TWO ROADBLOCKS OF COMPUTATIONALISM 163


Volume 20, 2: 2019

TWO ROADBLOCKS OF COMPUTATIONALISM

Napoleon M. Mabaquiao, Jr.

De La Salle University, Manila

With its use of the powerful technology of computer, the computational

theory of mind or computationalism, which regards minds as computational
systems, has been widely hailed as the most promising theory that will
carry out the project of explaining the workings of the mind in purely
scientific terms. While it continues to serve as the primary framework for
scientifically inclined theorizing and investigations about the nature of
minds, especially in the area of cognitive science, it, however, continues to
face strong objections from its critics. And with the growing complexity
and sophistication of the arguments used to promote and reject the theory,
the debate has become intractable. It has become quite difficult to assess
which side of the dispute is gaining the upper hand. Such difficulty may be
due to a variety of reasons. In this essay, I critically examine two of such
reasons. The first concerns the ambiguity of the theory’s intended scope of
application: whether it is limited to the mind’s cognitive features only or it
also includes the mind’s phenomenal features. The second concerns the
vagueness of how the so-called computer modelling of human cognitive
processes is able to duplicate such processes. Accordingly, if insufficiently
addressed, they remain as two roadblocks to the entire project of
computationalism.

INTRODUCTION
With the continuous development of computer technology, computationalism
remains as the dominant and most promising framework for naturalizing the mind,
referring to the project of explaining the workings of the mind in purely scientific terms
(see Jerry Fodor 1991, 489; David Chalmers 1993, 18). While most scientifically inclined
investigations done on the mind, especially those done in the areas of cognitive science
and artificial intelligence, continue to have it as their overarching framework,
computationalism, nonetheless, remains controversial as a philosophical theory of mind.
It continues to face strong objections from its critics, which its supporters relentlessly
respond to (see, for instance, Milkowski 2017). Now with the growing complexity and
sophistication of the arguments of both proponents and critics of the theory, it has

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
164 NAPOLEON M. MABAQUIAO, JR.

become quite difficult to assess which side of the dispute is gaining the upper hand.
Inthis essay, among the possible reasons for this difficulty, I critically examine the
following two. The first concerns the intended scope of application of the theory, whether
it includes all features of the mind or only some of them. As a philosophical theory of
mind, it seems natural to suppose that it is an account of the mind in general, or, at least,
of the essential features of the mind. But this is far from clear. There is, in fact, an
ambiguity whether what is claimed to be computational about the mind is limited only to
the mind’s cognitive features or likewise includes the mind’s phenomenal features. The
second concerns the disagreement on what the computer modelling of human cognitive
processes is able to accomplish in so far as establishing the thesis of computationalism
is concerned. In particular, the computationalist claim that the computer simulation of
human thought processes duplicates such processes remains controversial.
On the whole, I will try to show in this essay that as long as these two reasons
remain insufficiently addressed, the computationalist thesis that minds are computers
will be controversial as ever. They, in effect, will remain as two roadblocks to the entire
project of computationalism. I shall divide my discussion into three parts. To put things
in perspective, I shall provide a brief background of the computational theory of mind in
the first part. In the second part, I shall examine how much of the mind’s capacities are
being claimed to be computational by computationalism. In the third, I shall deal with
the alleged role of the computer modelling of human cognitive process in advancing the
theses of computationalism.

COM PUTATIONALIS M AND THE PHILOS OPHY OF M IND

The computational theory of mind, or computationalism for short, is a philosophical
theory of mind that claims that minds are computers or computational systems (Rescoria
2017, 1; Milowski 2018b, 1). To better understand this theory, let us situate it in the
general discussion on the nature of the mind’s existence. Views on the nature of mind’s
existence are generally divided into two kinds: the non-materialist views, which attribute
a non-physical existence to minds; and materialist views, which attribute a physical
existence to minds. Non-materialist views include idealist and (substance) dualist views
about the mind. On the other hand, materialist views are generally divided into non-
realist (or irrealist) materialist views, which reject the idea that mental states have
separate existence or distinct reality from the physical states of either the brain or the
body; and realist materialist views, which maintains said idea.
Under non-realist materialism are the following views: the (mind-brain) identity
theory, which reduces mental states to brain states; behaviorism, which, in its strong
version, reduces mental states to behavioral dispositions; eliminative materialism,
which claims that the theory that posits the existence of minds is outdated and mis-
taken; and instrumentalism, which claims that the attribution of mental states to an
entity is purely metaphorical being a mere convenient device for predicting the entity’s
behavior. Under realist materialism, on the other hand, are the following views:
functionalism,which regards mental states as states occurring on the level of a system’s

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 165

functional organization and as definable in terms of their causal roles; computationalism,

