 1 A }”, and the player B’s
preferences reasoned by A are explained in the same form. Such a mental process
consists of two parts; the part of assumption and the part of analysis. In the first part,
the player A assumes one of the player B’s choices. In the second part, the player
A analyzes his preference based on the assumption. Note that such an analysis is
contextually the same as the posterior analysis of someone who already knows the
result of B’s choice. All of the shifts seen in Fig. 6.1 are provided by such posterior
analysis. On the other hand, the additional shift in Fig. 6.2 is never induced by the
posterior analysis, since the player B’s choice is not fixed on this shift.
In principle, each player is not informed about another player’s choice, that
is, each player feels uncertainty about another player’s choice. A prior analysis
should be done in the presence of uncertainty. The prior analysis in the conventional
game theory is just a mixture of the posterior analysis based on a mixed strategy, a
probabilistic distribution given for another player’s choice: The “rational” player A
will consider the posterior analysis of the possibilities for two strategies mixed with
probabilities P(0 B ) and P(1 B ). In each posterior analysis, the player A can estimate
probabilities of his own choices, P(i A |0 B ), (i = 0, 1) and P(i A |1 B ), (i = 0, 1).
For the mixed strategy, the player A makes his own choice with the probability given
by FTP:
P(i A ) = P(i A |0 B )P(0 B ) + P(i A |1 B )P(1 B ).
For example, in PD game the P(1 A ) always becomes 1. As was mentioned above,
the additional shift in Fig. 6.2 is not induced by posterior analysis. Therefore, if the
additional shift affects the decision making, the total probability law will be violated;
Actually, such non-Kolmogorovian feature can be seen in the statistical data obtained
in some experiments, see [16, 17]. The additional effect comes from a “deeper”
uncertainty, where representation of possible actions of another player fluctuates.
Such mental fluctuation cannot be represented in terms of classical probability theory.
Quantum probability is one of the most advanced mathematical models for non-
classical probability, and it is useful to represent the psychological uncertainty, which
a real player feels. In this section, we introduce “a mental state”, which represents
mental fluctuations modelled as quantum fluctuations and describe the process of
decision making as the dynamic of the mental state. This dynamics is also represented
in the quantum framework.
By following the tradition of quantum information theory we call the A-player Alice
and the B-player Bob. We focus on Alice’s reasoning. In principle, Alice does not
know which action Bob will choose, that is, she feels uncertainty about Bob’s action.
We represent such uncertainty with aid of the following state vector:
The orthogonal basis |0 B and |0 B corresponds to Alice’s judgings “Bob will choose
the action 0” and “Bob will choose the action 1”. The structure of the state of |φ B
implies that Alice essentially cannot deny either of the two possibilities. The values
of α and β determine the weights assigned to these possibilities. We call the state
|φ B φ B | ≡ σ B “a prediction state”. (Here we represented the pure state (6.1), a
vector in complex Hilbert space, by the corresponding density operator, see Sect. 2.2.
In our coming modeling it is better to proceed with density operators. The process
of the decision making will be described by the quantum master equation which
96 6 Application to Decision Making Theory and Cognitive Science
transfers pure states given by vectors into in general mixed states given by density
operators.)
Alice will be aware of two consequences of “0 A 0 B ” and “0 A 1 B ” with the weights
(probability amplitudes) of α and β, for her own action “0 A ”. This situation is rep-
resented by the vector
Similarly,
|Φ1 A = |1 A ⊗ |φ B , (6.3)
is given for the action “1 A ”. We call |Φ0 A and |Φ1 A “alternative state vectors”. In
more general forms, the alternative state vectors are given as
|Φ0 A = |0 A ⊗ |φ 0B ,
|Φ1 A = |1 A ⊗ |φ 1B , (6.4)
|φ 0B = α0 |0 B + β0 |1 B ,
|φ 1B = α1 |0 B + β1 |1 B . (|αi |2 + |βi |2 = 1) (6.5)
and it specifies Alice’s mental state such that her action 0 A (or 1 A ) is decided with
the probability tr(Θ|Φ0 A Φ0 A |) ≡ P0 A (or tr(Θ|Φ1 A Φ1 A |) ≡ P1 A ). We call Θ
the “mental state”.
The process of decision making in a game is described as the dynamics of the
mental state Θ.
Θ0 = |Φ(0)Φ(0)|, (6.8)
where
|Φ(0) = x0 |Φ0 A + x1 |Φ1A .
x0 and x1 are complex numbers with |x0 |2 + |x1 |2 = 1. The values of x0 and x1 are
not very important because Alice has not started her thinking for decision making
yet. In general, they are determined randomly with the normalization. To simplify
the discussion, we assume x0 = 1, that is, Θ0 = |Φ0 A Φ0 A |.
{|R, |T } is the orthonormal basis in the space M , and |R, |T represent the dotted
√ The value of Δτ (0 ≤ Δτ ≤ 1) is√the degree of
path and solid path, respectively.
intensity to shift from 0 to 1. ( Δτ is the complex number satisfying | Δτ |2 = Δτ .)
In a similar way, the vector |Φ1→0 , which represents the branching to choice of 1,
is written as
|Φ1→0 = 1 − Δ̃τ |Φ1 A ⊗ |R + Δ̃τ |Φ0 A ⊗ |T
= 1 − Δ̃τ |Φ1 A R + Δ̃τ |Φ0 A T
The parameter Δ̃τ specifies the degree of intensity to shift from 1 to 0. It is assumed
that such branching is occurred in the time interval τ , so Δτ and Δ̃τ are functions of
τ in general.
As the next step, the “branching map” Z τ : H A ⊗ H B → H A ⊗ H B ⊗ M is
introduced. This map generates the state change of branching;
Z τ |Φ0 A = |Φ0→1 ,
Z τ |Φ1 A = |Φ1→0 . (6.9)
Vτ |Φ0 A R = |Φ0→1 ,
Vτ |Φ1 A R = |Φ1→0 . (6.11)
Vτ = e−iλH τ . (6.12)
The operator H is the Hamiltonian of the interaction between the “operative mental
state” and the state of the memory. It represents the generator of Alice’s decision
6.2 Decision Making Process in Game 99
making. In Sect. 7.1.4, we shall discuss the construction of H in detail. The parameter
λ in e−iH λ determines the strength of interaction.
Θ0 → Z τ Θ0 Z τ∗
tr M (Z τ Θ0 Z τ∗ ) ≡ Θ(τ ) = Θ1 .
As was mentioned in the third assumption, in the process of decision making, this
shift is repeated. In other words, the application of Z τ is repeated. In the next time
interval τ , the following state is produced:
(Z τ ⊗ I )Z τ Θ0 Z τ∗ (Z τ∗ ⊗ I ) ∈ S (H A ⊗ H B ⊗ M ⊗ M ).
Here, we note that the new memory space M is added for memorizing the paths of
the new branching. When the initial mental state is given by Θ0 = |Φ0 A Φ0 A |, the
following mental state is produced:
tr M ⊗M ((Z τ ⊗ I )Z τ Θ0 Z τ∗ (Z τ∗ ⊗ I )) = tr M (Z τ Θ1 Z τ∗ ) ≡ Θ(2τ ) = Θ2 .
E j∗ (·) = (Z ⊗ I ⊗ · · · ⊗ I ) · (Z ∗ ⊗ I ⊗ · · · ⊗ I ). (6.14)
100 6 Application to Decision Making Theory and Cognitive Science
The diagonal parts x00 and x11 correspond to the probabilities of choices P0 A and
P1 A . From Eq. (6.13) we can obtain the following results:
Δ̃τ
P0 A (n) = (1 − Δτ − Δ̃τ )n P0 A (0) + (1 − (1 − Δτ − Δ̃τ )n )
Δτ + Δ̃τ
= (1 − ξ )n P0 A (0) + (1 − (1 − ξ )n )P0SA ,
Δτ
P1 A (n) = (1 − Δτ − Δ̃τ )n P1 A (0) + (1 − (1 − Δτ − Δ̃τ )n )
Δτ + Δ̃τ
= (1 − ξ ) P1 A (0) + (1 − (1 − ξ )
n n
)P1SA . (6.15)
Δ̃τ Δτ
Here, Δτ + Δ̃τ = ξ , = P0SA and = P1SA . Noting that 0 ≤ ξ ≤ 2,
Δτ +Δ̃τ Δτ +Δ̃τ
we can find that the probabilities P0 A (n) and P1 A (n) approach the stable values P0SA
and P1SA after a large numbers n of iterations. We also note that the non-diagonal
parts x01 (n) and x10 (n) approach zero. Thus in this model in the process of decision
making the mental state is stabilized to the state represented by the density matrix:
P0SA 0
ΘE = .
0 P1SA
6.2 Decision Making Process in Game 101
Construction of Hamiltonian H
In Eq. (6.11), we introduced the time evolution operator Vτ that plays a main role of
the structure of the branching operator Z τ . Alice’s intentions for reasonings (0 → 1)
and (1 → 0) is encoded in this operator Vτ . In Eq. (6.12), we defined Vτ as the
time evolution operator eiλH τ . In this section, the corresponding Hamiltonian H is
constructed by using an operator that specifies Alice’s so to say “decision making
processor”.
Alice has to compare all potential consequences of the game
{0 A 0 B , 0 A 1 B , 1 A 0 B , 1 A 1 B }.
in the basis {|0 A 0 B , |0 A 1 B , |1 A 0 B , |1 A 1 B }. The elements {μi } and {μ̃ j } are real
numbers. We call this operator the “comparison operator”. For example, when Alice
has the tendency to prefer the consequence 1 A 0 B to 0 A 0 B , its degree of preference is
given by the value of μ1 = 1 A 0 B | G |0 A 0 B . The degree of the opposite tendency
is given by the quantity μ̃1 = 0 A 0 B | G |1 A 0 B . (In general, it is allowed that these
two tendencies coexist in Alice’s mental representation of the game.)
Here, we recall that Alice is not informed about Bob’s choice. Alice’s intention
to choose 0 A always fluctuates between 0 A 0 B and 0 A 1 B . This is a consequence
of the form of the alternative vector |Φ0 A = α0 |0 A 0 B + β0 |0 A 1 B . Similarly,
Alice with the mind to choose 1 A is fluctuated between 1 A 0 B and 1 A 1 B , as seen
in |Φ1 A = α1 |1 A 0 B + β1 |1 A 1 B . We consider the values of Φ1 A |G|Φ0 A and
Φ0 A |G|Φ1 A , which are calculated as
These values μ and μ̃ specify the degrees intentions for transitions (0 → 1) and
(1 → 0).
By using the comparison operator, we define the Hamiltonian H in Eq. (6.12);
H = Φ1 A T Φ1 A T K Φ0 A R Φ0 A R
+ Φ0 A T Φ0 A T K Φ1 A R Φ1 A R
+ Φ0 A R Φ0 A R K ∗ Φ 1 A T Φ 1 A T
+ Φ 1 R Φ1 R K ∗ Φ 0 T Φ 0 T ,
A A A A (6.17)
102 6 Application to Decision Making Theory and Cognitive Science
where
K = G ⊗ |T R| .
1 eiθ
|Ψ± = √ Φ0 A R ± √ Φ1 A T
2 2
μ
are eigenvectors of H for eigenvalues ±|μ| = ±ω (eiθ is equal to |μ| ), and the
vectors
eiθ̃
Ψ = √1 Φ1 R ± √ Φ0 T
± A A
2 2
μ̃
are eigenvectors for eigenvalues ±|μ̃| = ±ω̃ (eiθ̃ = |μ̃| ). Further, one can check that
the time evolution operator Vτ = e −iλH τ transforms Φ0 A R and Φ1 A R as
Vτ Φ0 A R = cos(ωλτ ) Φ0 A R − ieiθ sin(ωλτ ) Φ1 A T ,
Vτ Φ1 A R = cos(ω̃λτ ) Φ1 A R − ieiθ̃ sin(ω̃λτ ) Φ0 A T . (6.19)
Here, the relations Φ0 A R = √1 |Ψ+ + √1 |Ψ− and Φ1 A R = √1 Ψ+ +
2 2 2
√1 Ψ− are used. Since the above results correspond to Vτ Φ0 A R = |Φ0→1 and
2
√
Vτ Φ1 A , R = |Φ1→0 , we give | Δτ | and | Δ̃τ | by
| Δτ | = sin(ωλτ ), | Δ̃τ | = sin(ω̃λτ ).
Setting of Parameters
In this section, numerical analyses of the dynamics of decision making in PD games
are presented. Let us consider a two-player game with the following pay-off table.
A/B 0 B 1 B
0 A a/a b/b
1 A c/c d/d
6.2 Decision Making Process in Game 103
The parameters μi and μ̃i represent the degrees of Alice’s preferences in the
comparison of consequences. Here, it is assumed that the values of (μi2 , μ̃i2 ) ≡
(ki , k̃i ) are determined by the differences of pay-offs: For the comparisons
μ1 μ4
0 A0B 1A0B , 0A1B 1A1B ,
μ̃1 μ̃4
(c − a, 0) if a < c
(k1 , k̃1 ) = ,
(0, a − c) if a > c
(d − b, 0) if b < d
(k4 , k̃4 ) = . (6.20)
(0, b − d) if b > d
μ2 μ3
0A1B 1A0B , 0A0B 1A1B ,
μ̃2 μ̃3
⎧ √
⎪ ( (c − b)(c − b ), 0) if b < c and b < c
⎨ √
(k2 , k̃2 ) = (0, (b − c)(b − c )) if b > c and b > c
⎪
⎩ (0, 0) in other cases
⎧ √
⎨ ( (d
⎪ − a)(d − a ), 0) if a < d and a < d
√
(k3 , k̃3 ) = (0, (a − d)(a − d )) if a > d and a > d . (6.21)
⎪
⎩
(0, 0) in other cases
In the case of PD game (c > a > d > b and b > a > d > c ),
k1 = c − a, k̃3 = (a − d)(a − d )
k4 = d − b,
k̃1 = k2 = k̃2 = k3 = k̃4 = 0. (6.22)
decision making, at n = 0, Alice does not feel the possibility of the irrational choice
“0”, that is, P0 A (n = 0) = 0. At the condition (α02 , α12 ) = (0, 0), Alice judges that
“Bob will choose the rational 1.” She will not feel uncertainty about Bob’s choice and
choose “1” since the pay-off 3 is larger than 2. At the condition (α02 , α12 ) = (1, 0),
Alice judges that “Bob’s strategy will be the same as mine.” Then, Alice will choose
“0” by comparing the pay-offs 4 and 3. As seen in Fig. 6.5, the probability P0 A
develops and approaches 1. At the condition parametrized by (α02 , α12 ) = (0.5, 0.5)
or (0.75, 0.25), Alice feels uncertainty and chooses 0 with 0 < P0 A < 1. In the
above calculation, as n becomes large, the probability P0 A approaches
⎧ ω̃
⎪ sin2 ( 4π
5 ω)
⎨ sin2 ( 4π )+sin 2 ( 4π ω̃ ) (if ω ≥ ω̃)
5 ω
P0EA = lim P0 A (n) = 5
.
n→∞ ⎪
⎩ sin2 ( 4π
5 )
ω (if ω̃ ≥ ω)
sin2 ( 4π 2 4π
5 ω̃ )+sin ( 5 )
In a sense, this function gives a boundary between rational solutions (P0EA < 1/2)
and irrational solutions (P0EA > 1/2), see Fig. 6.6.
