Beyond Data Literacy:

DATA-POP ALLIANCE Reinventing Community

WHITE PAPER SERIES Engagement and Empowerment
in the Age of Data

October 2015

1
 Beyond Data Literacy:
DATA-POP ALLIANCE Reinventing Community
WHITE PAPER SERIES Engagement and
Empowerment in the Age
of Data
September 2015

Working Paper for Discussion

1
 Table of Contents
Foreword .................................................................................................................................... i
Glossary of key terms and concepts ......................................................................................... ii
Executive Summary ................................................................................................................. iv
Introduction ............................................................................................................................... 1
1 Genesis, contours and limits of ‘data literacy’................................................................... 3
1.1 Data literacy: an emerging concept of the ‘Data Revolution’ ................................................ 3
1.2 Data literacy as competencies of an extractive and transformative industry? .............. 4
1.3 Reconsidering ‘data literacy’ through the lens of history ...................................................... 5
2 Moving from ‘data literacy’ towards ‘literacy in the age of data’ ...................................... 6
2.1 Attempt at (re)defining ‘data literacy’ .......................................................................................... 6
2.2 Foundational pillars of ‘data literacy’ ........................................................................................... 9
2.3 Conceptualizing ‘data literacy’ as ‘literacy in the age of data’ ............................................ 14
3 Promoting data literacy for and via social inclusion ........................................................ 15
3.1 Making Big Data small(er) .............................................................................................................. 15
3.2 Understanding and designing for data literacy and inclusion using human- centered
approaches ...................................................................................................................................................... 18
4 Fostering social inclusion as data inclusion .................................................................... 19
4.1 Understanding and leveraging the power of words and language(s) .............................. 19
4.2 Politicizing the (Data) Revolution: towards data inclusion ................................................ 22
Concluding Remarks: The data revolution, data inclusion and data generations .................24
Appendices ............................................................................................................................... a
Appendix 1: “Data science without conscience….”............................................................................... a
Appendix 2: Claude Lévi-Strauss on writing and illiteracy programs in the original .............. c
Appendix 3: Literacy throughout history ............................................................................................... d
Appendix 4: The evolution of programming languages ...................................................................... f
Endnotes ................................................................................................................................... g

1
 Foreword
About this document
This document is part of Data-Pop Alliance’s White Papers Series developed in collaboration with
our partners. This White Paper was developed in collaboration with the Internews Center for
Innovation and Learning—who also provided funding—and researchers from the MIT Media Lab
Center for Civic Media, as well as Data-Pop Alliance.
Data-Pop Alliance is a coalition on Big Data and development jointly created by the Harvard
Humanitarian Initiative (HHI), the MIT Media Lab, and the Overseas Development Institute (ODI)
to promote a people-centred Big Data revolution.

About the co-authors

This paper was written by the following co-authors:
• Rahul Bhargava, Research Scientist, MIT Media Lab Center for Civic Media
• Erica Deahl, Innovation Specialist, 18F, General Services Administration; formerly graduate
student and research assistant, MIT Media Lab Center for Civic Media
• Emmanuel Letouzé (Lead and corresponding author), Director and co-Founder, Data-Pop
Alliance; Visiting Scholar, MIT Media Lab. [email protected]
• Amanda Noonan, Director of Research Design & Learning Internews Center for Innovation &
Learning
• David Sangokoya, Research Specialist, Data-Pop Alliance
• Natalie Shoup, Program Manager, Data-Pop Alliance

Acknowledgements
This paper benefited from guidance from Mark Frohardt (Internews) as well as comments from
William Hoffman (World Economic Forum), Alex ‘Sandy’ Pentland (MIT and Data-Pop Alliance),
and Alessia Lefébure (Columbia University).
Box 3 was written by Lauren Barrett, Communication Strategist, Data-Pop Alliance, who also
provided comments. Appendix 3 was developed by Gabriel Pestre, Research Scientist, Data-Pop
Alliance, and Carson Martinez, Research Intern, Data-Pop Alliance, who also edited the document.

Funding
Funding for this paper was provided by Internews Center for Innovation and Learning, whose
support is gratefully acknowledged, as well as the Rockefeller Foundation as part of their core
support to Data-Pop Alliance’s activities.

Disclaimer
The views presented in this paper are those of the authors and do not represent those of their
institutions.

Suggested Citation
“Beyond Data Literacy: Reinventing Community Engagement and Empowerment in the Age of
Data.” Data-Pop Alliance White Paper Series. Data-Pop Alliance (Harvard Humanitarian
Initiative, MIT Media Lab and Overseas Development Institute) and Internews. September 2015.

i
 Glossary of key terms and concepts

Algorithms: In mathematics and computer science, an algorithm is a series of predefined

instructions or rules – often written in a programming language intended for use by a computer –
designed to define how to sequentially solve a recurrent problem through calculations and data
processing. The use of algorithms for decision-making has grown in several sectors and services
such as policing and banking.
Big Data: The ecosystem created by the concomitant emergence of ‘the 3 Cs of Big Data’:
• Digital Crumbs—pieces of data passively emitted and/or collected by digital devices which
constitute very large data sets and streams and contain unique insights about their behaviors
and beliefs;
• Big Data Capacities—what has also been referred to as Big Data Analytics, that is the set of
tools and methods, hardware and software, know-how and skills, necessary to process and
analyse these new kinds of data—including visualization techniques, statistical machine-
learning and algorithms, etc.;
• Big Data Communities—which describe the various actors involved in the Big Data ecosystem,
from the generators of data to their analysts and end-users—i.e. potentially the whole
population.
Civic technology: A type of technology that enables citizen engagement or makes government
more accessible, effective, and efficient for the economic and social good of society. This specific
type of technology helps to connect people to resources, ideas, and other people needed to
improve their societies or communities.
Data: An object, variable, or piece of information that has the perceived capacity to be collected,
stored, and identifiable. It comes largely in two forms: structured and unstructured.
Structured data are essentially answers to questions asked by the collector of data, are
generally easy to organize and identify and have a strict hierarchy that is not easily
manipulated (i.e. responses to a survey organized in a table format and information about
people’s years of education and income in a chart).
Unstructured data are not readily amenable to automated analysis and often are used in ways
that differ from the intended purpose when collected (such as photos, videos, tweets), and
do not need to follow a hierarchical method of identification.
Data is also used as a policy concept and social phenomena (e.g. “data is changing the world”), or
as a shortcut for data ecosystems, Big Data, etc.
Data ecosystems: Complex adaptive systems that include data infrastructure, tools, media,
producers, consumers, curators, and sharers. They are complex organizations of dynamic social
relationships through which data/information moves and transforms in flows.
Data exhaust: Data that are passively emitted from cell phones, sensors, social media and other
platforms as digital translations of human actions and interactions.
Data inclusion: The universal ability of people to create, control, access and use data.
Data journalism: A new form of journalism stimulated by the open data movement, in which
stories are presented or supplemented through graphics or visualizations of analyzed datasets.
These static or interactive graphics include databases, maps, diagrams, grids, charts and many
other forms of illustrations that have transformed the look of mainstream news media.
Data literacy: The desire and ability to engage constructively in society through and with data.