which regards mental states as computational states or as states of a computational
system; biological naturalism, which regards mental states as higher-level biological
states; and the quantum view of consciousness, which regards mental states as quan-
tum states in the brain. (See Mabaquiao 2012, chap. 2.)
Computationalism, in its philosophical form, is a development from functionalism.
For this reason, let us look into some of the details of functionalism. Functionalism
regards mental states as functional states. But what exactly are functional states; and
how are they different from the physical states of the brain or the body? Functional
states have three key features. First, they refer to the physical states of a physical
system on the level of the system’s functional organization (see Putnam 1991), which in
turn refers to the specific way that the materials making up the system are arranged to
perform certain functions. An example is the way the parts of a clock are arranged to
perform the function of telling time. With regard to the human brain, its functional
organization refers to the way the brain’s neurons are arranged in order for the brain to
perform its various functions. The brain’s functional states thus do not refer to the
physical states of the brain’s neurons but to the physical states of its functional
organization.
The second key feature of functional states is that they are definable in terms
of their causal roles as input, output, and intervening internal states of a physical
system (see Block 1991, 211-212). On this point, functionalism converges with the causal
theory of mind developed by David Lewis (1991a, 1991b) and David Armstrong (1968,
1991) in the course of justifying the (mind-body) identity theory. Causally defined, a
functional state is caused by an input state B, is causally interacting with an intervening
internal state C, and causes an output state D.
The third key feature of functional states is the principle of multiple realizability,
according to which the same functional states can occur in physical systems consisting
of different materials but having the same functional organization. This makes functional
states substrate neutral, in that they are not affected by the nature of the materials that
make up the system. And being functional states, mental states are, thus, substrate
neutral in that while they occur in brains, they can also occur in other physical systems
having the same causal or functional organization as the human brain.
Functionalism has developed into several forms. For our purposes, let us
distinguish between the following two. The first version is general functionalism, which
is non-committal as to the kind of physical system deemed appropriate for mental states.
It does not specify the particular type of system that mental states can occur. The
second is computer functionalism, advanced by Hilary Putnam (1991), which specifies
that the particular type of physical system appropriate for mental states to occur is one
that instantiates the Turing machine. The Turing machine was conceived by Alan Turing
(1936) originally to resolve foundational issues in mathematics but which later on also
served as the abstract model of the modern-day digital computer. In this consideration,
what machine functionalism in effect claims is that mental states can occur only
incomputing machines, which then makes the human brain as a kind of computer. Needless

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
166 NAPOLEON M. MABAQUIAO, JR.

to say, it was this version of functionalism that paved the way for the development of
computationalism (Rescoria 2017). In this consideration, computationalism is sometimes
referred to as “computer functionalism” or “computational functionalism,” implying
that computationalism is a version of functionalism.
Computationalism contends that the functional states with which functionalism
identifies mental states are of the computational kind. Being so, computational states
also have the key features of functional states but with some modifications. First, while
functional states are the physical states of a physical system on the level of the system’s
functional organization, computational states, in its classical form, are the physical
states of a computing system on the level of software. The mind is treated as the
software of the computational system, while the brain as its implementing hardware.
Mental states, on this view, specifically refer to the states of the mind software when
being run or implemented by the brain hardware. Second, as functional states can be
defined in terms of the causal relations of inputs, intervening internal states, and outputs,
so are computational states. And third, as functional states are multiply realizable, so
are computational states. But while in functionalism this means that two physical systems
consisting of different materials but having the same functional organization will have
the same functional states; in computationalism, this would mean that two computers or
computing systems different in their hardware materials (though the same in
sophistication in terms of software-implementing capacities) but the same in their
software will have the same computational states.
In addition to the machine functionalism of Putnam, computationalism further
developed when Jerry Fodor (1979), in collaboration with Zenon Pylyshyn (1990), linked
it with his representational theory of mind (also known as the language of thought
hypothesis). According to the representational theory of the mind, the brain has an
inherent language or system of representation that has a language-like structure called
the “language of thought,” which serves as the vehicle or medium of the mind in
performing its computations or thinking operations. This language of thought consists
of mental symbols or internal representations which the mind manipulates according to
rules inherent in the brain.
While all this was happening in the discipline of philosophy, a parallel development
was occurring in the discipline of artificial discipline, a branch of computer science
devoted to the construction of intelligent machines. In the discipline of artificial
intelligence, a distinction was made between the general and neutral claim that computers
are powerful tools for understanding the workings of the human mind and the specific
and bold claim that human minds are themselves computer programs. The former has
been called “Weak AI,” while the latter “Strong AI” (such terms are due to John Searle
1980). A major influence in the development of Strong AI was the physical symbol
system hypothesis introduced by Allen Newell and Herbert Simon, according to which
“a physical symbol system has the necessary and sufficient means for general intelligent
action” (Newell and Simon 1976, 116). On this view, intelligent action is possible only
for a physical system that computes in terms of symbols. Human minds and computers
are such physical symbol systems, which explain their capacity for intelligent actions.

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 167

Computationalism, in this regard, has then been closely associated, if not equated, with
Strong AI.
Computationalism has evolved in different forms. The kind of computationalism
that we have described above, the initial version of computationalism, has been called
symbolic or classical computationalism. It is described as symbolic since it defines the
computing that occurs in thinking as a process of manipulating symbols or
representations according to some rules. It is described as classical, on the other hand,
in recognition of the fact that it was in this form that computationalism became a popular
and dominant theory of mind. The general idea of computationalism, that thinking is a
kind of computing, may have its roots to the ideas of philosophers before classical
computationalism (such as Leibniz), but it was during the introduction of classical
computationalism that computationalism became an influential theory of the mind.
While having its roots also in pre-classical computationalism, another influential
form of computationalism developed partly as a reaction to classical computationalism,
namely the view called the artificial neural network approach, also known as
connectionist computationalism or simply as connectionism. This approach uses a
computer model that is based on the neural networks of human brains to explain the
computing processes of the mind. It explains such processes as interactions among the
units of networks. There are also variations within this approach, and one popular
version is the parallel distributed processing network developed by James McClelland
and David Rumelhart, among others. Furthermore, if for the symbolic approach computing
is serial (that is, a step-by-step process and hence one operation occurring at a time),
for connectionism it is parallel (that is, simultaneous operations occurring at the same
time) (see Crane 1995, 154-162). There is an on-going debate among the proponents of
each approach on which of these two is the correct or superior approach to
computationalism (see, for instance, Fodor and Pylyshyn 1990, Smolensky 1993,
Rumelhart 1990, and McClelland and Rumelhart 1993). Some prominent AI scientists,
such as Marvin Minsky (1995, 649), however, has called for a synthesis of these two
approaches. In either form, however, the basic thesis of computationalism remains, that
thinking is a species of computing and that mental states are computational states.
Another theory of mind that has recently emerged that is associated with
computationalism is the so-called computational neuroscience, which claims that it is
the brain, not the mind, that is a computer (Rescoria 2017, 7). Computational neuroscience
is somewhat of a merger between the identity theory and computationalism. As the mind
is nothing but the brain, then it is the brain that is more appropriately a computer. Some,
however, believe that computational neuroscience more of a method than a theory—as
Wein and Pona (2015, 2) states: “Computational neuroscience is a branch of neuroscience
that uses computer simulations to study the brain; it is not a theory of mind, but a
method of studying the brain.” In any case, acknowledged key figures in the development
of computer neuroscience includes Patricia Churchland (1986), Paul Churchland (1995,
2007), and Chris Eliasmith (2013). This theory specifically claims that the human brain
processes themselves, not the physical states of the functional organization of the
brain nor the software allegedly being implemented by the brain hardware, are the