6.2 Decision Making Process in Game 105
α1
P0A<1/2
α0
If Alice makes her decision with (α0 , α1 ) on the boundary, she may strongly feel
a “dilemma”: See Fig. 6.7, which shows P0 A (n) at n = 1, 2, 3, 4, 7, 8 and 19. One
can see that the probability P0 A (n) is intensively fluctuated near the boundary.
The irrational choice in PD comes from the fact that the pay-off assigned for the
consequence 0 A 0 B is larger than the one for 1 A 1 B . In the case of the game I, the
difference of pay-offs of these consequences is only 4 − 3 = 1. Therefore, as can be
seen from Fig. 6.6, the domain of irrational solutions is relatively small. In Fig. 6.8,
we introduce the two games denoted by game II and game III in which pay-offs
assigned for 0 A 0 B are 6 and 10. The boundaries in Fig. 6.8 are given by
⎛ ⎛ ⎞⎞
1 − α02
II: α1 = sin ⎝arctan ⎝3 − ⎠⎠ ,
α0
⎛ ⎛ ⎞⎞
1 − α02
III: α1 = sin ⎝arctan ⎝10 − ⎠⎠ ,
α0
and these domains are larger than those in the game I. The PD game discussed in
introduction to this chapter is an extreme example with a large domain. It is expected
that in the above game,
A/B 0B 1B
0 A 100000/100000 2 /100000 + 1
1 A 100000 + 1/ 2 3/3
many real players will choose 0. Let us consider Alice at the state parametrized by
(α0 , α1 ) ≈ (0, 0): here she feels that Bob will choose 1 B with probability nearly 1.
(Alice may be a believer in the classical game theory.) When the game I is given, Alice
will choose the rational “1 A ” with high probability, because the boundary of Fig. 6.6 is
106 6 Application to Decision Making Theory and Cognitive Science
Fig. 6.7 The probability P0 A (n) fluctuated in the process of decision making
far from Alice’s point (α0 , α1 ) ≈ (0, 0). In the game III or the above extreme game,
Alice may touch the boundary and develop the possibility of the irrational choice.
6.3 Bayesian Updating Biased by Psychological Factors 107
2 Of course, every probability in this case is subjective, so that we do not need to use classical
probability theory in order to estimate this posterior probability. Instead of Bayes’ formulae, humans
often use more heuristic inference, and there are many reports in psychology.
108 6 Application to Decision Making Theory and Cognitive Science
P(C|A)P(A)
P(A|C) = .
P(C|A)P(A) + P(C|B)P(B)
When Alice sees the occurrence of D in S2 , she can update her prediction for the
event A from P(A) to
P(D|A)P(A)
P(A|D) = .
P(D|A)P(A) + P(D|B)P(B)
In the same way she updates her prediction for the event B. These conditional (updat-
ing) probabilities are called the posterior probabilities. The change of prediction is
described as “updating” from the prior probabilities P(A), P(B) to the posterior
probabilities, and it is a classical scheme of Bayesian updating.
In the paper [43], we redescribed the process of Bayesian updating in the frame-
work of “quantum-like representation”, where we introduced the following state
vector belonging to the Hilbert space H = H1 ⊗ H2 = C2 ⊗ C2 ;
|Φ = P(A ) A ⊗ ( P(C |A ) C + P(D |A ) D )
+ P(B ) B ⊗ ( P(C |B ) C + P(D |B ) D ). (6.23)
We call this vector the prediction vector. The set of vectors { A , B } is an
state
orthogonal basis on H1 , and { C , D } is another orthogonal basis on H2 . The A ,
B , C and D represent the events defined as
Event A (B ): Alice judges “the event A(B) occurs in the system S1 .”
Event C (D ): Alice judges “the event C(D) occurs in the system S2 .”
These events
are subjective
events
(judgments) in Alice’s “mental space” and
the vectors A , B , C and D give a quantum-like representation of the above
judgments. The vector |Φ represents the coexistence
of these judgments
√ in Alice’s
brain. For example, Alice is conscious of A with the weight P(A ), and under
, she sets the weights
√ √
the condition of the event
A
P(C |A ) and P(D |A ) for
the judgments C and D . Such an assignment of weights implies that Alice feels
6.3 Bayesian Updating Biased by Psychological Factors 109
causality between S1 and √ S2 : The events in S1 are causes and the events in S2
are results. The square of P(A ) corresponds to a prior probability P(A) in the
Bayesian theory. If Alice knows the objective
√ conditional
√ probabilities P(C|A) and
P(C|B), she can set the weights of P(C |A ) and P(C |B ) from P(C |A ) =
P(C|A) and P(C |B ) = P(C|B). If Alice has the prediction state |Φ Φ| ≡ ρ
and sees the occurrence of the event C in S2 , the event D vanishes instantaneously
(in her mental representation). This
vanishing
is represented as the reduction by the
projection operator MC = I ⊗ C C ;
MC ρ MC
≡ ρC
tr(MC ρ)
The posterior probability P(A|C) is calculated by
tr(M A ρC ),
where M A = A A ⊗ I .
The inference based on the Bayesian updating is rational—from the viewpoint of
classical probability theory, game theory and economics (the Savage sure thing princi-
ple). However, in cognitive psychology and economics one can find extended empir-
ical data showing that sometimes human inference seems to be irrational, see [23]
for the review. Typically this happens in contexts such that there are (often hid-
den) psychological factors disturbing the rational inferences. Our aim is to provide a
mathematical description of such an irrational inference; the concept of lifting will
be used. Let us introduce a lifting from S(H ) to S(H ⊗ K ) by
Eσ V (ρ) = Vρ ⊗ σ V ∗ .
ρσ V ≡ tr K (Lσ V (ρ)),
the prediction state biased from the rational prediction ρ. From this ρσ V , the joint
probability is defined as
tr (M X MY ρσ V MY ) ≡ Pσ V (X , Y ),
Pσ V (X , Y )
Pσ V (X |Y ) =
tr(MC ρσ V )
110 6 Application to Decision Making Theory and Cognitive Science
In this section, we discuss the process of visual perception of human beings from the
adaptive dynamical point of view.
Most people may have the experience that when they watch the moon in the evening,
the moon near the horizon is larger than when it is high in the sky. Nevertheless,
in reality, the size of the moon does not change. This phenomenon has been known
since the ancient Greek era, and nowadays it is called moon illusion. Why do we
feel such an unreality? Many psychologists or other scientists have been discussing
the mechanism of such an illusion, and several hypotheses have been proposed so
far; for example, the moon illusion is caused by the atmospheric refraction, see the
Ptolemy’s book of Almagest (ca. 150 C.E.), or by the movement of eyes [44], or by
others physiological factors. However, all such hypotheses are not definitive, and the
debate is still in progress.
What we can clearly say is that our perception of an object is not exactly the same
as it exists in reality. A human being does not recognize just a raw snapshot of what
comes into our eyes. Our recognition is unconsciously biased. How to bias depends
on how it exists in certain surroundings. In Fig. 6.9, there are two pictures which
have a single letter “A”. These letters are of the same brightness of color in reality.
However they seem different to us, which is biased by the background colors. This
is also one of the optical illusions.
An ambiguous figure is a figure which has multiple different meanings. See Fig. 6.10,
called Schröder’s stair. The ambiguous figure induces optical illusion [45]. Our brain
can switch between the two alternative interpretations of this figure; (i) The surface
of ‘L’ is at the front, and the surface of ‘R’ is at the back; (ii) The surface of ‘R’
is at the front, and the surface of ‘L’ is at the back. This switch-like process of
human perception is called depth inversion, and many experimental proofs of this
phenomenon have been reported. However, similarly to other optical illusions, the
details of its mechanism are not completely figured out even in recent studies.
Recently, Atmanspacher et al. have explained the frequent change of human per-
ception for another ambiguous figure, which is called Necker’s cube by means of
quantum-like model on two-level system [46]. They introduce the neutral state for
Necker’s cube, that is, they consider the pre-observed state, which is not changed to
one of two perceptions. The mathematical expression of the neutral state is given by
a superposition of two orthogonal vectors in the two dimensional space, and its oscil-
lation is given by a certain Hamiltonian, a Pauli matrix. They also consider frequent
measurements given by two projections with respect to the orthogonal vectors. By a
series of measurements, they described a stable recognition of the figure.
On the other hand, we focus on the context of the perception process from the
adaptive-dynamical point of view. Let us come back to the discussion on Schröder’s
stair. It is a well-known fact that the depth inversion depends on various contexts
of the figure; e.g., the relative size of the surface ‘L’ for ‘R’, color or shadow of
the figure, the angle to the horizon, etc. We discuss the adaptive system where such
contextual dependencies emerge. We calculate the contextual dependent probability
that a person answers either (i) or (ii) in the experiment.
We showed the picture of Schröder’s stair which is inclined at a certain angle θ
(see Fig. 6.10) to the 151 subjects. We prepared the 11 pictures which are inclined at
different angles: θ = 0, 10, 20, 30, 40, 45, 50, 60, 70, 80, 90. A subject must answer
either (i) “L is the front side” or (ii) “R is the front side” for every picture. We arranged
the computer experiment to change the pictures and to record their answers.
Before the experiment, we divided the 151 subjects into three groups: (A) 55
persons, (B) 48 persons, (C) 48 persons. For the first group (A), the order of showing
is randomly selected for each person. To assume statistically uniform randomness of
this selection, we used the computer-implemented function (e.g. java.rand). For the
second group (B), the angle θ is changed in the increasing sequence: 0, 10, . . . , 90.
Inversely, for the third group (C), the angle θ is changed in the decreasing sequence:
90, 80, . . . , 0.
Figure 6.11 shows the probability that a subject says (i) “L is the front side”
with respect to angle θ . We denote it by P(X θ = L). The probability P(X θ = L) is
decreased as the value of θ is increased. However, the speed of decreasing is different
in the three cases. Moreover, we can see that the probability at θ = 0 is nearly equal
to one in each cases. That is, almost every subject says “L is the front side” at the
angle θ = 0. Conversely, the probability θ = 90 is nearly equal to zero, that is,
almost every subject says “R is the front side” at the angle θ = 90. Therefore it can
be considered that subjects feel little ambiguity of the picture for θ = 0 and 90. In
fact, we observed that it took very short time (less than 1 s) to make his/her decision
when the angle θ is around 0 or 90◦ .
In fact, the experiments on recognition of ambiguous figures were the first experi-
ments outside of physics (in cognitive psychology), which were designed to demon-
strate violation of the formula of total probability, see [8, 9]. In this section we show
that the data from the experiment with the Schröder stair, Sect. 6.4.2, also violate this
basic law of the Kolmogorov theory of probability.
As explained in the previous section, we consider the three situations (A), (B) and
(C) in the experiments. We can write the violation of total probability law as follows:
PA (X θ = L) = PB (X θ = L , X θ = L) + PB (X θ = L , X θ = R) + Δ AB ,
PA (X θ = L) = PC (X θ = L , X θ = L) + PC (X θ = L , X θ = R) + Δ AC ,
PC (X θ = L) = PB (X θ = L , X θ = L) + PB (X θ = L , X θ = R) + ΔC B ,
where θ and θ are different angles. Note that, with experimental results, we can
estimate the degree of violation Δ AB , Δ AC and ΔC B , for instance, Δ AB at the angle
θ is given by
Δ AB = PA (X θ = L) − PB (X θ = L , X θ = L) + PB (X θ = L , X θ = L)
= PA (X θ = L) − PB (X θ = L).
We can see that Δ AB does not depend on θ since only θ appears in the RHS of the
above equation. This comes from Kolmogorovness within the context B s.t.
PB (X θ = L) = PB (X θ = L , X θ = L) + PB (X θ = L , X θ = L).
From similar discussions, one can find that ΔAC and ΔCB depend only on θ but not
on θ . As seen in Fig. 6.12, the strong violation becomes manifest in the middle range
of the angle θ .
Here, let us compare the above ΔAB , ΔAC or ΔBC with the conventional trigono-
metric interference in quantum mechanics. If ΔAB , ΔAC and ΔBC have the form of
trigonometric interference
Δ AB = 2 cos (φ AB ) PB (X θ = L , X θ = L)PB (X θ = L , X θ = R),
Δ AC = 2 cos (φ AC ) PC (X θ = L , X θ = L)PC (X θ = L , X θ = R),
Δ BC = 2 cos (φ BC ) PC (X θ = L , X θ = L)PC (X θ = L , X θ = R), (6.24)
then the following quantities should take values between −1 and +1.
Δ AB
δ AB = √ ,
2 PB (X θ = L , X θ = L)PB (X θ = L , X θ = R)
Δ AC
δ AC = √ ,
2 PC (X θ = L , X θ = L)PC (X θ = L , X θ = R)
Δ BC
δ BC = √
2 PC (X θ = L , X θ = L)PC (X θ = L , X θ = R)
However,
as shown in Fig. 6.13, the values of δ BC exceed one at several points
θ, θ . Therefore, the probability PB (X θ = L) can not be written as the conventional
form of quantum mechanical interference:
PB (X θ = L) = PC (X θ = L , X θ = L)
2
+ exp (iφ BC ) PC (X θ = L , X θ = R) .
Subjects have to answer either (i) “L is the front side” or (ii) “R is the front
side”,
1
so that we take the Hilbert space H = C2 for this model. Let |L = and
0
0
|R = be the orthogonal vectors describing the answers (i) and (ii), respectively.
1
The initial state of a subject’s mind is given by
ρ ≡ |x x| ,
where |x is a state vector √1 (|L + |R), which represents the neutral mind for |L
2
and |R before the decision making starts.
When a subject is shown a picture inclined at an angle θ , he recognizes the lean
of the picture. Such a recognition process is given by the operator
cos θ 0
M (θ ) = .
0 sin θ
After the recognition, the state of mind changes from the initial state ρ to an adaptive
state
M ∗ρ M cos2 θ cos θ sin θ
ρθ = Λ∗M (ρ) ≡ = .
tr |M|2 ρ cos θ sin θ sin2 θ
σ = ρθ ⊗ · · · ⊗ ρθ ,
N
where N is the number of the agents created by the subject in the process of approach-
ing to the answer (i) or (ii). If it takes him little time to answer, then N might be
small. After the creation of agents (or on the way to create them), the subject can
talk with the imaginary agents in the brain, and he/she knows the answer of each
agent. Through this dialogue, the subject can know the opinions of all the agents.