ii
Data modeling: Using existing datasets to infer current conditions or predict future outcomes.
The process involves resolving complex relationships among datasets in order to understand what
data means and how the elements relate.
Data Revolution: A term that has become mainstream in the policy and development discourse
since the High-Level Panel of Eminent Persons on the Post-2015 Development Agenda called for a
“Data Revolution” to “strengthen data and statistics for accountability and decision-making purposes”. It refers to
the applications and implications of data as a social phenomenon. The term “Industrial Revolution of
Data” was coined by Computer Scientist Joseph Hellerstein in 2008.
Data science: A field of research and practice that focuses on solving real-world problems using
large amounts of data by combining skills from often distinct areas of expertise: math, computer
science (hacking and coding), statistics, social science, and even storytelling or art.
Digital divide: The differential access and ability to use information and communications
technologies between individuals, communities and countries — and the resulting socioeconomic
and political inequalities.
Literacy: As defined by UNESCO, "the ability to identify, understand, interpret, create, communicate and
compute, using printed and written materials associated with varying contexts. Literacy involves a continuum of
learning in enabling individuals to achieve their goals, to develop their knowledge and potential, and to participate fully
in their community and wider society."4
Literacy in the age of data: See Literacy in a post-2015 world.
Open data- Data that is easily accessible, machine-readable, accessible for free or at negligible cost,
and with minimal limitations on its use, transformation, and distribution
Popular data: The practice of engaging, empowering and participatory approaches to data-driven
presentation and decision-making (R. Bhargava).
Small data: Explicitly collected data – the data is collected in the open, with notice, and on purpose.
Small Data can be analyzed by interested laymen. Small Data doesn’t depend on technology-assisted
analysis, but can engage it as appropriate." (R. Bhargava).
(Statistical) Machine learning- A subset of data science, falling at the intersection of traditional
statistics and machine learning. Machine learning refers to the construction and study of computer
algorithms — step-by-step procedures used for calculations and classification — that can ‘learn’
when exposed to new data. This enables better predictions and decisions to be made based on what
was experienced in the past, as with filtering spam emails, for example. The addition of “statistical”
reflects the emphasis on statistical analysis and methodology, which is the main approach to modern
machine learning.

iii
 Executive Summary

The term ‘data literacy’ has gradually emerged as a mainstream term and potential buzzword of the
‘Data Revolution’ discussions, as experts, policymakers and advocates began considering what it
would take to enable citizens to make better use of the vast amount of data available to them.
Policymakers have advocated for more data science skills-training programs. Schools and non-profit
organizations (such as Code for America, Girls Who Code, School of Data, etc.) have emerged to
tackle the digital divide by providing coding programs and technical curricula for vulnerable
populations, specifically for women and minorities. An increasing number of data journalists are
using and writing about data. Open data and civic technology advocates have organized hackathons
for civic hackers to use technical skills and foster new conversations on data for social good.
Despite its growing popularity as a much-needed “bottom-up” solution, data literacy is ill-defined or
ambiguous at best. Are current conceptualizations of ‘data literacy’ adequate—or do they put too
much emphasis on technical requirements and fail to challenge deeper structural and more politically
controversial issues? What does it mean to be “data literate” in an age where data is everywhere—
and how does it differ from being literate? Why and how should it be promoted? How might the
promotion of ‘data literacy’ empower individuals and communities to keep governments
accountable, solve local problems, and navigate their own data ecosystems? In a world of ubiquitous
digital connectivity and rising inequity, should we in fact be concerned with and talking about data
inclusion instead?
We first discuss ‘data literacy’ as an emerging concept within a much longer historical narrative of
literacy promotion. History sheds light on how defining and promoting literacy—who was literate
and who was not—has been often entrenched with the constructs and perpetuation of power
structures within societies—at odds with the notion of literacy as a necessarily empowering and
enlightenment force. There is a risk that the same processes may play out in the age of data, at a
speed and scope commensurate with those of the spread of data as a social phenomenon.
We define data literacy as “the desire and ability to constructively engage in society through and about data.” Five
observations emerge from this definition:
1. “Desire and ability” highlights technology as a magnifier of human intent and capacity.
2. “Ability” underlines literacy as a continuum, moving away from the dichotomy of literate
and illiterate.
3. “Data” is understood broadly as “individual facts, statistics, or items of information.”
4. “Constructively engage in society” suggests an active purpose driving the desire and ability.
5. And “through or about data” offers the possibility for individuals to engage as data literate
individuals without being able to conduct advanced analytics.
This definition—as well as the nature of data itself—encompasses elements and principles from
each of these sub-kinds of literacy (such as media, statistical, scientific computational, information
and digital literacies), moving away from medium-centred definitions of literacy towards a more
encompassing one.
In utilizing a definition of data literacy that builds on the elements of current sub-categories of
literacy and expands beyond particular media—and their technocrats—we describe four key pillars
that form its foundation: data education, data visualizations, data modelling, and data participation.

iv
Our exploration of data literacy pushes us to further consider what it would mean to be “literate in
the age of data” and denote four core pillars in literacy promotion:
- Data literacy promotion must be agile and adaptive, focusing on helping to foster adaptive
capacities and resilience rather than teaching platforms and technical languages that are
bound to become outdated.
- Data literacy promotion must build on the key features and pillars from all core sub-
categories of literacy, viewing literacy as a continuum.
- Data literacy promotion must involve empowering people to navigate their current
ecosystems and societies in ways that are meaningful and effective for them.
- Data literacy promotion must involve providing multiple pathways for people with different
data literacy needs and capacities to interact within a complex system.
At the center of the rationale and attention around data literacy promotion should be the goal of
empowering citizens and communities as free agents. This can only be achieved by considering data
literacy as a significant means and metric for social inclusion—where data literacy as defined and
conceptualized above is promoted for and via greater social inclusion—or, more appropriately, data
inclusion.
Here we highlight the following three critical challenges in designing data literacy programs:
- Making Big Data smaller, on scale where most or many more people are willing and able to
engage than is the case today
- Understanding the importance of context and utilizing elements of human-centered design;
- Understanding and leveraging the power of words and language in communicating and
visualizing data
As we revisit the larger context of the Data Revolution in the last section and concluding remarks in
the light of data literacy and social inclusion, it becomes clear that if this Data Revolution is to bring
about positive change, it has to be an evolution towards social inclusion in the age of data – towards
data inclusion. If a ‘business-as-usual’ framing for the Data Revolution continues unabated, our
efforts toward greater data literacy may reinforce existing power dynamics that promote social
exclusion. This transitional period is the opportune time to create a path towards
empowerment. Data literacy focused on building data inclusion offers a doorway to understanding,
interpreting, and managing data-driven decisions and arguments for all people.
Supporting data literacy is not primarily about enabling individuals to master a particular skill or to
become proficient in a certain technology platform. Rather it is about equipping individuals to
understand the underlying principles and challenges of data. This understanding will in turn
empower people to comprehend, interpret, and use the data they encounter—and even to produce
and analyze their own data. This can only be achieved by considering data literacy as a means toward
a necessary reinvention of community engagement and empowerment—towards what we term data
inclusion.