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
168 NAPOLEON M. MABAQUIAO, JR.

computational states. Unlike the classical and connectionist computational theories,

computational neuroscience, by grounding computationalism in biology, abandons the
thesis of multiple realizability (Wein and Pona 2015, 2). This means that their findings
cannot be made as bases for constructing artificial or machine intelligence or a silicon-
based creature. More importantly, it is susceptible to the same objections levelled against
the identity theory, especially the charge of “neuro-chauvinism”—that mentality is
exclusive to creatures whose brains consist of neurons.
Computationalism has been adopted as the main framework of cognitive science,
referring to the projected ultimate science of the mind. While multidisciplinary in
approach for utilizing the methods and findings of several disciplines that include
philosophy, computer science, psychology, anthropology, biology, neuroscience, and
linguistics, cognitive science uses computationalism as its overarching framework for
its scientific investigations about the mind. As Milkowski (2018b, 1) writes: “It is
generally assumed that CTM is the main working hypothesis of cognitive science.”
And as Freidenberg and Silverman (2006, 2) explain: “In order to really understand what
cognitive science is all about we need to know what its theoretical perspective on the
mind is. This perspective centers on the idea of computation, which may alternatively
be called information processing” (see also Howard Gardner 1985, 6-7, 384-85; and
Robert Harnish 2002, 2-3).

THE S COPE OF THE COMPUTATIONAL MIND

How much of the mind’s capacities are claimed to be computational by
computationalism? To facilitate our analysis, let us distinguish among three kinds of
mind: the general, cognitive, and phenomenal minds. The general mind is the mind
construed in terms of both its cognitive and phenomenal features. The cognitive mind
is the mind construed in terms of its cognitive features only. And the phenomenal mind
is the mind construed in terms of its phenomenal features only. When computationalists
claim that the mind is a computer, which among these minds are they referring to? Some
believe that it should refer to the general mind. As Wein and Pona (2015, 2), referring to
computationalism as CTM, states: “CTM is supposed to be a universal theory of mind:
it should be possible to explain every aspect of the mind by providing a suitable
computational description ….” But some believe that it only refers to the cognitive
mind, thereby excluding the phenomenal mind. As Milkowski (2018b, 2) notes: “The
generic claim that the mind is a computer may be understood in various ways, depending
on how the basic terms are understood. In particular, some theorists claimed that only
cognition is computation, while emotional process are not computational (Harnish 2002,
6), yet some theorists explain neither motor nor sensory processes in computational
terms (Newell and Simon 1972).”
The classic objections to functionalism contend that functionalism fails as a theory
of mind because it leaves out certain fundamental features of the mind. Such arguments,
for instance, point out that functionalism leaves out the mind’s phenomenal features.
These arguments include the China brain argument (see Block 1991, 211-229), knowledge
argument (see Jackson 1991, 291-295), and inverted qualia argument (see Schoemaker

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 169

1982, 357-381). As computationalism is a variant of functionalism, these objections

equally apply to it. Searle’s Chinese room argument (see Searle 1980, 417-457), on the
other hand, points to the failure of computationalism to account for an important feature
of the cognitive mind, namely its inherent intentionality. The previous arguments all
point out that computationalism fails because it leaves out the phenomenal mind. Searle,
however, in effect shows that even when limited to the cognitive mind, computationalism
still fails because it leaves out the intentional mind. The presupposition of these
objections is that computationalism, to be successful, should account for the general
mind.
Defenders of computationalism usually respond to these arguments in a variety
of ways. One is by exposing the flaws in the reasoning of these arguments. An example
of this kind of reaction is Chalmers’s fading and dancing qualia arguments, which are
intended to show the implausibility of the absent and inverted qualia arguments. Another
is by rejecting the assumption that certain features of the mind, such as qualia and
intentionality, are really essential to mentality. A classic example of this kind of approach
is Dennett’s instrumentalist theory of mind, according to which the attribution of
intentionality, or consciousness in general, is merely metaphorical for it is done solely
for its usefulness in predicting future behaviors of organisms (see Dennett 1989, 1991).
Still another is by clarifying that computationalism is not really intended to account for
the general mind (which includes the intentional mind) or the phenomenal mind, for it is
only intended to account for the cognitive mind. In this respect, some computationalists
contend that these “objections fail because they make computationalism a straw man”
(Milkowski 2018a, 58).
But even if we grant the construal of computationalism as a theory of the cognitive
mind, thereby restricting mentality to intelligence, there is still the problem of how much
of the cognitive mind computationalism is intended to account for. We earlier saw that
this was the main point of Searle’s Chinese room argument, showing that the intentional
mind is left out by computationalism. Let us further elaborate on this. When
computationalists speak of the mind, they usually mean intelligence or the cognitive
aspect of the mind. What they claim to be computational are only the mind’s cognitive
states (the propositional attitudes). This means that emotions or the affective aspect of
mentality, at least originally, are not part of the computationalist project. Now, in the
case of humans, we normally understand intelligence both in terms of functionality and
consciousness. The functionality of intelligence generally refers to abilities or capacities
to perform certain tasks such as solving problems, performing operations, answering
questions, and following rules. We say, for instance, that someone is intelligent if she
can solve certain mathematical problems or perform certain mathematical operations. Its
consciousness, on the other hand, generally refers to awareness of having certain
mental states or processes such as understanding and reasoning. We normally think
that both aspects should be present for someone to be truly intelligent. If someone
claims to understand something but cannot perform the associated task, we doubt
whether she is really intelligent. Or she actually performs the task but solely because
she has been conditioned to do so, we also doubt whether she is really intelligent.