For N agents, there are 2 N possible opinions. The subject’s answer is determined by
reference to the opinions of all the agents.
At the final step
⊗N of this dialogue, we additionally assumed that σ changed into
L or R ⊗N since no subject in this experiment can answer anything except (i) or
116 6 Application to Decision Making Theory and Cognitive Science
Q (2) = |L L L L| + |R R R R| ,
By applying the operator Q to the state of agents σ , the minority of opinions of the
agents are ignored and changed to the opinion of the majority agents. In this sense,
our model is considered as the
majority
voting
system for the N agents.
Here, the lifting Eσ∗Q : S C2 → S C2N is defined as
⊗N ∗
Qσ Q ∗ Q Λ∗M (ρ) Q
Eσ∗Q (ρ) = 2 = ⊗N ∈ S C2N
,
tr |Q| σ tr |Q|2 Λ∗M (ρ)
cos4 θ
P (2) (X θ = L) = ,
cos4 θ + sin4 θ
3 2
cos θ + 3 cos2 sin θ
P (3) (X θ = L) = 2 2 ,
cos3 θ + 3 cos2 sin θ + sin3 θ + 3 sin2 cos θ
4 2
(4) cos θ + 4 cos3 θ sin θ
P (X θ = L) = 2 2 .
cos4 θ + 4 cos3 θ sin θ + sin4 θ + 4 sin3 cos θ
6.4 Optical Illusions for an Ambiguous Figure 117
Fig. 6.14 Values of P (X θ = L) in our model and its comparison with experimental data (A)
We compared these probabilities with the experimental data of case (A), see Fig. 6.14.
One can find that the probabilities P(X θ = L) in N = 2, 3 and 4 coincide with the
experimental data of A.
In the situation (B) or (C), we take another adaptive operator M(θ + φ), whose
angle is shifted by an unknown psychological bias φ instead of M(θ ). This value of
φ can also be calculated by the experimental data. We show this in Fig. 6.15. The
values of φ in (B) are higher than those in (C), and especially the difference between
φ (B) and φ (C) is clearly seen in the middle range of the angle θ . By increasing N ,
this difference becomes smaller.
In physics the Leggett-Garg (LG) inequality [47] is one of the fundamental statistical
tests demonstrating the impossibility to use the classical realistic model in quantum
physics. This is a kind of the Bell inequality [48]. Opposite to spatial versions of
the Bell inequality emphasizing the role of nonlocality (such as Clauser-Horne-
118 6 Application to Decision Making Theory and Cognitive Science
3 Similar interpretation for the Bell inequality was presented in the paper [50], the first paper which
demonstrated that the Bell-type inequalities can be violated outside of physics, for the data collected
from recognition of ambiguous figures; see also [51]; see [49] for the detailed presentation of the
contextual viewpoint on quantum theory in general and on the Bell-type inequalities in particular.
6.5 Contextual Leggett-Garg Inequality for Statistical Data … 119
Conditions of Derivation
At the beginning of the discussion in the paper [47], Leggett and Garg (LG) postulated
the following two assumptions:
(A1) Macroscopic realism: A macroscopic system with two or more macroscopi-
cally distinct states available to it will at all times be in one or the other of
these states.
(A2) Noninvasive measurability: It is possible, in principle, to determine the state
of the system without arbitrarily small perturbation on its subsequent dynam-
ics.
Under these assumptions, the correlation functions must satisfy the LG inequality,
which will be presented in the next section. However, quantum mechanics violates
the LG inequality as well as spatial versions of the Bell inequality such as Clauser-
Horne-Shimony-Holt inequality. Therefore this violation means that at least one of
the two assumptions fails for quantum systems.
Although in the derivation of the LG inequality, Sect. 6.5.1, both conditions play
important roles, their foundational values are different. The main issue is realism,
whether one can still proceed with (A1), macroscopic realism, in the quantum world.
Therefore the main part of the LG paper [47] is devoted to the discussion about pos-
sible physical experimental schemes, which may lead to noninvasive measurements
or at least measurements in which invasiveness is small as compared with the degree
of violation of the LG inequality. There are claims, e.g., the experiment in [56], that
such negligibly invasive measurements were performed experimentally and the LG
inequality was violated. This is an important argument in favor of non-objectivity of
quantum observables.
However, the LG approach plays an important foundational role even if the possi-
bility that measurements are non-negligibly invasive cannot be excluded. We know
that classical systems and measurements on such systems satisfy conditions (A1)
and (A2); e.g., airplane’s trajectory. Therefore by violating LG we at least know that
a phenomenon under study cannot be described classically.
In cognitive science it is not easy (if possible at all) to come with an experimental
scheme, which would lead to (at least approximately) noninvasive measurements.
The brain is a kind of self-measurement device, by giving an answer to a question
the brain definitely perturbs its mental state non-negligibly. And the introspective
measurements definitely have the lowest degree of non-invasiveness. Therefore it
seems that violations of the LG type inequalities for the data collected, e.g., in cog-
nitive psychology, cannot lead to the conclusion that mental realism is questionable.
Here realism is understood in the sense of objectivity of mental observables: their
values can be assigned, say to the brain, a priori, i.e., before measurements. Never-
theless, such violations show that the data under consideration is nonclassical, i.e.,
it is not similar to the data collected, e.g., from an ensemble of moving airplanes.
120 6 Application to Decision Making Theory and Cognitive Science
However, our main point is that in relation to the problem of cognition the standard
physical viewpoint on conditions for derivation of the LG inequality, namely, the mix-
ture of macrorealism and non-invasiveness, does not match the mental situation well
enough. As was emphasized, the very notion of (non)invasive measurement loses its
clearness for self-measuring devices, and the brain is one of such devices. This book
advertises the contextual viewpoint on mental phenomena. It seems that Bell type
inequalities, including the temporal ones, can be used to distinguish between con-
textual and non-contextual realism and more generally (since mental processes are
fundamentally random) contextual and non-contextual probabilistic representations.
As will be seen from coming presentation, non-contextuality of the representation
of probabilistic data implies constraints on the data in the form of various inequal-
ities (or equalities such as FTP). By using the contextual representation a system
(including the brain) can violate such constraints.
Contextual Viewpoint on the Proof of LG Inequality
To provide a possibility to compare the original LG inequality with our contextual
generalization, see Sect. 6.5.2, and at the same time to add the contextual flavor to
the LG approach, we present the original LG derivation by considering time as a
context parameter.
Let Q be an observable quantity which takes the values +1 or −1. In the original
discussion by LG, Q is the observable of position of a particle in the two potential
wells. However we can discuss another two-level system, e.g., the spin- 21 system.
The measurement of a two-level system is performed on a single system at different
times t1 < t2 < t3 . We denote the observable at time tk by Q k (k = 1, 2, 3). By
repeating a series of three measurements, we can estimate the values of the correlation
functions by
1 (n) (n)
N
Ci j = qi q j ,
N
n=1
where qi(n) (or q (n) j ) is the result of the n-th measurement of Q i (or Q j ). Note
that the correlation between Q i and Q j takes the maximum value Ci j = 1 when
(n) (n)
qi q j equals 1 for all the repeated trials. Here, consider the assumption A1, then
the state of the system is determined at all times even when the measurement is
not performed on the system. Therefore, the values of joint probabilities of Q 1 , Q 2
and Q 3 are determined a priori at the initial time t0 . We denote it by the symbol
Pi, j (Q 1 , Q 2 , Q 3 ). Remark that the pairs of indexes i, j encode the situation that
only two observables Q i and Q j are measured. In other words, the joint probability
depends on the situations for which a pair of observables is measured. (We can
consider pairs of indexes, instances of time, as parameters encoding three temporal
contexts, Ct1 t2 , Ct1 t3 , Ct2 t3 , cf. Sect. 6.5.2.) However, if one considers A2, then the
joint probabilities do not depend on temporal contexts:
6.5 Contextual Leggett-Garg Inequality for Statistical Data … 121
Pi, j (Q 1 , Q 2 , Q 3 ) = P (Q 1 , Q 2 , Q 3 ) ∀i, j
Moreover,
the correlation functions are written with the joint probabilities
P Q i , Q j as
Ci j = P Q i = 1, Q j = 1 + P Q i = −1, Q j = −1
− P Q i = −1, Q j = 1 − P Q i = 1, Q j = −1
= 2 P Q i = 1, Q j = 1 + P Q i = −1, Q j = −1 − 1.
K ≤ 1. (6.26)
As we know, e.g., [47, 56] for quantum correlation functions Cij the above inequal-
ity can be violated (theoretically and experimentally).
and
PC (Q 1 ) = PC (Q 1 , Q 2 ) = PC (Q 1 , Q 3 )
Q 2 =±1 Q 3 =±1
= PC (Q 1 , Q 2 , Q 3 ) ,
Q 2 =±1 Q 3 =±1
PC (Q 2 ) = PC (Q 1 , Q 2 ) = PC (Q 2 , Q 3 )
Q 1 =±1 Q 3 =±1
= PC (Q 1 , Q 2 , Q 3 ) ,
Q 1 =±1 Q 3 =±1
PC (Q 3 ) = PC (Q 2 , Q 3 ) = PC (Q 1 , Q 3 )
Q 2 =±1 Q 1 =±1
= PC (Q 1 , Q 2 , Q 3 ) .
Q 1 =±1 Q 2 =±1
(A2) Consider three different contexts C A , C B and CC , then there exists a context
C unifying the above contexts C A , C B and CC such that
PC A (Q 1 , Q 2 ) = PC (Q 1 , Q 2 , Q 3 ) ,
Q 3 =±1
PC B (Q 2 , Q 3 ) = PC (Q 1 , Q 2 , Q 3 ) ,
Q 1 =±1
PCC (Q 1 , Q 3 ) = PC (Q 1 , Q 2 , Q 3 ) .
Q 2 =±1
Under these assumptions, one can obtain inequality (6.25) for K given by
In this section we use the data from the experiment presented in Sect. 6.4.2. Thus we
have the three kinds of experimental data: (A) the angle of Schröder stair changes
randomly, (B) from 0 to 90, (C) from 90 to 0. These contexts of experiments are
6.5 Contextual Leggett-Garg Inequality for Statistical Data … 123
denoted by C A , C B and CC . Let X θ be a random variable which takes the values ±1.
The event that a subject says “the left side is the front side” corresponds to the result
X θ = +1. Then, from the repeated trials for each experimental context, we have the
experimentally obtained values of joint probabilities:
PC A (X 0 , X 10 , . . . , X 90 ) , PC B (X 0 , X 10 , . . . , X 90 ) , PCC (X 0 , X 10 , . . . , X 90 ) .
Tables 6.4 and 6.5 shows that the value of K with respect to (θ1 , θ2 , θ3 ) = (0, 10, 20)
and (40, 45, 50). The value of K exceeding one is seen in several cases.
The violation of the contextual LG inequality by the statistical data collected for
observations of the Schröder stair rotated for different angles supports the contextual
cognition paradigm presented in the previous chapters of this book. Biological
probabilistic data are fundamentally contextual. The brain does not have a priori
prepared “answers” to the question about the R/L structure of the Schröder stair for
the fixed angle θ. Answers are generated depending on the mental context. Thus
mental realism is a kind of contextual realism. There are practically no (at least not
Table 6.3 (Left) (θ1 , θ2 , θ3 ) = (0, 10, 20), (Right) (θ1 , θ2 , θ3 ) = (40, 45, 50)
so many), so to say, “absolute mental quantities”, the “answers” to the same question
vary essentially depending on context. This conclusion is not surprising in the frame-
work of cognitive science and psychology, where various framing effects are well
known. Thus the main contribution of this study is the demonstration of applicability
of a statistical test of contextuality borrowed from quantum physics.
References
1. Fox, C., Rogers, B., Tversky, A.: Option traders exhibit subadditive decision weights. J. Risk
Uncertain. 13, 5–17 (1996)
2. Rottenstreich, Y., Tversky, A.: Unpacking, repacking and anchoring: advances in support the-
ory. Psychol. Rev. 104, 406–415 (1997)
3. Bearden, J.N., Wallsten, T., Fox, C.: Error and subadditivity: a stochastic model of subadditivity.
University of Arizona—Departmentt Management and Policy, preprint (2005)
4. Tversky, A., Koehler, D.: Support theory: a nonextential representation of subjective probability.
Psychol. Rev. 101, 547–567 (1994)
5. Khrennikov, A., Haven, E.: The importance of probability interference in social science: ratio-
nale and experiment (2007). arXiv:0709.2802v1 [physics.gen-ph]
6. Khrennikova, P., Khrennikov, A., Haven, E.: The quantum-like description of the dynamics of
party governance in the US political system (2012). arXiv:1206.2888 [physics.gen-ph]
7. Khrennikova, P.: Evolution of quantum-like modeling in decision making processes. AIP Conf.
Proc. 1508, 108–112 (2012)
8. Conte, E., Todarello, O., Federici, A., Vitiello, F., Lopane, M., Khrennikov, A.: A prelimi-
nary evidence of quantum-like behaviour in measurements of mental states. Quantum Theory:
Reconsideration of Foundations. Series Mathematical Model in Physics, Engineering and Cog-
nitive Science, vol. 10, pp. 679–702. Växjö University Press, Växjö (2004)
9. Conte, E., Khrennikov, A., Todarello, O., Federici, A., Zbilut, J.P.: Mental states follow quantum
mechanics during perception and cognition of ambiguous figures. Open Syst. Inf. Dyn. 16, 1–17
(2009)
10. Khrennikov, A.: Classical and quantum mechanics on information spaces with applications
to cognitive, psychological, social and anomalous phenomena. Found. Phys. 29, 1065–1098
(1999)
11. Khrennikov, A.: On quantum-like probabilistic structure of mental information. Open Syst.
Inf. Dyn. 11, 267–275 (2004)
12. Khrennikov, A.: Quantum-like brain: interference of minds. BioSystems 84, 225–241 (2006)
13. Busemeyer, J.B., Wang, Z., Townsend, J.T.: Quantum dynamics of human decision making. J.
Math. Psychol. 50, 220–241 (2006)
14. Busemeyer, J., Bruza, P.D.: Quantum Cognition and Decision. Cambridge University Press,
Cambridge (2012)
15. Hofstader, D. R.: Dilemmas for superrational thinkers, leading up to a luring lottery. Sci. Am.
6 (1983)
References 125
16. Hofstader, D.R.: Metamagical Themes: Questing for the Essence of Mind and Pattern. Basic
Books, New York (1985)
17. Shafir, E., Tversky, A.: Thinking through uncertainty: nonconsequential reasoning and choice.
Cogn. Psychol. 24, 449–474 (1992)
18. Tversky, A., Shafir, E.: The disjunction effect in choice under uncertainty. Psychol. Sci. 3,
305–309 (1992)
19. Croson, R.: The disjunction effect and reasoning-based choice in games. Organ. Behav. Hum.
Dec. Process. 80, 118–133 (1999)
20. Cheon, T., Takahashi, T.: Interference and inequality in quantum decision theory. Phys. Lett.
A 375, 100–104 (2010)
21. Cheon, T., Tsutsui, I.: Classical and quantum contents of solvable game theory on Hilbert
space. Phys. Lett. A 348, 147–152 (2006)
22. Khrennikov, A.: Quantum-like model of cognitive decision making and information processing.
Biosystems 95, 179–187 (2009)
23. Khrennikov, A.: Ubiquitous Quantum Structure: From Psychology to Finance. Springer, Hei-
delberg (2010)
24. Fichtner, K.-H., Fichtner, L., Freudenberg, W., Ohya, M.: On a quantum model of the recog-
nition process. In: QP-PQ: Probability and White Noise Analysis, vol. 21, pp. 64–84 (2008)
25. Franco, F.: The conjunction fallacy and interference effects. J. Math. Psychol. 53, 415–422
(2009)
26. Khrennikov, A., Haven, E.: Quantum mechanics and violations of the sure-thing principle: the
use of probability interference and other concepts. J. Math. Psychol. 53, 378–388 (2009)
27. Pothos, E.M., Busemeyer, J.R.: Can quantum probability provide a new direction for cognitive
modeling? Behav. Brain Sci. 36, 255–274 (2013)
28. Pothos, E. M., Busemeyer, J. R., Trueblood, J. S.: A quantum geometric model of similarity.
Psychol. Rev. (in press)
29. Lambert-Mogiliansky, A., Busemeyer, J.R.: Emergence and instability of individual identity. In:
Busemeyer, J.R., Dubois, F., Lambert-Mogiliansky, A., Melucci, M. (eds.) Quantum Interaction
6th International Symposium, QI 2012. Lecture Notes in Computer Science, vol. 7620, pp.