v
 Introduction

There is no shortage of discussions and initiatives about the promise and perils of leveraging
data in various sizes and forms to meet the world’s challenges as part of the “Data
Revolution” called for by the United Nations and others.1 But how exactly is data expected
to change the world we live in? What is the ‘theory of change’? In February 2010, about a
century ago in data years, The Economist published a widely cited article titled “The Data
Deluge: Businesses, Governments and Society Are Only Starting to Tap Its Vast Potential”
(Figure 1). One of the first online comments read, “Here’s our 21st century jobs, America. Please
understand and educate the next
generation accordingly.”
Over the past couple of years, Figure 1: “The Data Deluge” as depicted in 2010
the concept of ‘data literacy’ has
emerged as a key priority.
Schools and nonprofit
organizations have developed
programs to teach children how
to code at an early age.
Advocates in the open data
movement have long argued for
expanding use of public data
beyond experts and trained
journalists. Millennial job
seekers are taking courses on
Coursera, edX, and other open
online courses to develop data
science skills and increase their
Source: http://www.economist.com/node/15579717
competitiveness in the data era.
The international development
and civic technology communities have also emphasized the need for data literacy as a
requirement of the data revolution. These organizations highlight both the potential
economic and social impact of data literacy in the physical world and, to a lesser extent, its
potential democratizing effect.
However, when it comes to the revolutionary potential of data—and the nature and features
of the ‘data revolution’—we often miss the big picture. For the most part, the ‘data
revolution’ discourse is based on the notion that what the world misses (and therefore needs
most) is more and better data, and more people who are able to collect, analyze, crunch data,
to make better decisions. This line of argumentation, on which most calls for enhancing
‘data literacy’ rests, is not entirely wrong, but it leaves out many complex and controversial
questions about why the world of 2015 is in such a bad shape.
As always, valuable lessons can be drawn from history. Claude Lévi-Strauss in his seminal
book Tristes tropiques, argued that writing and the early decades of literacy promotion served
the purposes of power elites. A recent partial evaluation of major data science programs by
Stanford researchers also points to major shortcomings in the training of future data
scientists.2

1
The rationale for promoting data literacy may seem straightforward. However, society as a
whole has little clarity about what data literacy is, much less what they should expect from it.
Vital questions require answers before we begin to promote data literacy as an answer to the
world’s pressing problems.
1. What is “data literacy”? What does it entail, and how is it distinguished from statistical
literacy, mathematical literacy, digital literacy, numeracy, and similar concepts?
2. Why does data literacy matter? What societal goals is data literacy expected to serve?
What is the theory of change that moves from improved data literacy to the achievement
of those goals?
3. How adequate are current conceptualizations of data literacy? Does the current emphasis
on technical requirements fail to address deeper structural issues? Are we moving toward
a dystopian future in which we have to rely on world-class data scientists to fix all our
problems for us? (Appendix 1)
4. How might we foster more inclusive approaches to data literacy? How can pervasive
data literacy be a force for social inclusion – for data inclusion?
This paper argues for an expansion of the concept of data literacy. We argue that data
literacy as a term is inadequate, reinforces existing inequities and should be replaced by the
larger concept of inclusion. Fulfilling that vision will be much more demanding and
disruptive than developing popular new software systems and delivering face-to-face
trainings and MOOCS on statistical packages. Rather, it will involve understanding and
defining data literacy in terms of how to effectively empower individuals to navigate their
own data/information ecosystems to produce, engage with, communicate and use data.
Additionally, as we promote data literacy, we must incorporate human-centered approaches
by design, understanding the dynamic and appropriate context involved in curating,
synthesizing and communicating data.
As we move forward with the Data Revolution, this is an opportune time to go beyond what
we’ve described as ‘data literacy’ today and reconsider literacy in the age of data. Further, we
must recognize data literacy as the means and metric towards a social inclusion revolution—
the deeper goals that make the Data Revolution truly “revolutionary”—towards what we
term data inclusion.
To make these points, this paper continues with a discussion of current mainstream
approaches to the concept of data literacy. Section 2 advocates for a broader definition of
data literacy, and proposes to conceptualize it as literacy in the age of data. Section 3 argues that
data literacy ultimately ought to be the means and metric of greater social inclusion and vice-
versa. Section 4 presents options and requirements to support this desirable evolution
towards greater social inclusion in the age of data that we term data inclusion. Finally, we
provide concluding thoughts on today’s data generation and its contribution to the data
revolution.

2
1 Genesis, contours and limits of ‘data literacy’
1.1 Data literacy: an emerging concept of the ‘Data Revolution’
In the new “Industrial Revolution of Data,”3 more and more actors have become interested in
tapping into data to solve complex problems. From open government data to sensor data to
data exhaust from social media, cell-phones and other digital devices, the vast amount of
data available should allow for policy-makers, experts, businesses and activists to ask more
informed questions and thereby develop more effective policies and programs.
The term ‘data literacy’ has gradually emerged as a mainstream term and potential buzzword
of the ‘data revolution’ discussions (Box 1), as experts, policymakers and advocates begin
considering what it would take to enable citizens to make better use of the vast amount of
data available to them. Arguments commonly put forth include the following:
1. ‘Data literacy’ increases the economic impact of Big, Small and Open Data. As
companies aim to capitalize on the potential business value generated from data,
employees with data science skills have become highly valuable in today’s economy.
Businesses have begun investing in skill-based trainings to help their analysts “conduct
data-driven experiments, to interpret data, and to create innovative data-based products
and services.” 4 For many managers and business owners, the more “data literate” their
workforce, the bigger their profit margins;
2. ‘Data literacy’ enables local populations to understand and solve local problems.
Development actors and community advocates push data literacy as an opportunity to
increase the efficiency and resilience of local actors and communities in solving local
problems. Data literate local actors would need to be able to “work…with very granular data,
or data limited in geographic scope, as opposed to statistics that are often aggregated to a higher level.” 5
More critically, data literacy would empower local actors with the ability to not only work
with existing data, but generate, own, use and monetize data;
3. ‘Data literacy’ empowers citizens to keep governments accountable and
transparent. Increased access to government data does not inherently create societal
impact. Rather, citizens must be able to interpret, understand and effectively use the data
in order to keep governments accountable and “spread the benefits of open government
to marginalized communities.”6 Data literacy can help civil society groups catalogue rights
violations, fuel data-driven journalism and spur citizen engagement in transparency and
anti-corruption efforts. Additionally, advocates voice that increasing ‘data literacy’ can help
bridge an ever-increasing digital divide.