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
170 NAPOLEON M. MABAQUIAO, JR.

The kind of intelligence, however, that computationalists speak about is a general

one, one that does not just apply to humans but to machines as well. It seems obvious
that machines, in principle, can share the functional aspect of human intelligence; but
it is not clear whether they can also share its conscious aspect. The consideration of
the possibility of machine intelligence, thus, puts the fundamentality of consciousness
in the definition of intelligence into question. Three views can be distinguished
concerning the relative importance of consciousness and functionality to the nature of
intelligence or mentality in general. The first is the default view, which we shall refer to
as the ordinary view, which regards functionality and consciousness as equally
fundamental in defining intelligence. The second, which we shall call the purely
functional view, regards the mind’s functionality as adequate in defining mentality. The
third, which we shall call the purely conscious view, regards the mind’s consciousness
as adequate in defining intelligence. In the contemporary scene, the debate is mainly
between the purely functional view and the ordinary view. The purely conscious view,
which can be attributed to the idealists and substance dualists, is generally no longer
regarded as a strong contender. For as contemporary philosophy of mind is geared
towards the naturalization of the mind, the fundamentality of functionality is taken as a
given and the only question is whether or not consciousness is equally fundamental.
Consequently, we can adopt either a generalized construal or a specialized construal
of the claims of computationalism. We adopt a generalized construal of these claims
when we attribute a normal view to these claims; that is to say, we understand these
claims as applying to both functionality and consciousness of minds. On the other
hand, we adopt a specialized construal of these claims when we attribute a purely
functional view to these claims; that is to say, we understand these claims as only
applying to the functionality of minds. The inadequacy arguments obviously work only
under a generalized construal of the claims of computationalism; they are misplaced
under a specialized construal of the computationalist claims.
Understandably, most proponents of computationalism take a specialized construal.
In a footnote in his book The Mind Doesn’t Work That Way, Jerry Fodor (2000, 1), for
instance, writes: “This is not to claim that CTM is any of the truth about consciousness,
not even when the cognition is conscious. There are diehard fans of CTM who think it
is; but I’m not of their ranks” (CTM refers to the Computational Theory of Mind).
Herbert Simon and Craig Kaplan (1990, 1-2) also write: “Intelligence is to be judged by
the ability to perform intellectual tasks, independently of the nature of the physical
system that exhibits this ability.” Still, Roger Schank and Peter Childers (1984, 51) write:
“When we ask What is intelligence? we are really only asking What does an entity,
human or machine, have to do or say for us to call it intelligent.”
On the other hand, most of the critics of computationalism, such as John Searle
and Roger Penrose, prefer a generalized one. Penrose (1989, 525-26), for instance, writes:
“There is also the question of what one means by the term ‘intelligence’. This, after all,
is what the AI people are concerned with, rather than the perhaps more nebulous issueof
‘consciousness’.... In my own way of looking at things, the question of intelligence is a
subsidiary one to that of consciousness. I do not think that I would believe that true

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 171

intelligence could be actually present unless accompanied by consciousness.” In his

Chinese room argument, Searle (1980) objects to the claim that machines that are able to
simulate the intelligent behavior of humans are genuinely intelligent themselves. Searle
explains that in the case of humans, there is awareness of what their mental states mean
or represent in the world; while in the case of machines they just manipulate symbols
according to the rules of the given program without any awareness of what these
symbols mean. This means, for Searle, consciousness (in the form of intentionality) is
as fundamental to intelligence as the mind’s functionality.
In some cases, however, supporters of computationalism themselves are not clear
which construal they adopt as they seem unsure about the relevance of consciousness
to the computationalist project. In her informal survey, Drew McDermott (2007) shows
that most serious AI researchers are ambivalent on the importance of consciousness in
the computationalist project. She (2007, 119) writes: “Although one might expect AI
researchers to adopt a computationalist position on most issues, they tend to shy away
from questions about consciousness.” Interestingly, she herself, as a supporter of
computationalism, seems ambivalent on the subject matter. She counts herself as one
of those who regard consciousness as unimportant to the computational project; thus
she (2007, 119) writes: “When it comes to the problem of phenomenal consciousness,
however, the AI researchers who care about the problem and believe that AI can solve
it are a tiny minority … I count myself in that minority ….” But then she (2007, 119)
believes that despite the various objections hurled against it, “the basic computationalist
working hypothesis survived intact: that the embodied brain is an ‘embedded’ com-
puter, and that a reasonably accurate simulation of it would have whatever mental
properties it has, including phenomenal consciousness.” So it seems that conscious-
ness is important after all; otherwise, why emphasize the point that consciousness
would be part of what would be duplicated in a computer simulation of the brain.