102–113 (2012)
30. Dzhafarov, E.N., Kujala, J.V.: Selectivity in probabilistic causality: where psychology runs into
quantum physics. J. Math. Psychol. 56, 54–63 (2012)
31. Dzhafarov, E.N., Kujala, J.V.: Random variables recorded under mutually exclusive conditions:
contextuality-by-default, to appear in the Advances in Cognitive Neurodynamics. In: Proceed-
ings of the 4th International Conference on Cognitive Neurodynamics (2013). arXiv:1309.0962
[quant-ph]
32. Wang, Z., Busemeyer, J.R.: A quantum question order model supported by empirical tests of
an a priori and precise prediction. Top. Cogn. Sci. (2013) (to be published)
33. Wang, Zh, Busemeyer, J.R.: A quantum question order model supported by empirical tests of
an a priori and precise prediction. Top. Cogn. Sci. 5, 689–710 (2013)
34. D’Ariano, M., Fei, S.M., Haven, E., Hiesmayr, B., Jaeger, G., Khrennikov, A., Larsson, J.A.:
Foundations of Probability and Physics-6. Series Conference Proceedings, vol. 1424. American
Institute of Physics, Melville (2012)
35. Pothos, E.M., Busemeyer, J.R.: A quantum probability explanation for violation of rational
decision theory. Proc. R. Soc. B 276(1165), 2171–2178 (2009)
36. Asano, M., Ohya, M., Khrennikov, A.: Quantum-like model for decision making process in
two players game. Found. Phys. 41(3), 538–548 (2010)
37. Asano, M., Ohya, M., Tanaka, Y., Basieva, I., Khrennikov, A.: Quantum-like model of brain’s
functioning: decision making from decoherence. J. Theor. Biol. 281(1), 56–64 (2011)
38. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Tanaka, Y.: Quantum-like generalization
of the Bayesian updating scheme for objective and subjective mental uncertainties. J. Math.
Psychol. 56(3), 168–175 (2012)
39. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Tanaka, Y.: Quantum-like representation of
irrational inference. In: Busemeyer, J.R., Dubois, F., Lambert-Mogiliansky, A., Melucci, M.
126 6 Application to Decision Making Theory and Cognitive Science
(eds.) Quantum Interaction 6th International Symposium (QI-2012). Lecture Notes in Com-
puter Science, vol. 7620, pp. 138–147 (2012)
40. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Tanaka, Y., Yamato, I.: Adaptive dynamics
and its application to context dependent systems breaking the classical probability law. In:
Busemeyer, J.R., Dubois, F., Lambert-Mogiliansky, A., Melucci, M. (eds.) Quantum Interaction
6th International Symposium (QI-2012). Lecture Notes in Computer Science, vol. 7620, pp.
160–171 (2012)
41. Asano, M., Basieva, I., Khrennikov, A., Ohya, M., Yamato, I.: Non-Kolmogorovian approach
to the context-dependent systems breaking the classical probability law. Found. Phys. 43(7),
895–911 (2013)
42. Accardi, L., Ohya, M.: Compound channels, transition expectations, and liftings. Appl. Math.
Optim. 39, 33–59 (1999)
43. Asano, M., Ohya, M., Tanaka, Y., Khrennikov, A., Basieva, I.: Quantum-like representation
of Bayesian updating. In: Proceedings of the International Conference Advances in Quantum
Theory, vol. 1327, pp. 57–62. American Institute of Physics (2011)
44. Smith, R.A.: A Compleat System of Opticks. Cambridge, Published by the author (1738)
45. Schröder, H.: Ueber eine optische Inversion bei Betrachtung verkehrter, durch optische Vor-
richtung entworfener physischer Bilder. Annalen der Physik 181(10), 298–311 (1858)
46. Atmanspacher, H., Filk, T., Röme, H.: Complementarity in Bistable Perception, Recasting
Reality, pp. 135–150. Springer, Berlin (2009)
47. Leggett, A., Garg, A.: Quantum mechanics versus macroscopic realism: is the flux there when
nobody looks? Phys. Rev. Lett. 54, 857–860 (1985)
48. Bell, J.: Speakable and Unspeakable in Quantum Mechanics. Cambridge University Press,
Cambridge (1987)
49. Khrennikov, A.: Contextual Approach to Quantum Formalism. Springer, Berlin (2009)
50. Conte, E., Khrennikov, A., Todarello, O., Federici, A., Zbilut, J.P.: A preliminary experimental
verification on the possibility of Bell inequality violation in mental states. Neuroquantology 6,
214–221 (2008)
51. Dzhafarov, E.N., Kujala, J.V.: Quantum entanglement and the issue of selective influences in
psychology: an overview. Lect. Notes Comput. Sci. 7620, 184–195 (2012)
52. Atmanspacher, H., Filk, T.: Temporal nonlocality in bistable perception. In: Khrennikov, A.,
Atmanspacher, H., Migdall, A., Polyakov, S. (eds.) Quantum Theory: Reconsiderations of
Foundations-6. Special Section: Quantum-like Decision Making: From Biology to Behavioral
Economics, AIP Conference Proceedings, vol. 1508, pp. 79–88 (2012)
53. Atmanspacher, H., Filk, T., Römer, H.: Weak quantum theory: formal framework and selected
applications. In: Adenier, G., Khrennikov, A., Nieuwenhuizen, T.M. (eds.) Quantum Theory:
Reconsideration of Foundations, vol. 3, pp. 34–46. American Institute of Physics, New York
(2006)
54. Kolmogoroff, A. N.: Grundbegriffe der Wahrscheinlichkeitsrechnung. Springer, Berlin (1933).
English translation: Kolmogorov, A.N.: Foundations of Theory of Probability. Chelsea Pub-
lishing Company, New York (1956)
55. Acacio de Barros, J.: Joint probabilities and quantum cognition. In: Proceedings of the Interna-
tional Conference on Quantum Theory: Reconsiderations of Foundations-6, Växjö, Sweden,
11–14 June (2012)
56. Goggin, M.E., Almeida, M.P., Barbieri, M., Lanyon, B.P., O’Briend, J.L., White, A.G., Pryde,
G.J.: Violation of the Leggett-Garg inequality with weak measurements of photons. Proc. Natl.
Acad. Sci. 108(4), 1256–1261 (2010)
Chapter 7
Operational Approach to Modern
Theory of Evolution
Abstract In this chapter we discuss the possibility to apply the quantum formalism
to model biological evolution, especially at the cellular level—genetic and epigenetic
evolutions. We start with an extended historical introduction which can be useful for
non-biologists, physicists, mathematicians, psychologists, sociologists. In particu-
lar, detailed comparative analysis of Darwinism and Lamarckism is presented. We
discuss the process of elimination of the Lamarckian viewpoint from evolutionary
biology and its sudden reappearance in theory of epigenetic evolution. One of the
main messages of this chapter is that the ideology of quantum adaptive dynamics
(and, in particular, of theory of open quantum systems) matches perfectly with both
approaches (Darwinism and Lamarckism) and it can be used to unify them in a new
general theory, theory of quantum adaptive evolution.
7.1 Lamarckism
Lamarck was one of the first who presented evolution in biology as a scientific
theory [1]. One of the basic ideas of J.B. Lamarck is that an organism can pass
on characteristics that it acquired during its lifetime to its offspring. According to
Lamarck, evolution is fundamentally adaptive in its nature. He emphasized the role
of so to say an adaptive force in evolution:
Using the modern language, in any bio-system, its subsystems (organs in
Lamarck’s terminology) communicate actively with each other to inform about their
activity and matching with the environmental pressure. This information is trans-
formed even to germ cells which transform the results of adaptation to the next
generation.
In the framework of the present book this viewpoint is very important, since it
matches well with the idea of the authors that the dynamics of the QL-states1 of living
systems is mathematically described by the theory of adaptive dynamical systems
(Chap. 2) and, in particular, the theory of open quantum systems.
However, Lamarckian theory of evolution is not reduced to adaptive inheritance.
According to Lamarck, the adaptive dynamics increases information and, conse-
quently, biological complexity of bio-systems. This is so to say complexifying force.
Our adaptive dynamics approach models the “adaptive force”. However, the place
of the “complexifying force” in theory of adaptive and, in particular, open quantum
systems is not completely clear.
To match with the Lamarck model of complexity increasing evolution, we have to
find a “natural class” of QL adaptive dynamical systems, in the simplest case a class
of quantum Markov dynamical systems, which increases the complexity of states.
We remark that, in spite of the world-wide success of Darwinism, ideas of
Lamarck, especially the ideas about inheritance of the markers of adaptivity, were
supported by many authoritative biologists. We can mention I. Pavlov who claimed
that in his experimental studies on conditioned reflex he found out that this reflex can
be inherited and thus the reaction to conditioning is improved from one generation
to another. Unfortunately, the validity of his claim has never been tested by other
researchers (although it seems to be a routine test).
Lamarckism combined with genetics is known as neo-Lamarckism: the environ-
mental pressure induces adaptive mutations in genes which are transferred to the off-
spring. There exist a lot of experimental evidence both for and against neo-Darwinism
(Sect. 7.3). Recently, the ideas of neo-Lamarckism have been extended from genome
to epigenome. Opposite to genetics, in epigenetics the presence of the Lamarckian
component in the cellular evolution is well recognized in the community of evolu-
tionary biologists (Sect. 7.3).
7.2 Darwinism
1 We remind that quantum-like (QL) states are information states representing probabilistic
information about possible measurements on bio-systems, including self-measurements.
7.2 Darwinism 129
In this book we shall use both terms, neo-Darwinism and evolutionary synthesis,
as equivalent. The most important thing for our further consideration is the total
liberation of the theory from elements of the Lamarckism.
From ES1–ES3 and the citation of Futuyma, one can see that the evolutionary
dynamics is a combination of many counterparts of totally different origins. How to
model the resulting output of the interaction of so many factors? There have been
elaborated advanced mathematical models of genetic drift and natural selection and
their inter-relation. One can model effects of separation and so on. However, the main
problem of modeling synthesis of its counterparts is the impossibility to estimate their
contributions. As an example, we can consider the debates on the role of genetic drift.
This notion was introduced by Wright [6], who emphasized the role of the sample
errors in the processes of evolution. However, he was strongly criticized by Fisher [7],
who proposed a number of mathematical models of genetic drift and claimed that,
although it affects genetic variation in populations, the magnitude of its effect is small
comparing with the effect of natural selection. As the result of Fisher’s critique (as
well as other critical attacks), genetic drift was considered as a minor contribution
to evolution. However, later Kimura [8] coupled the debate on the importance of
genetic drift with his neutral theory of molecular evolution, which claims that most
of the genetic changes are caused by the genetic drift acting on neutral mutations.
And recently the role of genetic drift has been criticized by Gillespie [9].
In our opinion, this debate can continue for ever. Contributions of the genetic drift
and natural selection may vary essentially depending on context. It is difficult, if at all
7.2 Darwinism 131
possible, to estimate relative contributions of these two factors. The same is valid for
accounting contributions of other factors of evolution (both known and unknown).
Nowadays even the specter of Lamarck walks freely in halls and labs of biological
departments of the world’s top universities. It declares that the acquired characteris-
tics of cells play a crucial role in cellular evolution, see Koonin and Wolf [10].
2 The famous sole illustration of the “Origin of Species” shows a Tree of Life (or more precisely, a
series of trees presumably depicting the evolution of different divisions of organisms).
132 7 Operational Approach to Modern Theory of Evolution
a genetic change into the genome that is immediately adaptive with respect to that
particular cue.”
We also point to some other claims of experimental evidence in favor of
(neo-)Lamarckism directly on the genetic level. In a few articles there was claimed
that special environments can stimulate mutations in genome exceeding the predic-
tions based on Darwin’s model. The first paper in this direction was published by
Ryan [11]. His primary aim was to confirm (for a new experimental context) the
famous fluctuation test which was first performed by Luria and Delbruck [12]. In
the literature on foundations of neo-Darwinism, this test is considered as one of the
basic genetic confirmations of the neo-Darwinian mechanism of adaptive evolution:
First the totally random mutation and then the selection induced by an environment.
The main idea of this test can be presented as follows3 :
Take a large family of totally independent bacterial cultures and plate them onto separate
Petri dishes to grow under the pressure of some special “unfriendly environment”. Only
those who have special mutation would survive and produce large populations. If mutations
are totally random, then one can expect high variation in sizes of final populations. In one
culture, a mutant may appear very early and give rise to numerous colonies, in another, it
may appear very late and give rise to fewer colonies. As a result, there will be much more
fluctuations from one plate to another than in the case of environment induced mutations.
Suppose now that mutations appear only as a direct response to the environment. One can
also expect some kind of variation, however, the variations from one culture to another would
not be very high. And the experiment of Luria and Delbruck demonstrated that so to say
“Darwin was right (in fact, Wallace and Weismann), but Lamarck not”. The variation in
populations was too high to be explained by consistent environment driven mutations in all
populations. We also point out that simple probabilistic reasoning implies that in the latter
case the probability distribution of variations between populations has to be Poissonian. And
the Luria-Delbruck distribution differs non-negligibly from the Poissonian distribution.