Box 1: From the data revolution to data literacy in two UN-commissioned reports
In May 2013, as the UN began moving toward the post-2015 Sustainable Development Goals, a ‘High-Level
Panel of Eminent Persons on the Post-2015 Development Agenda’ appointed by the UN Secretary-General
published a report “call[ing] for data revolution for sustainable development…to improve the quality of statistics and
information available to people and governments.” 7 The report contained three mentions of “literacy”, and those
referred to “basic” literacy explicitly distinguished from “numeracy”.

By contrast, a 2014 report by the UN Secretary-General’s Independent Expert Advisory Group on a Data
Revolution for Sustainable Development (IEAG) used the term data literacy four times and treated it as one of
five pillars of its suggested action plan. The report called for an “education program aimed at improving people’s,
infomediaries’ and public servants’ capacity and data literacy to break down barriers between people and data.” 8

3
Despite this attention, descriptions of what exactly is meant by and expected from data
literacy have been absent or unclear. For example, although the concept featured
prominently in the IEAG report (Box 1), no definition was provided. In particular, it
remained ambiguous whether and how the report distinguished the roles of “capacity” and
“data literacy” in “break(ing) down barriers between people and data”—especially given that “capacity”
was absent in an otherwise similar statement. It was also not clear whether and how the
report distinguished data literacy and statistical literacy, and whether either separately or
combined these concepts could be assimilated to “numeracy”. Last, it was not clear whether
one or both should “ensur(e) that all people have capacity to input into and evaluate the quality of data
and use them for their own decisions, as well as to fully participate in initiatives to foster citizenship in the
information age”. Specifically, how much of the capacity “to input into and evaluate the quality of
data and use them for their own decisions” versus “to fully participate in initiatives to foster citizenship in
the information age” ought to be a result or expression of being data literate (versus statistically
literate or a combination of both).
The point is not to criticize a report that contained many points and proposals that are
currently shaping the global discussions about data, but to highlight how its ambiguities
underline the inherent complexity of the issue at hand, beyond and beneath its surface. To
date, the questions elaborated in the introduction do not have satisfactory answers.

1.2 Data literacy as competencies of an extractive and transformative industry?

As suggested above, the most commonly voiced argument in favor of ‘data literacy’ is that
people must know how to understand and use data for it to have an impact. As alluded to,
the Open Data community has long stressed that making government data ‘open’—i.e.
available on a website—is not enough; the ‘theory of change’ involves people having an
incentive and the ability to access the site and data, the capability to download it, and the
competencies and tools required to analyse it, etc. Similarly, when data in general and Big
Data in particular were compared to “the new oil of the digital economy”9 a few years ago, one
implicit but obvious step was the ‘refinement’ or
‘transformation’ stage. And indeed, Big Data has Figure 2: The Classic DIKW Pyramid
been defined as “a mindset (...) to turn mess into
meaning”10. The cover of The Economist’s Data
Deluge article conveys a similar argument
about transforming a raw material into a source
of growth.
These arguments reflect and fuel the main
stages of the widely-used vision of a process of
transformation from data to information,
information to knowledge—all the way up to
wisdom, summarized by the DIKW model
(Figure 2). 11 The choice of a pyramid and its
highly symbolic dimension may actually not be
entirely coincidental.
‘Data literacy’ would refer to the set of skills
and conditions for the first step up to occur. This approach and focus have been taken in
most popular press articles on the subject. For example, a 2014 Forbes article defined data
literacy as the “ability or ease” to “collect, analyse and visualize 18,446,744,073,709,600,000 data
points per person in less than one second.”12

4
Even when the bar is set lower than being able to process over 1018 data points per second,
current conceptualizations of data literacy revolve closely around some version of “the ability
to use and analyse data’”. As we shall see, this definition and its underlying assumptions and
expectations about data’s potential to bring about change and inherent obstacles to this
process is not flat-out wrong—but it focuses on skills required to perform tasks.
This conceptualization has implications that must be interrogated and challenged. For one, it
says nothing about the ultimate objectives of the transformative process at play—at the top
of the pyramid. It doesn’t question either the level of data collection—taking the availability
of data as a given—like oil sitting there to be extracted and processed. It leaves hardly any
room for ethical and political considerations. And yet the analogy with the ‘old oil’ should
serve as a warning: oil fuels economies and emergency vehicles as much as corruption, elite
capture, and global warming.
To hammer in the point, it is tempting to bring up the Godwin’s point 13 of Big Data:
Edward Snowden’s revelations about the nature and scope of the surveillance activities of
the US National Security Agency. If ‘data literacy’ were just the ability to turn data into
information, a society of junior NSA analysts (and their Amazon or Google counterparts)
would be a highly data literate society. With quite a few caveats discussed below, in an era
where concerns over data analytics-enabled government surveillance and corporate
manipulations (as in the cases of the Facebook social experimentation 14 or the recent
Volkswagen scandal15) are rising, one can feel intuitively that such a society may not be the
most progressive and inclusive. This, a minima, suggests that current conceptualizations of
what data literacy is, means and entails, are not fully adequate.