THE ROLE OF THE COMPUTER MODELLING OF THE MIND

Milkowski (2013, vii) writes: “The mind can be explained computationally

because it is computational…. My central claim reflects my adherence to realism: a
computational account of the mind can constitute a genuine explanation only insofar as
the mind is itself computational.” If the mind can be explained computationally and the
only way that it can be done, aside from the sufficient sophistication of the needed
technology to carry it out, is if the mind itself is computational, then a computational
explanation of the mind would constitute a powerful argument for the computationality
of the mind. There are, however, questions here. One, how does one establish that the
only way one can computationally explain the mind is that if the mind itself is
computational? Isn’t it possible that the mind is non-computational despite the fact that
it lends itself to a computational explanation? Isn’t it possible that the computational
explanation of the mind is just a convenient and practical way (a convenient stance, ,
following the language of Dennett) of accounting for the various activities and
manifestations of the mind?

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
172 NAPOLEON M. MABAQUIAO, JR.

Another way of stating the view that the only way a computer simulation of the mind
can be done is if the mind itself is computational is that the computer simulation of the
mind duplicates, not just merely simulates or models, the mind itself. As Herbert Simon
(1995, 676) straightforwardly puts it: “a computer simulation of thinking thinks.” Simon
(1995, 676), in what follows, explains how this is so by distinguishing between a computer
simulation of digestion which does not duplicate digestion and a computer simulation
of the mind which duplicates the mind: “The materials of digestion are chemical
substances, which are not replicated in a computer simulation. The materials of thought
are symbols—patterns, which can be replicated in a great variety of materials (including
neurons and chips), thereby enabling physical symbol systems fashioned of these
materials to think.” In gist, the reasoning of Simon is that a computer simulation of
human thought duplicates human thought because the materials of human thought are
symbols which can be duplicated by a computer. Put in another way, the computer’s
symbol-manipulating process can duplicate human thought because human thought is
itself a symbol-manipulating process. With some computationalists who are not inclined
to understand thinking in terms of symbol manipulation but in terms of information
processing, the argument of Simon can put as follows. Computers process information,
and so do minds. When computers simulate the information-processing activities of
minds, these simulations are themselves information-processing activities.
What enables, it shall be observed, the computer simulation of thinking to duplicate
thinking is that the thinking process is multiply realizable, in that it is substrate neutral
or that is independent of the physical medium in which it occurs. It is unlike the process
of digestion which is biologically dependent in that it cannot occur without the necessary
biological elements. Referring to the principle of multiple realizability as the principle of
organizational invariance, Chalmers (2010, 37-38), in what follows, seems to echo
Simon in explaining that what a computer simulation duplicates about human thought
includes consciousness:
In general, if a property is not an organizational invariant, we should not
expect it be preserved in a computer simulation (a simulated rainstorm is not
wet). But if a property is an organizational invariant, we should expect it be
preserved in a computer simulation (a simulated computer is a computer). So
given that consciousness is an organizational invariant, we should expect a
good enough computer simulation of a conscious system to be conscious,
and to have the same sorts of conscious states as the original system.

Accordingly, if one denies the principle of multiple realizability, arguing that

thinking is not substrate neutral, then the computer simulation of minds can merely
simulate minds, never duplicate them. Wein and Pona (2015, 2) state this point clearly as
follows:

It is uncontroversial that computation as understood in computer science

can be implemented via different media. Thus, if mind is Turing-style
computation, it can be realized not only in brain, but also using, for instance,

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 173

silicon chips. Clearly, for Artificial Intelligence only cognitive science

theories that imply Multiple Realizability are of interest. Later we will see
that in relation to consciousness, the rejection of Multiple Realizability thesis
is one of the main argument strategies against CTM.

And precisely, this is Searle’s line of reasoning when he argues that a computer
simulation of minds only simulates and never duplicates minds. Searle (1980, 29) contends
as the computer simulation of digestion is not itself a process of digestion (or the
computer simulation of a hurricane is not itself a hurricane), the computer simulation of
human thinking process is not itself a thinking process. In short, he, contra Simon,
argues that a computer simulation of thinking does not think, or, contra Chalmers, a
computer simulation of a conscious entity is not itself conscious. And this is because
Searle does not believe that thinking or minds are substrate neutral. In his biological
naturalism, he argues that consciousness is a higher-level biological phenomenon,
making the biological elements of consciousness essential to the occurrence of
consciousness. Be this as it may, the issue then of whether a computer simulation of
minds merely simulate or duplicate minds is a crucial, if not the crucial, issue to deal with
in assessing the thesis of computationalism that minds are computers. For if such
simulations do not duplicate minds, then it may be the case that while minds lend
themselves to such simulations they are non-computational. Though his point about
Turing is controversial, Rescoria (2017, 3) expresses the tension regarding the role of
computer modelling as applied to minds in what follows:

Formalization and computation are thus closely related, and together yield
the result that reasoning that can can be formalized can also be duplicated
(or simulated) by the right type of machine. Turing himself seems to have
been of the opinion that a machine operating this way would literally be
doing the same things that human performing computations is doing—that
it would be ‘duplicating’ what the human computer does. But other writers
have suggested that what the computer does is merely a ‘simulation’ of
what the human computer does….