Ryan [11] repeated this test (with another environment4 ) and he found that the
probability distribution differs essentially from the Luria-Delbruck distribution and it
is closer to the Poissonian distribution corresponding to the neo-Lamarckian scenario.
This experimental result did not attract much attention; it was practically ignored.
However, later an influential paper confirming Ryan’s statistics was published in
Nature by Cairns et al. [14]. Then numerous experiments were performed and a
series of papers in good journals was published, see [15–22]. However, whether
the obtained statistical data can be explained by cellular natural selection or some
contribution of adaptive mutations is present, is still the subject of hot debates.
One of the explanations of the data of the “Lamarckian type” is that the pressure
from unfriendly environment is considered by cell population as a stress, which
increases the total rate of (random) mutations including the environment matching
mutations. This explanation was considered as satisfactory by the genetic community
3 Here we follow the presentation from the work [13] of Vasily Ogryzko. He was one of the pioneers
in attempts to use quantum physics to model cellular evolution.
4 He used a mutation that would enable the cells to grow on lactose, reverting a previously inactive
and (around 2007) the specter of Lamarck disappeared from labs. However, as we
have seen, he is back once again: now with the CRISPR-Cas phenomenon.
Finally, we cite Koonin and Wolf once again [10]:
More generally, recent empirical and theoretical studies of diverse processes of stochastic
and deterministic changes in genomes make it clear that evolution is not limited to the basic
Darwinian scheme of random variation that is subject to selection. Evolution can be more
adequately depicted as a continuum of processes from completely random ones, under the
Wrightean modality defined by random variation and random fixation of changes via genetic
drift; to the Darwinian modality with ran-dom changes fixed by the deterministic process of
selection; to the Lamarckian mode in which both variation and fixation are deterministic.
Galton felt that the small, incremental steps by which natural selection supposedly
proceeded would be thwarted by the phenomenon he had discovered, which he called
regression (or reversion) to the mean. Hence, Galton believed that evolution must
proceed via discontinuous steps.
This debate between Galton’s and Darwin’s adherents was later transformed into
the well known debate between adherents of Mendelianism and biometricians. Since
at the beginning Mendel’s theory was in visible contradiction with Darwin’s con-
tinuous evolution, Bateson, one of the most prominent aliens of Galton, see [25],
actively used Mendel’s laws as supporting evolution by jumps [26]. However, later it
became clear that Mendel’s approach also can be used to explain continuous changes
a la Darwin. Nowadays, although it is commonly accepted that there were no evolu-
tionary jumps, the debate between Galtonists and Darwinists continues. Recently, a
form of Galtonism has been reborn in the theory of punctuated equilibrium.
Although S. J. Gould and N. Eldredge, the creators of the theory of punctuated
equilibrium [27], wanted to distance from evolutionary models based on saltations
(mainly because nowadays such evolutionary models are actively used by adherents
of creationism in evolution), the theory of punctuated equilibrium essentially based
on Galton’s ideas. By this theory, evolution consists of periods of rapid change
implying creation of new species and long periods of statis—stability of species.
The main distinctions are (a) the emphasis of statis (such periods can take millions
of years and during periods of statis the environment can vary essentially without
changing species), (b) the duration of the discrete events, creation of new species.
By Galton, such events are really instantaneous (one generation), by Gould and
Eldredge, “creation” can take 50,000–100,000 years, so that its duration is short (of
the spot-type) only in comparison with the duration of the period of statis.
However, in the same way as Galton, Gould and Eldredge claimed [27] that new
species are created not via random mutations and then natural selection in the major
population, but through speciation in isolated (e.g., geographically) relatively small
sub-populations. In such isolated sub-populations profitable changes would not be
totally washed by stochasticity of inbreeding in very large populations.
Galtonism as well as the theory of punctuated equilibrium were created to oppose
Darwinian and neo-Darwinian viewpoint that evolution is based on small gradual
changes in huge populations and that profitable small changes (selected to match with
the environmental pressure) during very long periods lead to new species (Although
Gould and Eldredge, who were rather opportunistic in the presentation of their views,
tried to combine peacefully punctuated equilibrium with gradualism [27].)
Once again we see that it is very difficult to terminate the aforementioned dis-
pute and make the choice between the discrete and continuous versions of evolution.
Therefore the operational QL-approach may again be considered as a tool for uni-
fication. We shall see, Sect. 8.2.3, that the evolution described by quantum master
equation is combined of jumps and continuous drifts. Thus usage of the appara-
tus of the theory of open quantum(-like) and more generally adaptive systems can
provide a synthetic mathematical description of discrete and continuous components
of evolutionary changes.
7.5 Epigenetic Lamarckism 135
In recent years cell biologists have found out that non-genetic variation acquired
during the life of an organism can sometimes be passed on to offspring E-a phe-
nomenon known as epigenetic inheritance, see, e.g., Russell’s monograph [28] (see
also Chap. 3). Recently several examples of adaptive mutations in eukaryotes by the
epigenetic mechanism have been reported, see Jablonka and Raz [29] for a detailed
review. Under the influence of the environment, the epigenome structure including
DNA methylation and histone modification may change during growth and such
changes sometimes can be inherited by the offspring. This is the adaptive mutation
and a kind of neo-Larmarckism [29].
Everywhere below we shall use the term epimutation: A heritable change in gene
expression that does not affect the actual base pair sequence of DNA.
We concentrate modeling on cellular epigenetic inheritance (CEI). This is a nar-
rower aspect of epigenetic inheritance as discussed in the broad sense (see also
Sect. 3.3.3). It refers to epigenetic transmission in sexual or asexual cell lineages, and
the unit of this transmission is the cell. Four types of CEI are recognized today [29]:
1. The CEI based on self-sustaining regulatory loops.
2. The CEI based on three-dimensional templating.
3. The chromatin-marking CEI.
4. The RNA-mediated CEI.
These types of CEI are realized in cells with the aid of very different processes
whose mechanisms are known only in general (many important details still have to
be clarified). Nevertheless, all the aforementioned CEIs are parts of one universal
phenomenon, namely, the development of special adaptive features under the pres-
sure of the environment and transmission of these features from a mother cell to
daughter cells.
Therefore a perspective to create a universal model of CEI describing all its types
in the common framework is very attractive. Of course, such an activity does not
contradict to the continuation of intensive studies in cellular system biology aimed
to the creation of detailed models for each CEI and their interrelation. We present an
operational model, see Sect. 1.2, of CEI, which is applicable to all its possible types
(known and even yet unknown). Here the keyword is adaptive dynamics (Sect. 1.4).
Our aim is to create an adaptive dynamical model of CEI. This model, although it does
not describe concrete cellular mechanisms, can be interesting for cellular biology. It
presents a general mathematical structure of CEI and justifies the epigenetic mech-
anism from the viewpoint of the theory of adaptive dynamical systems. In the light
of our model the existence of CEI-mechanisms in cells is very natural, cf. Sect. 1.4.
Roughly speaking, if CEI were not discovered in experimental studies, it would be
a tricky problem to explain its absence. Thus by using the operational approach and
ignoring details of cellular processes we acquire knowledge on universal information
processes beyond CEI.
136 7 Operational Approach to Modern Theory of Evolution
References
This chapter is devoted to a model of the epigenetic cellular evolution based on the
mathematical formalism of open quantum systems. We emphasize that, although
in this book we restrict our QL-modeling to the epigenetic evolution, it is clear
that the structure of the model allows it to be extended to describe evolution of
biological organisms in general. We restrict the model to epigenetics, since here we
can use a closer analogy with quantum mechanics and mimic behavior of a cell as
behavior of a quantum particle. In general, we have to take into account cell death
(annihilation in quantum terminology) and birth (creation). Mathematically, such a
model is more complicated.
What kind of the mathematical apparatus can be used to model the adaptive
dynamics operationally? As pointed out in Sect. 2.4, an important class of adaptive
dynamical systems can be described by the apparatus of the theory of open quantum
systems. Now we apply this theory to the epigenetic dynamics of a cell. In accordance
with the theory of open quantum systems, dynamics of the epigenetic state of a
1Such kind of natural selection is performed not on the cellular, but on the molecular level.
2Mathematically, it is characterized by approaching a steady state solution of the quantum master
equation (Sect. 2.4).
8 Epigenetic Evolution and Theory of Open Quantum Systems … 139
One of the basic examples illustrating the need of the information interpretation of
the wave function (the quantum state) was proposed by E. Schrödinger. It is known
as Schrödinger’s cat.
3By using the information interpretation of the quantum formalism we need not keep to quantum
physical reductionism as, e.g., Ogryzko [5, 6], McFadden and Al-Khalili [7], McFadden [8].
140 8 Epigenetic Evolution and Theory of Open Quantum Systems …
Schrödinger’s cat itself may know its real state, alive or dead. But we, observers,
do not know this state. Therefore, we think it is at the superposed state of alive or dead.
In this sense, even the E. coli in glucose/lactose experiments is at the superposition
of the lactose oriented or glucose oriented states (Chap. 5). E. coli itself may know
its state, but we do not know it before “asking it the question” expressed in the
experiment on growing on lactose/glucose and assaying its β-galactosidase. Hence,
before the assay, we should say that E. coli is at the state of superposition, see
again Chap. 5.
We have already pointed that the mathematical formalism of quantum mechan-
ics delivers only statistical predictions to an observer. The information encoded in
the wave function is about probabilities of possible results of measurements. This
formalism cannot say anything about a possible result of measurement for a spe-
cific single system. Therefore our model of the QL-evolution is of fundamentally
probabilistic nature. It describes the evolution of probabilities for genes’ expres-
sions in large ensembles of cells. What happens in a single cell is the subject of
system biology.
4 As was already emphasized, our model treats all possible sources of CEI and their interrelations
Our proposal is to use the machinery of the theory of open quantum systems and
to describe the dynamics of the epigenetic QL-state by using the quantum master
equation (the GKSL-equation), cf. with QL-cognitive modeling [2–4] and Chap. 6.
This equation can be used to describe transitions from states of uncertainty given
by QL-superpositions to classical probability distributions. Hence, such an equation
cannot be an equation with respect to the quantum state represented as a vector
belonging to the complex Hilbert space (normalized by one)-a pure state. Such
vectors represent superpositions of possibilities, which have to disappear at the end
of the epigenetic evolution. We have to use more general quantum states represented
by density operators. Both the purely quantum superposition describing uncertainty
and the final classical probability distribution can be represented by density operators
(Sect. 2.2).
In the quantum Markovian approximation the dynamics of the state of a system
interacting with an environment is described by the GKSL-equation, see (2.50):
dρepi
γ = −i[H, ρepi (t)] + W ρepi (t), ρepi (0) = ρepi
0
, (8.1)
dt
where H is a Hermitian operator determining the internal dynamics of epimutational
changes in cells, which are isolated from the environmental pressure (“cell’s Hamil-
tonian”), and the linear operator W describes the environmental pressure. In general,
opposite to H, the operator W has a complex mathematical structure. It has such a
form that starting with a density operator ρepi 0 we shall get density operators at all
instances of time. For the time being, the specific structure of W is not important for
us, see Sect. 8.2.3 for mathematical details. Biologically, this operator is determined
by the properties of the environment, including the initial state of the environment.
Here γ is the time scale constant, which determines the temporal dimension of
the epigenetic evolution. By using such a scaling factor of the dimension of time,
we are able to proceed with dimensionless Hamiltonian H and the environmental
operator W .
For a very general class of GKSL-equations, the environmental operator W drives
(in the limit t → ∞) the epigenetic state of an ensemble of cells, ρepi (t), to the steady
solution: ρepi (t) → ρepi ;st . Typically, the uncertainty (in the form of superposition)
is eliminated from the asymptotic state ρepi ;st .
In our QL-model such a steady state is considered as the result of the epige-
netic evolution in the environment (mathematically represented by the operator W ).
The limiting probability distribution ρepi ;st describes the probability distribution of
epimutations, which took place in a cell population as a consequence of interaction
with the environment. The internal uncertainty whether to (epi)mutate or not mutate,
was resolved and a stable phenotype was created.
142 8 Epigenetic Evolution and Theory of Open Quantum Systems …
The quantum Markovian dynamics (8.1) is derived (see, e.g., [16, 17]) under the
requirement that the reaction of the environment to a system is negligibly small. This
requirement is natural in modeling of the epigenetic evolution. We can assume that a
cell cannot change essentially the structure of the environment. Another condition for
derivation of Eq. (8.1) is the factorization of the initial state of the compound system:
The cell in combination with the environment. In quantum terms this means that at
the beginning the states of cells and the environment were not entangled. In cell
biological terms it means that before being under the influence of the environment
the population of cells and the environment were not correlated, i.e., the previous
evolution of cells was independent from this specific environment.
This special form of the open quantum dynamics is derived under the assumption
of Markovness. In the cell biological framework this assumption has the following
form. A cell does not “remember” a long chain of interactions with the environ-
ment; its state at the instant of time t + δt, ψC (t + δt), where δt is a very small
(mathematically infinitely small) interval, is determined by its state at the instant of
time t, ψC (t), and not by the family of its states in previous instants of time, i.e.,
{ψC (s) : s < t}. This is the most questionable assumption. We cannot exclude the
presence of long term memory effects in the epigenome of a cell interacting with an
environment. As well as in physics, we treat the Markovian condition as an approxi-
mate condition, i.e., long term memory effects may be present in a cell, but they are
sufficiently weak to justify the approximation in use.
8.2 Dynamics of Cell Epigenetic State in the Process of Interaction … 143
1
ρ → ρ − {L ∗ L , ρ}δt + Lρ L ∗ δt + o(δt). (8.4)
2
This expression can be written as
1
|Lψ (8.6)
|Lψ
We regard this action as a jump, since for δt → 0 the output state does not
approach the input state |ψ. The probability of no jump is given by Pno jump =
1 − ψ|L ∗ L|ψδt. However, the absence of a jump does not imply that the input
state is preserved. The branch without jump evolves as |ψ → |ψ − 21 L ∗ L|ψδt
with the corresponding normalization. We call this branch the drift-type evolution,
since the output state approaches the input state for δt → 0.
Thus, in the QL-model the environment driven evolution can be considered as a
branching process with “evolutionary jumps” (“quantum jumps”) 5 and continuous
drift-evolution. We shall study this problem in more detail in Sect. 8.3.1, where the
simplest model of epigenetic mutation in a single gene will be considered.
The sum in representation (8.2) of the environment operator can contain a few
terms. The number of terms can be very large, it grows as N 2 − 1, where N is the
dimension of the state space. In our model the dimension of the epigenetic state space
grows as N = 2n , where n is the number of epigenetic markers under consideration,
Sects. 8.4 and 8.5. Therefore, in general, the QL evolution is a combination of about
2n quantum jumps to the states determined by the operators L j and the corresponding
drifts. This is a branching process of great complexity. The epigenetic state jumps
in different directions (determined by the environmental operators L j ), outputs of
jumps form superpositions; if no L j -jump occurs, the state deforms continuously,
and these continuous deformations are superposed with the superposition of jumps.