1.3 Reconsidering ‘data literacy’ through the lens of history

One way to probe the concept of ‘data literacy’ is to reflect on its metaphorical roots—i.e.
‘standard’ or ‘traditional’ or ‘basic’ literacy—to understand how both literacy and efforts to
promote it have been defined and conceptualized. History sheds light on how defining and
promoting literacy has been often entrenched with the constructs and perpetuation of power
structures within societies—at odds with the notion of literacy as a necessarily empowering
and enlightening force.
Box 2: Claude Lévi-Strauss on the function of writing and literacy programs in history
“Writing is a strange thing. It would seem as if its appearance could not have failed to wreak profound changes in the living conditions of
our race, and that these transformations must have been above all intellectual in character. (…) Yet nothing of what we know of writing,
or of its role in evolution, can be said to justify this conception. If my hypothesis is correct, the primary function of writing, as a means of
communication, is to facilitate the enslavement of other human beings. …The use of writing for disinterested ends, and with a view to
satisfactions of the mind in the fields either of science or the arts, is a secondary result of its invention and may even be no more than a
way of reinforcing, justifying, or dissimulating its primary function. [T]he European-wide movement towards compulsory education in the
nineteenth century went hand in hand with the extension of military service and the systematization of the proletariat. The struggle
against illiteracy is indistinguishable from the increased control exerted over the individual citizen by the holders of power.”
Claude Lévi-Strauss, Tristes Tropiques, 1955

For example, Claude Lévi-Strauss, in his famous book Tristes tropiques, studied the historical
role of writing and the rationale for literacy programs during the Industrial Revolution in
Europe. He reckoned that the notion that writing “could not have failed to wreak profound changes
in the living conditions of our race” was a misconception. Rather, he argued, writing— this “strange
thing”—was, for centuries, a means by which elites perpetuated and strengthened their
control of the masses. 16 Further, Lévi-Strauss described literacy campaigns as means of
making people able to serve the interests of the elites in power (Box 2 and Appendix 2).

5
Further, the literature on the effects of (and need for) literacy during the Industrial
Revolution is rather ambiguous. By the mid-nineteenth century, the majority of European
workers did not need to be literate, then measured by the ability “to sign one’s name,” but there
was a point below which the process of industrialization could not have happened, as “it was
useful to have a wide pool from which those who did need literacy—merchants, clerks, surveyors and
engineers, for instance (end of the quote is missing?).”17”As Lévi-Strauss points out too, this period
corresponded with the heyday of European Nation-State building 18 —which in parts of
Europe also implied the systematic and brutal cracking down on regional languages to
impose that of the central authority.19
Spurred by international organizations such as UNESCO as well as governments and civil
society organizations, efforts to promote universal mass literacy began in the 1950s. As
literacy became part of the agenda for an emerging international community post-World War
II, campaigns to eradicate “illiteracy” focused on promoting reading and writing as a basic
set of skills for autonomy in and across countries. Yet, definitions of literacy differed across
states and regions, and global campaigns against illiteracy became fragmented during the
Cold War. Since then, the development of new technologies and globalization introduced
new literacies, prompting literacy advocates to constantly reconsider the definition of the
lowest bar for basic literacy.
The world has changed significantly since Lévi-Strauss wrote these lines. Various forms of
literacy, including some related to the use of data, have undeniably made fundamental
contributions to people’s enlightenment and empowerment—from the civil rights
movement to fights for gender equality and environmental protection. But whereas it is not
completely clear whether these effects were ‘secondary’ or explicitly intended, it is evident
that literacy programs have always been embedded in local ontologies.
Taking an objective look at the state of affairs today suggests that advanced data analytics
techniques, despite their potential to spur human progress, have so far worked especially
well for governments and corporations. It is unclear whether and how promoting ‘data
literacy’ the way it is currently conceptualized—by providing skills without much in the way
of questioning their ends and means—may reverse or repeat the history of literacy
promotion. This invites us to reconsider current approaches to data literacy that are based
on overly mechanistic views of the world and its problems.

2 Moving from ‘data literacy’ towards ‘literacy in the age of

data’
2.1 Attempt at (re)defining ‘data literacy’
Two co-authors of this paper have previously20 proposed to define data literacy as the ability
to read, work with, analyse and argue with data:
• Reading data involves understanding what data is, and what aspects of the world it
represents;
• Working with data involves creating, acquiring, cleaning, and managing it;
• Analysing data involves filtering, sorting, aggregating, comparing, and performing other
such analytic operations on it;
• Arguing with data involves using data to support a larger narrative intended to
communicate some message to a particular audience.

6
In this paper, we put forward a new definition of data literacy that goes one step further. We
define data literacy as the “the desire and ability to constructively engage in society
through or about data.”
At least five observations can be made about this definition.
1. Desire and ability echoes Kentaro Toyama’s conceptualization of technology as a magnifier
of human intent and capacity. 21 Awareness and opportunity to engage are front and
center;
2. Ability allows for varying levels of data literacy, away from dichotomy between data
literate and data illiterate individuals. Obviously different positions and different goals
require different levels of data literacy. Certain basic thresholds might be established to
define minimal data literacy, and these could change over time;
3. Data is understood in its broader sense; data has been defined as “individual facts, statistics,
or items of information, or “a body of facts”, and in that sense a news article whether printed or
online, a tweet, an Instagram photo, a video – all of these are data. In the realm of data
analytics, the distinction overlaps in great part although not fully with the distinction
between unstructured data (such as files) and structured data (typically databases) (Box 3).
Though this notion may indeed seem very broad, as it suggests that potentially everything,
from music22 to a chair’s molecular structure and thus aspect, are or could be data, a
feature of the world’s future may very well be ubiquitous data-ification;
4. Constructively engage in society suggests an active sense of purpose—it suggests that literacy
must be sought, deployed and measured in relation to specific goals that are deemed
‘constructive’; of course these will be highly dependent on context but these rule out, for
instance, any goal that infringes on Human Rights;23
5. Through or about offers the possibility for individuals to engage in society through and/or
about data—i.e. one can be data literate without being able to conduct advanced analysis.
This definition also encompasses existing medium-based literacies. Evolutions in definitions
of literacy have been on par with the emergence of ‘sub-kinds’ of literacies with their own
specific definitions and requirements—statistical literacy, scientific literacy, media literacy,
digital literacy and more. Breaking down ‘literacy’ into its constitutive pieces has practical
value, but shaping various forms of literacy around emerging mediums increases the ‘silo-
ification’ and technocracy around these mediums.
The attitudes and skills implied by our definition of data literacy can be pulled from the
following sub-kinds of literacy:24
1. Information literacy is a pre-Internet era concept that emphasizes the importance
of being able to locate and determine the credibility of information.
2. Scientific literacy focuses on the application of scientific concepts and
experimentation methods needed for personal decision-making and civic
participation.25
3. Media literacy in contrast deemphasizes the acquisition of technical skills and
focuses instead on supporting media production and developing a critical
understanding of issues such as modes of representation, language, production, and
audience.26
4. Statistical literacy is about enabling individuals to critically assess and use statistics
within their everyday lives.
5. Computational literacy encourages individuals to seek algorithmic approaches to
problems, move between different levels of abstraction, and use modelling as a way
to identify relationships.27,28