Let us look more closely at what transpires in the so-called computer simulation of
minds? According to Anders Sandberg (2003, 3): “Simulations are processes that mimic
the relevant features of target processes… A computer simulation is an attempt to
model a particular system by creating a software representation that represent objects,
relations and dynamics of the system in such a way that relations between objects in
the simulation map onto relations between equivalence classes of objects in the original
system.” Now what about the computer simulation of human thought or intelligence?
How is it done? But first, what is really being simulated in this kind of simulation? There
seems to be three possible answers: first, thinking itself or the thought processes
themselves; second, the brain activities that correlate with thinking processes; and
third, intelligent behaviors or behaviors regarded as manifestations of thinking.

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
174 NAPOLEON M. MABAQUIAO, JR.

It seems apparent that thinking itself, being epistemically subjective (that is, being
directly knowable only by its bearer), is never directly simulated. Mind simulation is
merely inferred from either brain simulation or behavior simulation, which are the ones
that can be done directly. Simon (1984, 25), speaking in behalf of the computationalists,
wrote: “we all believe that machines simulate human thought.” Observe, however, how
he, along with Newell in their 1963 essay “GPS, A Program that Simulates Human
Thought,” explained how simulation of human thought proceeded in their research:
“We…conceive of an intelligent program that manipulates symbols in the same way
that our subject does—by taking as inputs the symbolic logic expressions, and producing
as outputs a sequence of rule applications that coincides with the subject’s…. If the fit
of such program were close enough to the overt behavior of our human subject…then
it would constitute a good theory of the subject’s problem solving” (Newell and Simon
1995, 419-450). The main point seems to be that if a computer program enables a machine
to perform certain actions which when performed by humans are considered intelligent,
such as solving logic problems, then such a program has allegedly succeeded in
simulating the thinking process of humans when performing the said actions.
We can gather from the explanation of Newell and Simon above that there are two
levels of simulation that take place. On the first level is the computer simulation of
human intelligent behavior; on the second level is the computer simulation of human
thought. The first level is a simulation of external outputs (the external outputs of a
computer simulate the behavioral outputs of the human mind) while the second level is
a simulation of internal processes (the computer software simulates the “human
software”). The first level is a direct kind of simulation while the second level is an
inferred one—inferred from the first level. What happens here is similar to our knowledge
of other minds. We cannot directly know the mental states of other persons; but we can
infer them based on the similarity of their behavior with ours when we are expressing
our own mental states. In any case, the point here is that human thought is never
directly simulated; its simulation is inferred, in the case of Newell and Simon, from a
behavioral simulation.
But before we deal with simulation on the level of intelligent behavior, let us look
into the simulation on the level of brain activities. It may be thought that the way to
computationally simulate human thought is by a computer simulation of the human
brain. If the brain activities are constitutive of mental activities, and brain activities are
computational (the main claim of computational neuroscience), then the computer
simulation of brain activities is a direct simulation of mental activities (see the discussion
of Milkowski on how neuroscience studies brain computation, Milkowski 2018a, 530-
32). It is the brain itself that is the computer or that does the computing, unlike in the
classical model where the brain is just the implementing device of the mind software that
does the computing. As McDermott (2007, 145) writes: “What I argue is that the essence
of computationalism is to believe (a) that brains are essentially computers; and (b)
digital computers can simulate them in all important respects, even if they aren’t digital
at all.” [My italics] The problem here is in treating the brains as the computers (not their
mental software) this will abandon the principle of multiple realizability for it will make

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 175

the cognitive process or information processing substrate relative in that it will be

biologically determined. Since computer modelling is done by a machine, which is not of
the same biological stuff as the human brain, how can then we claim that the computer
simulation done by the machine of the human biological brain duplicates the
computational process of the brain? If the computational process of the biological brain
is not multiply realizable, then it cannot be duplicated by a non-biological system.
Let us now examine the computer simulation done on the level of intelligent
behavior. Roger Schank and Peter Childers contend that the reasoning behind the
attribution of intelligence to machines would be basically the same when we are to
determine whether extraterrestrials or aliens are capable of understanding or intelligence,
for in both cases we just have to rely on behaviors. Referring to aliens (extraterrestrial
beings), they (1984, 55) write: “We would have no understanding of their civilization or
their physiology, and the sheer problem of communication with such entities would
limit sharply the accuracy of our assessment of their intelligence. Our basis for evaluating
their intelligence would rest solely on the outputs we received from them, and their
understanding of us would rest solely on our output to them.”
It shall be observed that the main point of Schank and Childers basically follows
the reasoning behind the Turing test (see Turing 1950). The Turing test uses a machine
simulation of human intelligent behavior, in the form of answering questions, to determine
whether machines can be said to be intelligent. This is precisely what we earlier identified
as the first level of simulation involved in the alleged computer simulation of human
thought. The goal of the machine is to mimic human intelligent behavior in ways such
that an interrogator would be unable to distinguish between a machine and a human
respondent. And in order to make the judgment of the interrogator objective, the features
of the machine and human respondents that are irrelevant for intelligence attribution—
such as physical features and sound of voice—the interrogator is separated from these
respondents by a wall. Here the interrogator’s only access to the respondents is through
textual communication via a teletype machine. The point of the test is consistency in the
attribution of intelligence: if the human respondent is considered intelligent through
his/her answers, the machine respondent whose answers cannot be distinguished from
those of the human respondent must therefore likewise be regarded as intelligent.
The Turing test, however, is only a test for machine intelligence; and not for the
computationality of human thought or of thinking in general (see Mabaquiao 2014). If a
machine passes the test, it is intelligent; but it does not mean that intelligence is
computational in nature. Take again the case of an alien, if the alien passes the Turing
test then it is intelligent. But should we suppose that the alien thinks like a computer?
We really do not know; all we know is that it manifests behaviors which when manifested
by humans are considered intelligent. The point of the test is simply to show that the
inner mechanism of a certain system is irrelevant to the ascription of intelligence to it.
Again, when Turing asked whether machines could be intelligent, he was not after
whether intelligence is computational or whether human intelligence is also governed
by some cognitive algorithm. He was only after whether given what machines can do,
we can legitimately say they are intelligent. And the test that he developed to determine