At the next step, the state directions of jumps and drifts are randomly changed…
5Quantum jump (leap) is a jump of an electron from one quantum state to another within an atom.
Quantum jumps were invented by Einstein, who postulated that electrons in an atom can absorb
and emit electromagnetic energy only by discrete portions, which later were called photons. Thus,
opposite to classical systems, electron energy cannot change continuously.
8.3 Dynamics of a Single Epimutation of the Chromatin-Marking Type 145
Its elements ρ00;st and ρ11;st give probabilities of the events: “No μ-epimutation”
and “μ-epimutation”. Thus in a large population of cells, say M cells, M >> 1,
the number of, e.g., cells with mutation is given (approximately) by Nm ≈ ρ11;st M.
The limiting QL-state (represented by the diagonal matrix (8.10) approached the
stability with respect to the influence of this (concrete) environment. We remark that
146 8 Epigenetic Evolution and Theory of Open Quantum Systems …
As was already pointed out, it is possible to construct QL dynamics such that starting
with any state the trajectory will stabilize to the same pure state (see Chap. 4 for
cognitive applications), e.g., ρst = π1 = |11|. Even if at the beginning only a
few cells were (epi)mutated, finally, cells will mutate with the unit probability. By
moving from a mixed state, e.g., from the state given by the diagonal matrix with
equal elements, to the pure state |1,
1/2 0 00
ρ0 = → ρst = , (8.11)
0 1/2 01
the von Neumann (quantum) entropy decreases. We emphasize the role of environ-
ment in such an evolution.
Consider now the environment operator W based on a single Lindblad operator,
see (8.2), acting in a qubit space and having the form
√
L= p|10|, (8.12)
√
i.e., L|ψ = p0|ψ|1. Thus during a small time interval δt, each state |ψ can
jump only towards the same state |1, see (8.6). For the pure state |ψ = c0 |0+c1 |1,
the probability of such a jump is equal to Pjump = p|c0 |2 δt.6
In such a process all evolutionary jumps are oriented towards mutation (the con-
crete mutation under consideration); cell state cannot jump back by eliminating this
mutation. However, the evolution is not reduced to jumps. There is also a continuous
evolutionary drift, which changes the probability of a jump.
6 As in Sect. 8.2.3, we proceed under the assumption that the time scale constant γ was set as γ = 1. If
we take γ into account, then the formula for the jump-probability takes the form: Pjump = p|c0 |2 δtγ .
Hence, the smallness of the jump duration is relative to the time scale of evolution.
8.3 Dynamics of a Single Epimutation of the Chromatin-Marking Type 147
d|φ p
(t) = − π1 |φ(t). (8.13)
dt 2
The solution of this linear differential equation is given by |φ(t) = e− pt/2 c00 |0 +
c10 |1 and the corresponding QL state evolves as
Thus the pure no-mutation state |0 is stationary with respect to the continuous
evolutionary drift. This state can always jump to the mutation state |1 and the
probability rate of evolutionary jumps is constant, p.
If the input (pure) state differs from the pure no-mutation state, then it drifts
towards the pure mutation state as given by (8.14) and the probability rate of sudden
jumps to the pure mutation state |1 decreases as
e− pt p|c00 |2 δt
Pjump = .
e− pt |c00 |2 + |c10 |2
Thus the evolution corresponding to the very simple environment operator given
by (8.12) has the complex branching structure combining evolutionary jumps with
continuous drifts.
The evolutions similar to one driven by the operator (8.12) are widely used in
quantum physics. For example, take
√
L= p|01|. (8.15)
where j |cg; j |2 = 1.
What is the meaning of this superposition from the biological viewpoint? Can a
gene really be in superposition of a few different epimutations?
Although our model is operational and in principle we are not interested in such
questions, we make a comment to clarify the coupling of operational and biological
descriptions of this situation. A cell by itself “knows its epigenome” at each instant
of time; so it is well aware which epimutations have taken place up to this instant of
time. However, a biologist performing an experiment with cells does not know the
situation inside an individual cell in such detail. And superposition is related to the
uncertainty of the observer’s information.
If epimutations in different genes are independent from each other, then the QL-
state of cell’s epigenome is represented as the tensor product of states |ψg :
However, in living cells, most of the genes/proteins are correlated somehow forming
a big network system. Thus, usually one epimutation affects other genes. Hence
the assumption of independent epimutations is nonbiological. Therefore we have to
consider more general states describing the consistent epimutations of all genes in
the genome of a cell. These are entangled states (Sect. 2.5.1), which are widely used
in quantum information theory:
|ψepi = c j1 ... jm | jg1 . . . jgm , (8.18)
j1 ... jm
where | jg1 . . . jgm is a short notation for the tensor product of states of superpositions
in various genes, | jg1 . . . jgm ≡ | jg1 ⊗ . . . ⊗ | jgm and the sum of all squared
coefficients is equal to 1.
7 Depending on the biological context, it is always possible to select a few epimutations of the
main importance. Hence, the number k g need not be very large. We state again that our model is
operational. It need not be very detailed.
8.4 “Entanglement” of Epimutations in Genome 149
We remark that the notion of entanglement is in the very heart of quantum mechanics.
However, although it is widely used in quantum information, the understanding
of the physical essence of entanglement is far from to be complete, see Sect. 9.2.
Nevertheless, there is complete consensus that entanglement implies correlations-
in our epigenetic modeling these are correlations between epimutations in different
genes. (Whether entanglement is simply the tensor product encoding correlations is
an open question, see Sect. 9.2.)
Typically, in quantum foundations experts emphasize that correlations corre-
sponding to an entangled state are “superstrong”, i.e., they have amplitudes exceeding
the amplitudes, which are possible for classical correlations. Experimentally, non-
classicality of correlations is demonstrated in the form of violation of Bell’s inequality
(Sects. 9.1 and 9.2). However for our modeling of epigenetic mutations, the debate
on nonclassicality and superstrength of entanglement-correlations is not important.
The key point of the application of entanglement in modeling of epimutations is its
“nonlocal feature”. We remark that the notion of quantum nonlocality is often used
vaguely. Typically, especially in the relation to Bell’s tests, nonlocality is understood
as nonlocality of hidden variables (Sect. 9.2). In our cell biological studies we cannot
assume such “real physical nonlocality” of variables describing gene expressions.
We do not appeal to physical quantum effects in a cell. We state again that the
origin of QL-representation is the uncertainty in information about cell behavior in
experiment.
We shall appeal to another sort of nonlocality, which may be called operational
nonlocality. The form of the tensor space representation (8.18) of potential epimuta-
tions in cell genome implies that epimutation in one gene implies consistent epimu-
tations in other genes. If the state (8.18) is not factorized, then by changing of the
environment of one gene, say g1 , and inducing, see Sect. 8.3, some epimutation in
it, we can induce consistent epimutations in other genes. In the operational approach
(endowed with the information interpretation) this global change of the genome state
happens in information space and not in physical space, cf. [20, 21]. This is simply
the update of information, which experimenter has to make by taking into account
changing of the state of the gene g1 . Since information is encoded in probabilities,
this global state updating is nothing else than an updating of probabilities. Such an
updating is in general non-Bayesian, see also Chap. 4.8
8 What is beyond such a global (“nonlocal”) structure of state update? This is the separate question
that can be ignored in the operational approach. However, biologists may want to have some picture
of what happens beyond the operational description. Our picture is that gene expressions in genome
are strongly correlated; these correlations can be nonlocal in the purely classical field manner:
strongly correlated waves, including electromagnetic and chemical waves, in a cell. However, we
state again that we shall proceed in the purely operational framework. Hence this classical wave
picture for QL entanglement, see [22–27] need not be taken into account (and, moreover, it may be
wrong). We state again that a model of brain functioning based on the representation of information
by purely classical electromagnetic waves was elaborated in [28–30].
150 8 Epigenetic Evolution and Theory of Open Quantum Systems …
9 We state once again that it has nothing to do with nonlocality of hidden variables.
8.4 “Entanglement” of Epimutations in Genome 151
In Sect. 8.2.1 we pointed that, although under the environment driven QL-
dynamics the superposition will be finally resolved and a steady state solution will be
approached, the complete stabilization is possible only in the limit t → ∞. Hence,
for any finite interval of time, the total stabilization is impossible. This feature of our
QL-model also matches well with observation of Sollars et al. [31], p. 73:
Because of the inherent instability of epigenetic inheritance, fixation of an epigenetically-
determined phenotype is probably less stable than fixation through a genetic selection mech-
anism. Waddington, for example, was unable to reduce the frequency of the crossveinless
phenotype in negative selection experiments once the phenotype was fixed [33]. In contrast,
after only two or three generations of negative selection, we observed a complete rever-
sion to wild-type frequency of ectopic outgrowth in our sensitized iso-KrIf-1 strain in the
geldanamycin selection experiment (data not shown). Similarly, epigenetic traits such as
color variegation or cold adaptation in plants are unstably inherited, Bender [34] Kohler and
Grossniklaus [36]. Therefore, a combination of both epigenetic and genetic mechanisms is
probably required to explain the rapid changes in body plans that are observed in the fossil
record, Gould and Eldredge [37].
This classical state space consists of 2n points. This space is the basis of the classical
information description of the process of epigenetic evolution. However, we move
to the quantum information description by assuming that classical states can form
superpositions. To match with the Dirac ket-vector notation, which is used in quantum
physics, we denote the classical state α as |α ≡ |α1 . . . αn . Then QL state space
of (possible) epigenetic mutations, Hepi , consists of superpositions of the form
|ψ = cα |α,
α
where α |cα |2 = 1. This is a complex Hilbert space of dimension 2n . Now we repeat
our previous considerations, see Sects. 8.3 and 8.4, for epimutations of the chromatin-
marking type. The QL adaptive dynamics described by the quantum master equation
can be considered as a mixture of neo-Darwinian, neo-Lamarckian, and Wrightean
evolutions. This cocktail of stochasticity and determinism is consistently represented
in the QL operational framework. The final steady state gives to experimenters the
classical probability distribution of the inherited epigenetic markers.
As we have seen in Sect. 8.4, entanglement plays a crucial role in the speedup of
the epigenetic evolution. Since epimutations of the chromatin-marking type can be
coupled to physical carriers, it was easy to use the standard notion of entanglement
(as entanglement of systems) in the epigenetic framework. In general, epigenetic
markers are merely information structures in a cell such as, e.g., self-sustaining reg-
ulatory loops. However, we are lucky, since recently a new general viewpoint on
entanglement has been elaborated in quantum information community (Sect. 2.5.2).
Entanglement can be considered not from the system viewpoint, but from the observer
viewpoint as well. One considers a family of algebras of observables, say {Ai }, on
the total state space, in our case on Hepi . Under some restrictions on these algebras
the state space can be represented as the tensor product of subspaces corresponding to
these algebras. (This is an algebraic representation of contextuality of some observ-
ables for a single cell, Sect. 2.5.2.) In our case we consider algebras of observables
corresponding to different epigenetic markers, the corresponding subspaces are two
dimensional qubit spaces, Hepi = ⊗nj=1 H j;qubit . Now we can use the notion of
entanglement corresponding to this tensor product decomposition of the state space
and repeat the speedup argument which was discussed in detail in Sect. 8.4.
References
1. Jablonka, E., Raz, G.: Transgenerational epigenetic inheritance: prevalence, mechanisms, and
implications for the study of heredity and evolution. Q. Rev. Biol. 84, 131–176 (2009)
2. Asano, M., Ohya, M., Khrennikov, A.: Quantum-like model for decision making process in
two players game. Found. Phys. 41(3), 538–548 (2010)
3. Asano, M., Ohya, M., Tanaka, Y., Khrennikov, A., Basieva, I.: On application of Gorini-
Kossakowski-Sudarshan-Lindblad equation in cognitive psychology. Open. Syst. Inf. Dyn. 17,
1–15 (2010)
References 153
4. Asano, M., Ohya, M., Tanaka, Y., Basieva, I., Khrennikov, A.: Dynamics of entropy in quantum-
like model of decision making. J. Theor. Biol. 281, 56–64 (2011)
5. Ogryzko, V.V.: A quantum-theoretical approach to the phenomenon of directed mutations in
bacteria (hypothesis). Biosystems 43, 83–95 (1997)
6. Ogryzko, V.V.: On two quantum approaches to adaptive mutations in bacteria. http://arxiv.org/
ftp/arxiv/papers/0805/0805.4316.pdf
7. McFadden, J.J., Al-Khalili, J.V.: A quantum mechanical model of adaptive mutation. Biosys-
tems 50, 203–211 (1999)
8. McFadden, J.: Quantum Evolution. Harper (2000)
9. Donald, M.J.: A Review of Quantum Evolution. (2001) http://arxiv.org/abs/quant-ph/0101019
10. Fuchs, C. A.: Quantum mechanics as quantum information (and only a little more). In: Khren-
nikov, A. (ed.), Quantum Theory: Reconsideration of Foundations, Ser. Math. Modeling 2,
Växjö University Press, Växjö, pp. 463–543 (2002)
11. Caves, C.M., Fuchs, C.A., Schack, R.: Quantum probabilities as Bayesian probabilities. Phys.
Rev. A. 65, 022–305 (2002)
12. Fuchs, ChA, Schack, R.: A quantum-Bayesian route to quantum-state space. Found. Phys. 41,
345–356 (2011)
13. Zeilinger, A.: Dance of the Photons: From Einstein to Quantum Teleportation. Farrar, Straus
and Giroux, New York (2010)
14. Balazs, A.: Some introductory formalizations on the affine Hilbert spaces model of the origin
of life. I. On quantum mechanical measurement and the origin of the genetic code: a general
physical framework theory. Biosystems 85, 114–125 (2006)
15. Karafyllidis, I.: Quantum mechanical model for information transfer from DNA to protein.
Biosystems 93, 191–198 (2008)
16. Ingarden, R.S., Kossakowski, A., Ohya, M.: Information Dynamics and Open Systems: Clas-
sical and Quantum Approach. Kluwer, Dordrecht (1997)
17. Ohya. M., Volovich I.V.: Mathematical Foundations of Quantum Information and Computation
and Its Applications to Nano- and Bio-systems, Springer (2011)
18. Rooney, P., Bloch, A., Rangan, C.H.: Decoherence control and purification of two-dimensional
quantum density matrices under Lindblad dissipation. (2012) http://arxiv.org/abs/1201.0399
19. Fisher, R.A.: The Genetical Theory of Natural Selection. Clarendon Press, Oxford (1930)
20. Khrennikov, A.: On quantum-like probabilistic structure of mental information. Open. Syst.
Inf. Dyn. 11, 267–275 (2004)
21. Khrennikov, A.: Quantum-like brain: Interference of minds. Biosystems 84, 225–241 (2006)
22. Acacio de Barros, J., Suppes, P.: Quantum mechanics, interference, and the brain. Math. Psy-
chology 53, 306–313 (2009)
23. Khrennikov, A.: Detection model based on representation of quantum particles by classical
random fields: Born’s rule and beyond. Found. Phys. 39, 997–1022 (2009)
24. Khrennikov, A., Ohya, M., Watanabe, N.: Classical signal model for quantum channels. J.
Russ. Laser Res. 31, 462–468 (2010)
25. Khrennikov, A.: Prequantum classical statistical field theory: Schrödinger dynamics of entan-
gled systems as a classical stochastic process. Found. Phys. 41, 317–329 (2011)
26. Khrennikov, A.: Description of composite quantum systems by means of classical random
fields. Found. Phys. 40, 1051–1064 (2010)
27. Khrennikov, A., Ohya, M., Watanabe, N.: Quantum probability from classical signal theory.
Int. J. Quantum Inf. 9, 281–292 (2011)
28. Khrennikov, A.: On the physical basis of theory of “mental waves”. Neuroquantology 8, S71–
S80 (2010)
29. Khrennikov, A.: Quantum-like model of processing of information in the brain based on clas-
sical electromagnetic field. Biosystems 105(3), 250–262 (2011)
30. Haven, E., Khrennikov, A.: Quantum Soc. Sci. Cambridge Press, Cambridge (2012)
31. Sollars, V., Lu, X., Xiao, L., Wang, X., Garfinkel, M.D., Ruden, D.M.: Evidence for an epige-
netic mechanism by which HSP90 acts as a capacitor for morphological evolution. Nat. Genet.