7
 6. Digital literacy involves “the ability to find, evaluate, utilize, share, and create
content using information technologies and the Internet.”29
Data literacy interacts with and builds on all six of these approaches and requires a
combination of the technical, critical, quantitative and conceptual skills on which they are
based (Figure 3). This definition—as well as the nature of data itself—encompasses elements
and principles from each of these sub-kinds of literacy, moving away from medium-centred
definitions of literacy towards a more encompassing one.
Figure 3: How different modern types of literacies interact

Source: authors’ elaboration based on secondary sources

Since data are used differently in various domains, researchers have proposed multiple,
possible definitions of the competencies required to be data literate. These definitions differ
in terms of the skills they emphasize, the level of technical proficiency they call for, and the
methods and technologies they specify. Data literacy demands pedagogical approaches that
are customized to the context of the learners, and every program will need to tailor its
approach to focus on the competencies appropriate for its particular mission and audience.
One facet of this definition of data literacy that is valuable in a pedagogical and research
context is that it emphasizes the importance of interdisciplinary thinking as a core
component of data literacy. Discipline-specific approaches to data literacy focus on either
quantitative or qualitative investigation, which can bias the resulting interpretations.
Quantitative analysis makes it possible to uncover hidden patterns and gain insight into
complex datasets, while qualitative analysis makes it possible to surface individual stories
within those aggregations. Increasingly, institutions have recognized that these methods can
be complementary; by learning both approaches, individuals can explore an issue from
multiple perspectives and reach more balanced and comprehensive conclusions.
Similar to the history of other literacy efforts, data literacy will not be a quick fix, but a rather
slow exercise in behaviour change. Spurring engagement and enhancing the universal
perceived value of data literacy will require marketing the skills as essential to everyday

8
functioning and long-term advancement and presenting data as accessible and applicable. It
will be an evolution in intellectual dynamics.
As alluded to, the ethical and political implications of this new data age, such as human
rights abuses, lie in our conceptualization of data literacy. To effectively engage participants in
data ecosystems there will be a need to understand, design and communicate approaches
that foster contextually relevant, human-centred, culturally resonant and effective
engagement and use. The primary protection against encroachment of rights lies in data
literate citizens with the desire and ability to comprehend and control the use of their data.
If we conceptualize—and indeed confine, as we will discuss below—data literacy within
these parameters, what does it look like, entail, and require? How can it be further unpacked?

2.2 Foundational pillars of ‘data literacy’

In employing a definition of data literacy that builds on the elements of current sub-
categories of literacy and expands beyond particular media—and their technocrats—we
describe four key pillars that form its foundation: data education, data visualizations, data
modelling, and data participation.
The “data” in data literacy: data education
First, data literacy involves understanding what data is, or are, and does, or do—that is, basic
data education. It entails, at minima, being able to define data; contrast and connect data
with statistics 30 ; distinguish structured and unstructured data, qualitative and quantitative
data, etc. (Box 3). It also implies critically assessing what they contain, convey, represent,
even at a fairly abstract level—pieces of information that result from the translation, the
coding, of some human experience into a language; numbers, words, pixels—with full or
limited awareness of the object (be it an individual, a community, a country) characterized in
the data.
It follows that central to basic data education is acknowledging the imperfect and biased
nature of data. The simplistic assumptions that data, especially when they get bigger, are a
neutral and unbiased representation of reality has been criticized by Crawford and others
who have brought attention to the ethical dilemmas inherent in these data practices. Datasets
are still “objects of human design” and therefore vulnerable to error and biases. 31 This can
happen at various stages of the process of data collection, analysis, and representation,
unintentionally or through deliberate manipulation.
Specifically, structured data are created intentionally to answer a particular question, and
their creators often bias the way the data are created to get the answers they want. This limits
the usefulness of the data to their viewpoint. (In other words, structured data is answers, not
raw data). Unstructured data on the other hand are primarily created with intentions other
than what they are being used for (photos are created for fun and memory, tweets for
connecting to friends, etc.). The interpretation for social purposes comes after the data
creation, and it is then that biases—notably political—can creep in. Unstructured data,
however, make it possible to revisit the data with different perspectives and purposes: they
thus support debate and a broad range of knowledge creation.32
These are only a handful of the features of data that a data literate individual—someone
willing and able to constructively engage in society through and about data—should grasp. These
features, for the most part, are post-collection and pre-analysis. We can then examine three
additional building blocks of data literacy: data visualization, data modelling, and data
participation.

9
Box 3: What is Data?
There is no agreed-upon definition of data. In general, data is an object, variable, or information that has
the perceived capacity to be collected, stored, and identified. According to Oxford Dictionaries, data is
“facts and statistics collected together for reference or analysis.”33
There are two main types of data: structured and unstructured. The former are created intentionally to answer
a particular question; as a result they are easy to search for, organize, and identify and have a strict
hierarchy. The hierarchy for a
person’s favorite food might be:
Structured Unstructured
food, fruit, apple, red delicious.
Each variable is clearly defined • Hierarchal structure • No set internal structure
and labeled in a way that fits the
• Least flexible • Most flexible
structure’s taxonomy. Relational
databases, popularized by IBM in • ~10% of data and decreasing • ~90% of data and increasing
the 1970s and 1980s, offered a • Each unit corresponds with • Each unit may have its own
significant improvement in the use a specific row and column, identifiable set of
of structured data in comparison i.e. hierarchy. Follows ACID information and does not
to earlier hierarchical models. model: Atomicity, correspond to a particular
Consistency, Isolation, hierarchy, such as film clips,
Unstructured data are everything Durability pictures, and text documents
else. It can be photos, word
documents, and other variables that do not need to follow a hierarchical method of identification. For
example, someone can input data, such as an ‘apple’, without having to sequence it under the category of
‘fruits’ or know that there is a subcategory of ‘red delicious.’
Is unstructured data completely disorganized then? No. Metadata can be used to describe unstructured
data. This can be .jpeg for example if it is used to describe a picture of an apple.
Over 90% of data is unstructured
data, and it is growing Qualitative Quantitative
exponentially in comparison to
structured data because of the Information about
Responses to a survey about
rapid creation of digital data, such people’s age (in years),

Structured
people’s activities during the
as videos and tweets. As a World years of education,
weekend organized in a table
Bank report notes, “a 10-minute income, and amount
format with columns and
video of cats uploaded on spent in a table format
rows
YouTube may be quite heavy in with columns and rows
terms of bytes but arguably contain
less value than say Walt Whitman’s
Unstructured

Photos from weekend Field notes about

Leaves of Grass.” 34
activities, which can be people’s income, age or
The Big Data revolution is a result organized or unorganized by other quantitative data;
of this rapidly growing size, type of photo (i.e. .jpeg), or scans of the table
unstructured data. Much of this and photo descriptions, etc. described directly above
unstructured data is qualitative,
however, the large majority of
tools used to derive insights from data are quantitative in nature, such as statistics.35 As greater techniques
and tools are needed to analyze and make use of this data – i.e. actionable insights – there is a greater need
to create quantitative, structured metadata surrounding these unstructured big data sets in order to employ
these analytical tools.