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
176 NAPOLEON M. MABAQUIAO, JR.

machine intelligence is in fact not exclusive to computing machines; for it can very
much apply to any kind of entity suspected of being intelligent such as aliens. Given
this, the fact that computing machines are governed by computer programs is irrelevant
to the ascription of intelligence to these machines. What is relevant is simply whether
they can perform the intelligent tasks. It is thus possible for a non-computational system
to exhibit intelligent behavior, pass the Turing test, and consequently be regarded as
intelligent. In other words, the machine needs only to simulate human intelligent behavior
to be considered intelligent, it does not need to duplicate human thought processes.
The computational simulation of human minds, on this level of simulation, does not
require that human minds be computational.

CONCLUSION

Two issues, one having to do with scope and the other with method, serve as
roadblocks to the project of computationalism. On the issue of scope, it is unclear how
much of the mind’s capacities computationalism intends to account for. The issue occurs
not just between computationalists and their critics but also among computationalists
themselves. There is a standing disagreement on whether computationalism should
account for the mind’s phenomenal and cognitive features or merely for its cognitive
features. As regards the mind’s cognitive features, there is a further disagreement on
whether it should account for the functionality of intelligence only or also of the
consciousness of intelligence. On the other hand, on the issue of method, it is not clear
how the duplication of minds can be arrived at through the routes of computer simulations
of brain processes and intelligent behavior. The computer simulation of brain processes
duplicates thought processes only if brain processes are taken to be constitutive of
thought processes. This, however, leads to the rejection of a necessary condition for
this duplication to occur: that thought processes be substrate neutral and thus be
multiply realizable. Taking the route of computationally simulating intelligent behavior
is likewise problematic, mainly for failing to sufficiently establish that minds are
computational. Given this, no duplication of minds can be sufficiently established.

REFERENCES
Armstrong, David M. 1991. The causal theory of mind. In The nature of mind. Edited by
David M. Rosenthal. Oxford: Oxford university press.
Block, Ned. 1991. Troubles with functionalism.” In The nature of mind, edited by David
Rosenthal, 211-229. Oxford: Oxford University Press.
Chalmers, David. 1993. A computational foundation for the study of cognition.
https://www.ida.liu.se/divisions/hcs/seminars/cogsciseminars/Papers/
Chalmers_Computa tional_foundations.pdf. [Journal of Cognitive Science 2012
(12): 323-357]. Accessed 9 June 2014.
________. 2010. The singularity: A philosophical analysis.” Journal of Consciousness
Studies (17): 7-65. http://consc.net/papers/singularity.pdf. Accessed 4 October 2012.

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 177

Churchland, Paul. 1995, The engine of reason, the seat of the soul. Cambridge: MIT
Press.
Churchland, Patricia. 1986, Neurophilosophy. Cambridge: MIT Press.
Crane, Tim. 1995. The mechanical mind: A philosophical introduction to minds,
machines and mental representation. London: Penguin Books Ltd.
Dennett, Daniel. 1989. The intentional stance. Massachusetts: The MIT Press
________. 1991. Three kinds of intentional psychology. In The nature of mind, edited
by David Rosenthal, 613-625. Oxford: Oxford University Press.
Eliasmith, Chris.,2013, How to Build a Brain, Oxford: Oxford: University Press.
Fodor, Jerry. 1979. The language of thought. Cambridge: Harvard university press.
________. 1991. Methodological solipsism considered as a research strategy in cognitive
psychology.” In The nature of mind, edited by David Rosenthal, 485-498. Oxford:
Oxford University Press.
_________. 2000. The mind doesn’t work that way: The scope and limits of
computational psychology. Massachusetts: The MIT Press.
Freidenberg, Jay & Silverman, Gordon. 2006. Cognitive science: An introduction to the
study of mind. California: Sage Publications, Inc.
Gardner, Howard. 1985. The mind’s new science: A history of the cognitive revolution.
U.S.A.: BasicBooks-A Division of HarperCollinsPublishers.
Harnish, Robert. 2002. Minds, brains, computers: A historical introduction to the
foundations of cognitive science. Oxford: Blackwell Publishers.
Jackson, Frank. 1991. What Mary didn’t know. In The nature of mind, edited by David
Rosenthal, 291-295. Oxford: Oxford University Press.
Mabaquiao, Napoleon. 2014. Turing and computationalism. Philosophia: An
International Journal of Philosophy 15 (1): 50-62.
_______. 2012. Mind, science and computation. Manila: De La Salle University
Publishing House and Vibal Foundation, Inc.
_______. 2011. Computer simulation of human thinking: An inquiry into its possibility
and implications.” Philosophia: An International Journal of Philosophy 40(1): 76-
87.
Milkowski, Marcin. 2013. Explaining the computational mind. Massachusetts: The
MIT Press.
_______. 2017. Objections to computationalism. A short survey. In Proceedings of the
39th Annual Meeting of the Cognitive Science Society. Computational Foundations
of Cognition (pp. 2723–2728). Presented at the 39th Annual Meeting of the Cognitive
Science Society, London: Cognitive Science Society. https://mindmodeling.org/
cogsci2017/papers/0515/index.html.
_______. 2018a. From computer metaphor to computer modeling: The evolution of
computationalism. Minds and Machines (28): 515-541.
_______. 2018b. The computational theory of mind. The Internet Enccyclopedia of
Philosophy, ISSN 2161-0002, https://www.iep.utm.edu/compmind/. Accessed 31
October 2018.
McDermott, Drew. 2007. Artificial intelligence and consciousness. In The Cambridge