33, 70–74 (2003)
154 8 Epigenetic Evolution and Theory of Open Quantum Systems …
32. Waddington, C.H.: Canalization of development and the inheritance of acquired characters.
Nature 150, 563–565 (1942)
33. Waddington, C.H.: Genetic assimilation of an acquired character. Evolution 7, 118–126 (1953)
34. Bender, J.: Plant epigenetics. Curr. Biol. 12, R412–R414 (2002)
35. Koonin, E. V., Wolf, Yu. I.: Evolution of microbes and viruses: A paradigm shift in evolutionary
biology? Frontiers in Cellular and Infection Microbiology. 1192 (2012)
36. Kohler, C., Grossniklaus, U.: Epigenetics: the flowers that come in from the cold. Curr. Biol.
12(4), R129–R131 (2002)
37. Gould, S.J., Eldredge, N.: Punctuated equilibrium comes of age. Nature 366, 223–227 (1993)
Chapter 9
Foundational Problems of Quantum
Mechanics
algebra and probability on it.) If such probability space cannot be constructed, then
the “Boole-Bell inequality” can be in principle violated. Later a Soviet probabilist
Vorobj’ev found the most general conditions for the existence of a single probability
space for a family of random variables, see again [4, 10] for details. Thus mathe-
maticians considered Bell’s type inequalities without any relation to nonlocality.
This viewpoint matches well with our discussion on non-Kolmogorovness of
quantum probability, see Sect. 4.1.1. In Sect. 9.2 we shall show that a few sources
of non-Kolmogorovness of collected statistical data can be associated with the real
experimental situation, once again [4] for details. Hence, addiction to nonlocality,
which is rather common in the quantum community, is not justified by the real exper-
imental situation. At the same time “death of realism”, nonobjectivity of quantum
observables (see Sect. 9.3), mathematically can be coupled to non-Kolmogorovness.
It is clear from the above presentation that Bell’s inequality can be used as a
statistical test of non-Kolmogorovness (and in this sense quantum-likeliness) of data
collected for pairs of observables of any origin, e.g., for biological observables. The
essence of this test is that observables can be measured pairwise, but it is impossible
(due to the specialty of experimental contexts) to measure these observables jointly.
Thus the experimenter gets a collection of pairwise probability distributions, but not
the total joint probability distribution. If the latter does not exist, one can expect
the violation of Bell’s inequality. Outside of physics the first Bell’s test was done
for recognition of ambiguous figures, Conte et al. [11]; for Bell’s test for semantic
models see Bruza et al. [12], see also the monograph of Bussemeyer and Bruza [13].
General analysis of usage of Bell’s tests in cognitive psychology was performed by
Dzhafarov et al. [14]. In cellular biology no Bell’s type tests have yet been performed.
And to find the corresponding experimental context and perform such an experiment
is an exciting problem.
The LG-inequality, see Sect. 6.4.2, is also an inequality of the Bell-type. Thus
our experimental test for it can be considered as one of the first Bell’s type tests
performed outside of physics.
This section is written for those who have already heard about the problem of hidden
variables, violation of Bell’s inequality, quantum nonlocality, (in)completeness of
quantum mechanics. We discuss interpretation problems related to these topics. Thus,
this section is written for biologists who may be afraid to use the quantum formalism
in biology as a consequence of the aforementioned foundational problems and who
are aware of these problems.
The debate on a possibility to go beyond the operational quantum formalism and
to introduce hidden (i.e., yet unknown) variables, which determine the results of
measurements, via the functional relation (1.1) was started about 90 years ago, by
9.2 Debate on Hidden Variables and Its Biological Dimension 157
Einstein and Bohr. Einstein was sure that sooner or latter a prequantum model will
be created and the quantum formalism would be considered as simply an operational
formalism appeared as a result of ignorance of the intrinsic degrees of freedom of
quantum systems. Bohr (as well as Heisenberg, Fock, Landau, von Neumann,…)
were sure that it is meaningless to look for such variables, since these are metaphys-
ical. Bohr presented motivations (based on the existence of the indivisible Planck
quant of action) that one cannot make better measurements than those described
by the quantum formalism. Hence, intrinsic variables, even if one were able to find
them, would be unmeasurable (metaphysical).
Later this position of fathers of quantum mechanics was strengthened, there
were “proven” various no-go theorems. The most known are “theorems” of von
Neumann [15], Bell [2], Kochen-Specker [16].1 By these theorems any attempt to
introduce hidden variables and to go beyond the operational quantum formalism
would imply either contradiction with the predictions of quantum mechanics or such
a pathological feature of hidden variables as nonlocality (Bell’s theorem [2]). We
remark that for Bohr and Heisenberg, such theorems had no real physical value,
since they were sure that Heisenberg’s uncertainty principle is incompatible with the
existence of a subquantum classical structure. The first no-go theorem was proved
by von Neumann in 1933 [15]. It was totally ignored by Bohr and Heisenberg. They
just did not need it.
In fact, the no-go statements are dangerous pitfalls on the way towards applications
of the quantum operational formalism in biology. (This problem is simply ignored
by the majority of researchers, who formally apply the mathematical formalism of
quantum mechanics, e.g., in psychology or cognitive science.) In biology intrinsic
variables definitely exist; e.g., the levels of genes expressions in systems biological
models of cell functioning or the neuronal variables in the brain science. It has been
done a lot with the aid of such variables, e.g., in systems biology. And we cannot buy
the idea of biological nonlocality, e.g., we cannot believe that human beings can be
coupled on the mental level in the absence of physical interactions. It seems that we
have to give up using the quantum operational formalism in biology. It is a powerful
operational formalism describing measurements. However, its usage contradicts the
possibility to introduce intrinsic variables in biological systems.
First of all, we note to the readers who are aware of the debate on hidden variables
in quantum theory (in particular, Bell’s inequality, quantum nonlocality, and so on)
that the standard presentation of the contemporary state of art is not fair at all. The
readers who are not experts in quantum foundations may get the impression that the
problem was completely clarified and any model with local hidden variables would
contradict the predictions of quantum mechanics. This is definitely not the case. The
experimental setup of tests for violation of Bell’s inequality involves a number of
“loopholes” which provide possibilities to violate Bell’s inequality and, moreover, to
obtain quantum correlations for purely classical models. In fact, these loopholes are
1 These statements are not really theorems in the rigorous mathematical sense. They are statements of
the physical nature. Apart from mathematical assumptions (which are typically vaguely formulated),
they contain a list of physical assumptions.
158 9 Foundational Problems of Quantum Mechanics
classical probabilistic models can violate Bell’s inequality and reproduce quantum
correlations, which are typically considered as irreducible to classical correlations.
Of course, it is a separate question whether such models with hidden variables
describe the real physical situation.
However, such models with hidden variables may match well with biological
phenomena. Hence, there is a wide field for studies in foundations of quantum bio-
informatics.
Another “do not be afraid of no-go theorems” argument was presented by Acacio
de Barros [19]. He rightly pointed that the size of a biological system (too small) and
the speed of information processing in it (too slow) are such that the discussion on
the problem of a superluminal action on a distance, quantum nonlocality, is simply
meaningless; see also his work [20] on QL-modeling of cognition.
Finally, we point to a recently developed prequantum model based on representa-
tion of quantum systems by random waves, prequantum classical statistical field the-
ory [21–25], cf. [26]. This model in combination with the measurement theory based
on interaction of random waves with detectors of the threshold type [27] produces
correlations coinciding with quantum correlations. Here detectors can have 100 %
efficiency; cf. Suppes and Acacio de Barros [28]. Applications of this approach to
the creation of QL-representation of brain functioning based on classical electro-
magnetic random waves in the brain were presented in works [29–31]. However,
we point out that in this book we do not advertise any particular model of physical
realization of “brain’s hardware”. We proceed in the operational QL approach and
we are interested, so to say, in “brain’s software.”
References
1. Feynman, R., Hibbs, A.: Quantum Mechanics and Path Integrals. McGraw-Hill, New York
(1965)
2. Bell, J.: Speakable and Unspeakable in Quantum Mechanics. Cambridge University Press,
Cambridge (1987)
3. Khrennikov, A.: Växjö interpretation-2003: realism of contexts. In: Khrennikov, A.Yu. (ed.)
Quantum Theory: Reconsideration of Foundations-2. Series Mathematical Modeling, vol. 10,
pp. 323–338. Växjö University Press, Växjö (2004)
4. Khrennikov, A.: Contextual Approach to Quantum Formalism. Springer, Berlin (2009)
5. Khrennikov, A.: Interpretations of Probability, 2nd edn. De Gruyter, Berlin (2009). (completed)
6. Fuchs, C.A.: Quantum mechanics as quantum information (and only a little more). In:
Khrennikov, A. (ed.) Quantum Theory: Reconsideration of Foundations. Series Mathemati-
cal Modeling, vol. 2, pp. 463–543. Växjö University Press, Växjö (2002)
2 However clear philosophical understanding of the process of transition from individual nonob-
7. Fuchs, C.: Delirium quantum (or, where I will take quantum mechanics if it will let me). In:
Adenier, G., Fuchs, C., Khrennikov, A.Yu. (eds.) Foundations of Probability and Physics-3.
Series Conference Proceedings, vol. 889, pp. 438–462. American Institute of Physics, Melville
(2007)
8. Garola, C., Khrennikov, A.: Dialog on beyond quantum. In: Khrennikov, A. (ed.) Quantum
Theory: Reconsideration of Foundations-5. Series Conference Proceedings, vol. 1232, pp.
355–359. American Institute of Physics, Melville (2010)
9. Boole, G.: An Investigation of the Laws of Thought. Dover Edition, New York (1958)
10. Vorob’ev, N.N.: Consistent Families of Measures and Their Extensions. Theory Probab. Appl.
7, 147–162 (1962)
11. Conte, E., Khrennikov, A., Todarello, O., Federici, A., Zbilut, J.P.: A preliminary experimental
verification on the possibility of Bell inequality violation in mental states. Neuroquantology 6,
214–221 (2008)
12. Bruza, P., Kitto, K., Ramm, B., Sitbon, L., Blomberg, S., Song, D.: Quantum-like non-
separability of concept combinations, emergent associates and abduction. Log. J. IGPL 20(2),
455–457 (2010)
13. Busemeyer, J., Bruza, P.D.: Quantum Cognition and Decision. Cambridge University Press,
Cambridge (2012)
14. Dzhafarov, E.N., Kujala, J.V.: Quantum entanglement and the issue of selective influences in
psychology: an overview. Lect. Notes Comput. Sci. 7620, 184–195 (2012)
15. Von Neuman, J.: Mathematical Foundations of Quantum Mechanics. Princeton University
Press, Princeton (1955)
16. Kochen, S., Specker, E.: The problem of hidden variables in quantum mechanical systems. J.
Math. Mech. 17, 59–87 (1967)
17. Khrennikov, AYu., Volovich, I.V.: Local realism, contextualism and loopholes in Bell’s exper-
iments. In: Khrennikov, AYu. (ed.) Foundations of Probability and Physics-2. Series Mathe-
matical Modeling, vol. 5, pp. 325–344. Växjö University Press, Växjö (2002)
18. Adenier, G., Khrennikov, A.Yu.: Is the fair sampling assumption supported by EPR experi-
ments? Phys. B: At., Mol. Opt. Phys. 40(1), 131–141 (2007)
19. Acacio de Barros, J.: Quantum-like model of behavioral response computation using neural
oscillators. Biosystems 110, 171–182 (2012)
20. Acacio de Barros, J.: Joint probabilities and quantum cognition. In: Proceedings of the Interna-
tional Conference on Quantum Theory: Reconsiderations of Foundations-6. Växjö, Sweden,
11–14 June (2012)
21. Khrennikov, A.: Detection model based on representation of quantum particles by classical
random fields: Born’s rule and beyond. Found. Phys. 39, 997–1022 (2009)
22. Khrennikov, A., Ohya, M., Watanabe, N.: Classical signal model for quantum channels. J.
Russ. Laser Res. 31, 462–468 (2010)
23. Khrennikov, A.: Prequantum classical statistical field theory: Schrödinger dynamics of entan-
gled systems as a classical stochastic process. Found. Phys. 41, 317–329 (2011)
24. Khrennikov, A.: Description of composite quantum systems by means of classical random
fields. Found. Phys. 40, 1051–1064 (2010)
25. Khrennikov, A., Ohya, M., Watanabe, N.: Quantum probability from classical signal theory.
Int. J. Quantum Inf. 9, 281–292 (2011)
26. Ohya, M., Watanabe, N.: A new treatment of communication processes with Gaussian channels.
Jpn. J. Ind. Appl. Math. 3, 197–206 (1986)
27. Khrennikov, A.: Role of detectors and their proper calibration in inter-relation between classical
and quantum optics. Opt. Eng. 51, Article number 069001 (2012)
28. Acacio de Barros, J., Suppes, P.: Quantum mechanics, interference, and the brain. Math. Psy-
chol. 53, 306–313 (2009)
29. Khrennikov, A.: On the physical basis of theory of “mental waves”. Neuroquantology 8, S71–
S80 (2010)
30. Khrennikov, A.: Quantum-like model of processing of information in the brain based on clas-
sical electromagnetic field. Biosystems 105(3), 250–262 (2011)
162 9 Foundational Problems of Quantum Mechanics
31. Haven, E., Khrennikov, A.: Quantum Social Science. Cambridge Press, Cambridge (2012)
32. Plotnitsky, A.: Reading Bohr: Physics and Philosophy. Springer, Heidelberg (2006)
33. Plotnitsky, A.: On the reasonable and unreasonable effectiveness of mathematics in classical
and quantum Physics. Found. Phys. 41, 466–491 (2011)
34. Khrennikov, A.: Towards a resolution of dilemma: nonlocality or nonobjectivity? Int. J. Theor.
Phys. 51, 2488–2502 (2012)
Chapter 10
Decision and Intention Operators
as Generalized Quantum Observables
Abstract Since the classical formula of total probability is violated for data from
both QM and cognitive psychology and the former has developed the advanced math-
ematical apparatus for the study of the interference effect, it is natural to apply this
apparatus to describe statistical data of the latter. However, there is one pitfall; one
problem has to be taken into account and carefully analyzed. The experimentally
obtained matrices of transition probabilities are not doubly stochastic as is not the
case for the matrices corresponding to the transitions from one orthonormal basis to
another (the bases of the eigenvectors of the Hermitian operators representing quan-
tum observables). Therefore “cognitive observables” (“mental observables”) cannot
be represented by Hermitian operators (at least with nondegenerate spectra). A possi-
bility to solve this problem, namely, to use lifting maps, was explored in the previous
chapters. A lifting map describes in detail the interaction of a system with an environ-
ment; this map contains, as parameters, the state of environment, the operator of the
interaction between the system and the environment, and the measurement device.