10
Data visualizations
Data visualizations, the typical vehicle through which data are conveyed to the public, are
not necessarily accurate, accessible or appropriate within their contexts. A data visualization
calls attention to a specific pattern or story within a dataset, illustrating one of many possible
interpretations; a visualization cannot communicate the full complexity of a dataset. This
raises the issue of the questions we ask of data – questions that inevitably will involve a
degree of bias at least to some extent.
Creating and understanding data visualizations requires graphicacy. Graphicacy is“the ability to
understand and present information in the form of sketches, photographs, diagrams, maps, plans, charts,
graphs and other non-textual, two-dimensional formats”. 36 It is a complementary skill that is
necessary for the effective communication of data-derived information. Beyond graphicacy,
it is also necessary to understand how different ways of communication may or may not be
cultural appropriate in terms of symbols, visuals and media. At present our understanding of
such effective and culturally appropriate communication is sparse. The challenge and
opportunity will be to work with communities and individuals to surface their contextual
understanding of data and the ways to understand, find, capture, use and communicate
these, as illustrated in Box 4.
Furthermore, the bias in visualizations is often deliberate.37,38 A highly data literate person or
public will understand not only how to interpret data visualizations, but also how to assess
the reliability and objectivity of the sources.
Box 3: Understanding data; case study in graphicacy
To test the accessibility and utility of data visualizations created for the Kenyan media, Internews led
a pilot investigation. In Kenya and elsewhere, the publication of data visualizations in print news
media is a recent phenomenon, linked to the popularization of data journalism.
The study tested four different textual and graphical representations of data. There was no significant
difference in participants’ comprehension of text-based analysis versus graphic. Among the graphic
representations, simple bar charts were most easily interpreted.
Pinker’s theory of graph comprehension contends that viewers must be able to recognize specific
types of graphs to be able to translate their visual information into quantitative information. “If this
type is unknown to the viewer, s/he will almost always struggle with interpretation at first glance.
According to Pinker, there are three routes to comprehension: “being told,” induction, and
deduction.
Improvement in the ability to read graphs may be best enhanced by explicit instruction. Furthermore,
aesthetic preference and even the ability to read a graph may be culturally determined. Picture stories
with which the audience can identify appear to facilitate recall. This finding aligns with the “active
audience” theory in media reception studies.

Data modelling
Data modelling—using existing datasets to infer current conditions or predict future
outcomes—has become a prominent practice among corporations and municipalities
because it has proven to be so profitable. Overreliance on data modelling often fails to fully
account for human error, oversimplifies complex factors, makes it difficult to verify the
quality of the original data, and points toward solutions that overlook human needs.
A well-known example of data modelling’s potential for failure due to human bias and
flawed methods is the series of devastating fires in the Bronx in the 1970s that resulted from

11
the RAND Corporation’s recommendation to close numerous fire stations in one of New
York City’s poorest neighbourhoods.39 Another common issue that is as old as statistical
analysis is spurious correlations and confusing correlation with causation; one new challenge
is the fact that, with more data, spurious correlations and meaningless patterns are easier to
find—this has been referred to as “apophenia40”—which some policymakers, salespersons and
various advocates have been known to use and abuse to advance their own agendas or
embellish their accomplishments.41
These examples illustrate the perils of an overreliance on data and data analytics when data
modelling is used without taking into consideration existing local knowledge and the agility
of human behaviours. Further, using abstruse methods of data analysis that seem
authoritative makes policies harder for opponents to verify and critique. These issues all have
a profound impact on individuals, most of whom do not know what predictive data
modelling is, let alone have the knowledge to evaluate and point out its shortcomings.
The public needs to be more data literate to interrogate and potentially challenge these very
decisions and processes. This highlights the critical need for usable tools and trusted
intermediaries that are able to open the ‘black boxes’ and unpack these processes and expose
their potential biases in comprehensible and engaging ways.
Participation
Most people are excluded from engaging with data for a host of technological, technical,
cognitive and practical barriers. As a result, they are unable to influence the types of
applications that are built, and to direct those efforts toward outcomes that may benefit their
communities. Even applications that are intended to engage diverse communities in
contributing data run the risk of overlooking underserved communities who may lack access
to the technology necessary to participate.42,43
The open data movement attempts to address this issue by making data free and readily
available, thereby increasing the transparency of public institutions and encouraging public
participation. As it stands, most of the individuals taking advantage of open data resources
are civic technologists with existing expertise who come with their own biases and
perspectives. Most people are still excluded from engaging with data since they require
access to education, infrastructure, and technology.
Data literacy as a concept involves the interaction of multiple ecosystems containing both
literate and illiterate actors. The ability to use data and to create actionable knowledge
requires understanding of local information ecosystems: how data is transformed into
information, then knowledge, as it flows through different points and channels in a dynamic,
non-linear, networked system. The function or role of any node or point within an
information ecosystem changes depending on the context. A farmer can be a consumer of
information received through a mobile phone alert, a producer of information as they
transcribe the information on a bag of rice, and a mover and influencer of information as
they share it with the rest of their community and at the market.
An information ecosystem is not a static entity; it is by nature constantly evolving and
changing. Nor is it a discrete form; it can be defined at many levels, from global to national
to community to interest-based groupings within communities. It is a complex, adaptive
system that includes information infrastructure, tools, media, producers, consumers,
curators, and sharers. Data and data-derived information and communications are
increasingly critical elements of information ecosystems. Research by the Internews Center
for Innovation & Learning has described eight critical dimensions common to any
data/information ecosystem (Figure 4).

12
 Figure 4: Eight Critical Dimensions of Information Ecosystems

Key features here not only include logistical aspects such as demand, structures, applications
and flow of information but also place a heavy focus on context; ease of accessing, finding,
using, sharing, and exchanging different types of information; barriers to interaction and
participation; and relevance of information. Extremely significant to these ecosystems as
well, and a resulting feature of their complexity, is social trust – the influence of trust
networks on the flow and use of information – which involves the data itself, the consumer
and the influencers of the system.
Understanding how data/information ecosystems function and evolve is critical to fostering
and expanding data literacy and therefore data engagement. This approach does not simply
empower voices “from the ground up” – it accounts for needs, challenges, and opportunities
for all nodes within a system to be appreciated and valued, be they governments, community
leaders, telecoms, epidemiologists, technologists, patients, farmers, or others. Trust,
transparency and better control of the flow of data and information are all supported as
feedback loops that continually feed the flow of data, information and impact. For the ever-
expanding communities of “Data Revolutionaries” there is an urgent and magnificent
opportunity to ensure the innovative uses of data are in the service of supporting more
inclusive, appropriate and transparent solutions. We need to ensure the inclusion and
empowerment of all members of the data ecosystem—producers, consumers, movers and
users—to expand everyone’s opportunities to make meaningful and relevant decisions.