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
178 NAPOLEON M. MABAQUIAO, JR.

handbook of consciousness, edited by Philip D. Zelano, Morris Moscovich, and

Evan Thompson, 117-150. Cambridge: Cambridge University Press.
McClelland, James and David Rumelhart. 1993. On learning the past tenses of English
verbs. In Readings in philosophy and cognitive science. Edited by Alvin Goldman.
Cambridge: The MIT press.
Minsky, Marvin. 1995. Logical versus analogical or symbolic versus connectionist or
neat versus scruffy. In Computation and intelligence: Collected readings. Edited
by George Luger. Cambridge: The MiT press.
Newell, Allen & Simon, Herbert. 1961. Computer simulation of human thinking. Science,
New Series, 134 (3495): 2011-2017.
________. 1976. Computer science as empirical inquiry: Symbols and search.
Communications of the ACM (Association for Computing Machinery) 19 (3): 113-
126.
________. 1995. GPS, a program that simulates human thought. In Computation and
intelligence: Collected readings, edited by G. Luger, 415-428. Cambridge: The MIT
Press.
Penrose, Roger. 1989. The Emperor’s new mind: Concerning computers, minds, and the
laws of physic. Oxford: Oxford University Press.
Peschl, Markus & Scheutz, Matthias. 2001. Some thoughts on computation and simulation
in cognitive science.” In Proceedings of the Sixth Congress of the Austrian
Philosophical Society. Online: https://hrilab.tufts.edu/publications/
scheutzpeschl00linz.pdf.Accessed February 14, 2017.
Putnam, Hilary. 1991. The nature of mental states. In The nature of mind. Edited by
David M. Rosenthal. Oxford: Oxford University Press.
Pylyshyn, Zenon. 1990. Computing in cognitive science. In Foundations of cognitive
science. Edited by Michael Posner. Cambridge: The MIT Press.
Rescoria, Michael. 2017. The computational theory of mind. The Stanford Encyclopedia
of Philosophy (Spring 2017 Edition), Edward N. Zalta (ed.), URL = <https://
plato.stanford.edu/archives/spr2017/entries/computational-mind/>. Accessed 4
March 2018.
Rumelhart, David. 1990. The architecture of mind: A connectionist approach. In
Foundations of cognitive science. Edited by Michael Posner. Cambridge: The MIT
Press
Sandberg, Anders. 2013. Feasibility of whole brain emulation. In Theory and philosophy
of artificial intelligence, edited by V. Müller. Berlin: Springer. http://
shanghailectures.org/sites/default/files/uploads/2013_Sandberg_Brain-
Simulation_34.pdf. Accessed 8 May 2014.
Sandberg, Anders & Bostrom, Nick. 2008. Whole brain emulation: A roadmap. Oxford
University: Technical Report, Future for Humanity Institute.http://www.fhi.ox.ac.uk/
Reports/2008-3.pdf. Accessed 8 October 2012.
Schank, Roger & Childers, Peter. 1984. The cognitive computer. Reading: Addision-
Wesley Publishing Company, Inc.
Searle, John. 1980. Minds, brains, and programs. Behavioral and Brain Sciences 3 (3):

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019
 TWO ROADBLOCKS OF COMPUTATIONALISM 179

417-457.http://www.cogsci.soton.ac.uk/bbs/Archive/bbs.searle2.html. Accessed 8
October 2011.
________ . 1990. Is the brain’s mind a computer program?” Scientific American, 26-31.
h t t p : / / w w w. c s . p r i n c e t o n . e d u / c o u r s e s / a r c h i v e / s p r 0 6 / c o s 11 6 /
Is_The_Brains_Mind_A_Computer_Program.pdf. Accessed 3 January 2011.
Schoemaker, Sydney. 1982. The inverted spectrum. The Journal of Philosophy 79 (7):
357-381.
Simon, Herbert. 1995. Machine as mind. In Computation and intelligence: Collected
readings, edited by George Luger, 675-692. Massachusetts: The MIT Press.
________. 1984. Why should machines learn. In Machine learning: An artificial
intelligence approach, edited by R. Michelski, J. Carbonell J, and T. Mitchell, 25-37.
Berlin: Springer.
Simon, Herbert & Kaplan, Craig. 1990. Foundations of cognitive science. In The
foundations of cognitive science, edited by Michael Posner, 1-47. Massachusetts:
The MIT Press.
Smolensky, Paul. 1993. On the proper treatment of connectionism. In Readings in
philosophy and cognitive science. Edited by Alvin M. Goldman. Cambridge: The
MIT press.
Sun, Ron. 2008. Introduction to computational cognitive modeling. In The Cambridge
handbook of computational psychology, edited by Ron Sun, 3-20. Cambridge:
Cambridge University Press, 2008.
Turing, Alan. 1950. Computing machinery and intelligence. Mind 59 (236): 433-460.
Wein, TU and Pona. 2015. Computational theory of consciousness. Seminar in artificial
intelligence. https://www.logic.at/lvas/SemAI/Consciousness.pdf. Accessed 10
October 2018.

Submitted: 25 August 2018; revised: 27 April 2019

Philosophia: International Journal of Philosophy ISSN 2244-1875

Vol. 20, No. 2, June 2019