However, sometimes such a detailed description of the environment can be reduced
to mere representing observables of systems by positive operator valued measures. In
this chapter we demonstrate the power of usage of positive operator valued measures
(POVMs) in biological science on the example of quantum modelling of cognition.
As noted in Chap. 2, since the classical formula of total probability (FTP) is vio-
lated for data from both QM and cognitive psychology and the former has devel-
oped the advanced mathematical apparatus for the study of the interference effect,
it is natural to apply this apparatus to describe statistical data of the latter. How-
ever, there is one pitfall; one problem has to be taken into account and carefully
analyzed. The experimentally obtained matrices of transition probabilities are not
doubly stochastic (Sect. 4.1.1) as is not the case for the matrices corresponding
to the transitions from one orthonormal basis to another (the bases of the eigen-
vectors of the Hermitian operators representing quantum observables). Therefore
“cognitive observables” (“mental observables”) cannot be represented by Hermitian
operators (at least with nondegenerate spectra). A possibility to solve this problem,
© Springer Science+Business Media Dordrecht 2015 163
M. Asano et al., Quantum Adaptivity in Biology: From Genetics to Cognition,
DOI 10.1007/978-94-017-9819-8_10
164 10 Decision and Intention Operators as Generalized Quantum Observables
namely, to use lifting maps, was explored in the previous chapters. A lifting map
describes in detail the interaction of a system with an environment; this map contains,
as parameters, the state of environment σ, the operator of the interaction between the
system and the environment V , and the measurement device M. However, sometimes
such a detailed description of the environment can be reduced to mere representing
observables of systems by positive operator valued measures (POVM). For biologi-
cal applications, it is important that matrices of transition probabilities of POVMs are
in general not doubly stochastic. Therefore it is natural to apply POVMs to describe
biological statistical data. In this chapter we present a QL-model of decision making
based on the representation of the decision and intention operators by POVM.
Although the interrelation between violation of the classical FTP and quantum
interference of probabilities is well studied, see Chap. 2, typically the situation is
restricted to probabilities corresponding to conventional quantum observables and
pure states (Hermitian operators and complex state vectors). The FTP with inter-
ference term for the POVM-observables and the mixed states (density operators)
was derived in [1]. In this case the interference term cannot be parameterized just
by the “interference angle” θ. It contains a few parameters admitting an interesting
cognitive interpretation, see Sect. 10.3.
We model decision making in the two party games: one party, say Bob, has to make
decision on some problem B. The payoff depends on the actions of another party,
say Alice. Therefore Bob has to take into account Alice’s intentions A, see Chap. 6.
Bob’s mental state is described by the density operator ρ, his decision operator
is given by the POVM M b ; Bob also has to analyze possible Alice’s actions, this
“measurement” is represented by the POVM M a . We stress that both the decision
making operator M b and the intention operator M a represent self-measurements, see
Chap. 1 for discussion.
We start with the presentation of the fundamentals of the theory of positive operator
valued measures (POVM). (POVMs were already mentioned at the end of Sect. 1.3
and they were used in Sect. 4.4 to represent “event systems”.)
10.1.1 POVMs
Let H be a finite dimensional complex Hilbert space endowed with the scalar product
·, ·. The space of density operators is denoted by the symbol D(H ).
(We considered the simplest case of a discrete operator valued measure on the set
of indexes J = {1, 2, ..., m}. ) It is convenient to use the following representation of
POVMs:
M j = V j V j , (10.1)
V j ρV j
ρj = . (10.3)
trVj ρVj
pa (α) = trρMa (α) = trV(α)ρV (α), p(β) = trρMb (β) = trW(β)ρW (β).
(10.4)
Now we consider two consecutive measurements: first the a-measurement and
then the b-measurement. If in the first measurement the value a = α was obtained,
then the initial state ρ was transformed into the state
V (α)ρV (α)
ραa = . (10.5)
trV(α)ρV (α)
166 10 Decision and Intention Operators as Generalized Quantum Observables
This is the conditional probability to obtain the result b = β under the condition of
the result a = α.
We set
p(α, β) = pa (α) p(β|α). (10.7)
This is the probability to obtain the result (α, β) in the consecutive measurement of
a and then b. To find p(α, β) with the aid of the probability p(α) and the conditional
probability p(β|α), we apply the classical Bayes formula. Thus “quantumness” is
in the probabilities p(α) and p(β|α); the probability p(α, β) is the result of their
classical combination.
We remark that the probability p(α, β) can be expressed directly in terms of
POVMs:
p(α, β) = trρV (α)W (β)W(β)V(α). (10.8)
This is the probability to obtain the result b = β in the first measurement and then
the result a = α. One can easily find examples of POVMs (and even Hermitian
operators) such that
p(α, β) = p(β, α). (10.10)
We recall (Chap. 2) that, for two classical random variables a and b, which can
be represented in the Kolmogorov measure-theoretic approach, the formula of total
probability (FTP) has the form
p b (β) = pa (α) p(β|α). (10.11)
α
10.2 Formula of Total Probability with the Interference Term … 167
This parameter is equal to the ratio of the ordered joint probabilities of the same
outcome, but in the different order, namely, “b then a” or “a then b”. Then [1]
1 p(α1 , β) p(α2 , β)
λβ = (γα1 β − 1) + (γα2 β − 1) . (10.16)
2 p(α2 , β) p(α1 , β)
In principle, this coefficient can be larger than one [1]. Hence it cannot be repre-
sented as
λβ = cos θβ (10.17)
for some angle (“phase”) θβ . However, if POVMs M a and M b are, in fact, spec-
tral decompositions of Hermitian operators, then the coefficients of interference are
always less than one, i.e., one can find phases θβ .
The transition from the classical FTP to the FTP with the interference term (10.12)
can be considered as an extension of the parameter space. Besides the probabilities
for the results a = α and b = β and the conditional probabilities p(β|α), new
parameters λβ , the coefficients of interference between observables a and b (for the
state ρ), are taken in consideration.
We present an interpretation of the structure of the coefficients of interference. We
start with terms γαβ . They express noncommutativity of measurements or nonclas-
sicality of the joint probability distribution. For classical probability all coefficients
γαβ = 1 (and hence all interference coefficients λβ = 0).
168 10 Decision and Intention Operators as Generalized Quantum Observables
1 √ √
λβ = [(γα1 β − 1) μβ + (γα2 β − 1)/ μβ ]. (10.19)
2
We now present again the schemes of decision making that are based on classical
and QL FTP. In the latter case we discuss the role of “interference parameters” in
QL FTP of Sect. 10.2.
There are two parties, Alice and Bob. They operate under some complex of conditions
(physical, social, financial) C, context. Each can act only in two ways: a = α1 , α2
and b = β1 , β2 . Bob’s decisions depend crucially on Alice’s actions. However, in
general Bob does not know precisely which action a = α1 or a = α2 will be chosen
by Alice. Therefore Bob can only guess. He estimates subjectively the probabilities
pa (α1 ) and pa (α2 ) of her possible actions. He can also estimate probabilities of
his own actions conditioned by Alice’s actions, p(β|α). Finally, Bob estimates the
probabilities p b (β1 ) and p b (β2 ) by using FTP (10.11). If, e.g., p b (β1 ) > p b (β2 ),
then Bob makes the decision b = β1 .
This scheme is realized in the brain on the unconscious level. Subjective estimates
of probabilities for Alice’s actions and conditional probabilities for his own actions
conditioned by her actions typically are not present on the conscious level.
the estimate of the effect of noncommutativity of Bob’s and Alice’s actions and
relative magnitudes of probabilistic influences of Alice’s actions on Bob’s actions.
To estimate γαβ , Bob has to estimate not only the ordered probability p(α, β) of his
own action b = β under the condition that Alice would act as a = α, but also the
ordered probability p(β, α) of possible Alice’s action a = α under the condition that
she assumes Bob’s action b = β. However, since the absolute magnitudes of these
probabilities are not important, Bob is interested only in their ratio.
Finally, Bob estimates probabilities of his actions by using the FTP with the
interference term (10.12).
This scheme is also realized on the unconscious level. In particular, all the afore-
mentioned measures of quantumness are estimated subjectively (on the basis of
collected experience).
We can say that in the QL-scheme the counterfactual arguments play a crucial
role. To estimate the coefficients of noncommutativity γαβ , Bob has to imagine how
Alice would act if he acted as b = β. These probabilities are compared with the
probabilities of Bob’s own actions conditioned by possible Alice’s actions. If Bob
guesses that Alice makes the same estimation of the probabilities of his actions
conditioned by her actions as for her own actions conditioned the corresponding
Bob’s actions, then p(α, β) = p(β, α) and γαβ = 1. In this case the QL and
classical FTP and, hence, the decision making schemes coincide. Thus in this decision
making scheme quantumness is related asymmetrically in the estimation of the action-
reaction probabilities. Bob can imagine that Alice is friendlier than him or that Alice
is more defect-inclined (we consider the Prisoner’s Dilemma game as the basic
example of the situation under consideration).
In the operational QL-formalism all probabilities and components of the coeffi-
cients of interference are encoded in the Bob’s mental state, the density operator ρ,
his decision making observable, POVM M b , and his observable for Alice’s inten-
tions, POVM M a (We remark that these are self-observables).
Reference
1. Khrennikov, A., Loubenets, E.: On relation between probabilities in quantum and classical
experiments. Found. Phys. 34, 689–704 (2004)
Index
A Chemical modification, 52
Acacio de Barros, 159 Chromatin-mark, 145
Adaptive dynamics, 6, 38 Chromatin-marking, 135
Adaptive force, 127 Codon, 45
Additivity, 14 Collapse, 23
Adjoint operator, 29 Completeness of quantum mechanics, 156
Algebra of sets, 14 Complex conjugation, 20
Allosteric, 52 Complexifying force, 128
Ancestor, 128 Complexity, 128
Anticommutator, 143 Complex numbers, 20
Archaea, 131 Computational biology, 31
Average, 15 Conditioned reflex, 128
Context of measurement, 4
Covariance, 16
B CRISPR-Cas system, 131
Bacteria, 131
Bateson, 134
Bayes’ formula, 17 D
Bayesian approach, 7 Darwin, 128
Bayesian updating, 107, 108 Darwinian, 53
Bell’s inequality, 156 Darwinism, 144
Biological nonlocality, 157 Decision making, 164
Bio-nonlocality, 160 Decision operator, 164
Bohr, 157 Decoherence, 29, 138
Borel sets, 15 Delbruck, 132
Density operators, 21
Detectors inefficiency loophole, 158
C Diauxie, 43, 49
Canalization model, 151 Differentiation, 9
Cartesian product, 32 DNA methylation, 53, 135
Catabolite repression, 49 Double stochasticity, 64
Cell cognition, 5 Doubly stochastic, 164
Cell differentiation, 5, 75
Cells signaling, 129
Cellular epigenetic inheritance, 135 E
Central dogma, 42, 43 Egg, 129
Channel, 68 Embryogenesis, 9
© Springer Science+Business Media Dordrecht 2015 171
M. Asano et al., Quantum Adaptivity in Biology: From Genetics to Cognition,
DOI 10.1007/978-94-017-9819-8
172 Index
I
Incompatible observables, 8 N
Inducerexclusion, 49 Natural selection, 129
Index 173
Neo-Darwinism, 138 S
Neo-Lamarckism, 128, 138 Saltations, 133
Neo-Larmarckism, 135 SAT algorithm, 6
Non-Kolmogorovian model, 8 Schrodinger’s cat, 139
Non-Markovian, 37 Scrapie, 9
Self-measurement, 5, 24
Signal transduction, 42, 47
O Single probability space, 156
Objective property, 159 Somatic cell, 129
Observable-adaptive, 6 Sperm, 129
Ogryzko, 132, 139 State-adaptive, 6
Open quantum systems, 29, 137 Statis, 134
Operational nonlocality, 149 Steady state, 29
Stem cell, 75, 79
Stochasticity, 64
P Subjective probability, 107
Pangenesis, 129 Superconducting Transition-Edge Sensors,
Partition, 19 158
Party governance, 31 Superposition, 22, 29
Phase, 63 Suppes, 159
Phenotype, 31 Systems biology, 51, 157
Phosphotransferase, 49
Planck constant, 27
T
Positive operator valued measures, 3, 63, 164
Tensor product, 31
Prion, 75
Theorem of Bell, 157
Probabilistically incompatible, 8
Theorem of Kochen-Specker, 157
Probability, 14, 15
Theorem of von Neumann, 157
Probability distribution, 15
Time window loophole, 158
Probability manifolds, 8
Transiliencies, 133
Probability space, 15
Transition probabilities, 64
Projection postulate, 23
Transition probability, 18
Protein, 43, 80
Tree of Life, 131
Protein folding, 46
Two slit experiment, 8
Punctuated equilibrium, 134
Pure state, 20
U
Unfair sampling, 158
Q Unitary operator, 28
Quantum bio-informatics, 9
Quantum biology, 9
Quantum brain, 9 V
Quantum jump, 144 Virus, 131
Quantum-like, 9 Von Neumann, 157
Quantum-like representation algorithm, 22 Von Neumann equation, 28, 37
Quantum master equation, 29
Quantum nonlocality, 156
Qubit, 21, 26 W
Waddington, 151
Wallace, 129
R Watson and Crick, 41
Random variable, 15, 16 Wave function, 26
Regulatory loop, 135 Weismann, 129
RNA world, 43 Weismann’s barrier, 129