13
2.3 Conceptualizing ‘data literacy’ as ‘literacy in the age of data’
Despite our attempt to clarify and broaden the definition of data literacy, it may not stand the
test of time. Artificial intelligence, virtual reality, and other new technologies threaten to
completely disrupt our current conceptualizations of data and how to use it. Data literacy,
defined and conceptualized as “the desire and ability to engage constructively in society through or about
data,” may not be enough to empower global citizens to use various kinds data to improve
their lives and strengthen their communities.
As new discoveries and media change, data literacy must be able to adapt as well, focusing
on fostering adaptive capacities and resilience rather than teaching platforms and technical
languages that are bound to become out-dated. An even broader concept is needed to ensure
citizens may identify, navigate and participate in the rapidly changing data ecosystem.
Promoting data literacy needs to move beyond the constraints of a sub-type of literacy and
expand to promoting literacy in the age of data.
Promoting literacy in the age of data must be adaptive. Despite our increasing capacities to
collect and capture data, we are still navigating the possibilities of data. Discovering the
impact of a current dataset could take years: the emergence of new technologies and datasets
may challenge our acceptance of current datasets and increase the risks involved in using
them.
Promoting literacy in the age of data should not solely be based in new technologies or
mediums, but involve empowering people to navigate their current ecosystems and societies
in ways that are meaningful and effective for them. In the age of data, new data and
technologies will continue to challenge and shape our individual and collective capacities to
learn, communicate and make decisions. Data literacy promotion must move beyond solely
focusing on platform-based skill development (e.g. writing, coding, etc.).
As it turns out, this is exactly in line with the evolution in the thinking about ‘standard’
literacy. In setting its goal for universal literacy under the motto of “Literacy as Freedom” in
the mid-2000s—before the emergence of data and Big Data as core policy concepts—
UNESCO noted:
“At first glance, ‘literacy’ would seem to be a term that everyone understands. But at the same time,
literacy as a concept has proved to be both complex and dynamic, continuing to be interpreted and
defined in a multiplicity of ways.”
It then proposed an expansive definition of literacy:
“[T]he conception of literacy has moved beyond its simple notion as the set of technical skills of
reading, writing and calculating—the so-called ‘three Rs’—to a plural notion encompassing the
manifold meanings and dimensions of these undeniably vital competencies. Such a view, attending
recent economic, political and social transformations, including globalization, and the advancement of
information and communication technologies (ICTs), recognizes that there are many practices of
literacy embedded in different cultural processes, personal circumstances and collective structures”
(UNESCO 2004, 6).44
Today, UNESCO defines literacy in a broad encompassing definition as follows:
Literacy refers to the "ability to identify, understand, interpret, create, communicate and
compute, using printed and written materials associated with varying contexts. Literacy involves a
continuum of learning in enabling individuals to achieve their goals, to develop their knowledge and
potential, and to participate fully in their community and wider society".4

14
 Although this definition did not make any reference to data, it is consistent with and includes
core aspects of data literacy—although less emphasis is placed on ‘desire’.
Promoting literacy in the age of data should build on the key features and pillars from all
core sub-categories of literacy – literacy as a continuum. Historically, a main feature of
literacy has been the evolutionary nature and instrumental dimension of its definition and
measurement: we noted earlier how it was once defined and measured by the ability to sign
one’s name as opposed to tracing a cross. Over time, the standards by which literacy has
been assessed have risen alongside literacy rates; at all times, literacy has been fundamentally
redefined according to its purpose. The definition, promotion and evaluation of literacy have
been and remain context and purpose-specific—not a-contextual, abstract and absolute.
Even more so, literacy is only relevant within shared ontologies. This makes literacy both an
instrument of power, and the condition for challenging it.
Further on this point, promoting literacy in the age of data must go beyond the binary
conceptualization of being literate or illiterate. There are perilous dangers in thinking that a
given individual or group are data literate, and thereby assuming the completeness of their
potential ability to engage with and use data. The subtleties and grades of literacy are
numerous and continue to evolve; as this evolution unfolds, so too must the fidelities of its
systems, tools and supports.
Promoting literacy in the age of data must involve providing multiple pathways for people
with different data literacy needs and capacities to interact within a complex system. In
understanding literacy as a “continuum of learning,” efforts to promote literacy in the age of
data must provide multiple entry-points for people to understand and consider data literacy
in conjunction with their own goals for knowledge development and participation in their
community and societies. In this sense, there are many levels of literacy as a way for people
with different capacities and needs to interact in the complex ecosystems that exist.

3 Promoting data literacy for and via social inclusion

At the center of the rationale for data literacy promotion must sit the goal of empowering
citizens and communities as free agents. This can only be achieved by considering data
literacy as a significant means and metric for social inclusion—where data literacy as defined
and conceptualized above is promoted for and via greater social inclusion; which we term data
inclusion.

3.1 Making Big Data small(er)

Data may reinforce existing power structures and processes. This risk is most evident in the
case of Big Data; the sources and features of Big Data’s potentially ‘disempowering’ effect
have by now been well identified, notably by Boyd and Crawford who as early as 2010 noted
the existence of “significant questions of truth, control and power in Big Data studies: researchers have the
tools and the access, while social media users as a whole do not.”45In a recent paper, Bhargava and
D’Ignazio summarized that “Big Data has an empowerment problem”46, adding “one might argue that
having Big Data be in service of the subjects needs is sufficient to argue it is beneficial, but empowerment is
not handed from those in power to those without, it bubbles from the bottom up.”
Some will argue that enhancing ‘(Big) Data literacy’ would precisely help veer the cursor
towards greater empowerment, depending on what is involved in and expected from data
literacy. Mastering Python is neither a sufficient nor a necessary condition to pass the data
literacy test if, for instance, the data used has been collected without the subjects’ informed
consent (Box 5), extracted without a solid understanding of the risks and implications

15
involved, or used to discriminate against people.
This argument implies that the conceptualization and promotion of data literacy should not
be disconnected from ethical considerations. Indeed it can be argued that data literacy is
essentially an ethical imperative.
Box 5: Do you know where your data are?
In 2014, the Federal Trade Commission’s report on data brokers—companies that collect and sell
personal consumer information—revealed that most consumers are unaware of how their personal
data is being collected and used.47Although the work of data brokers can benefit consumers and the
economy by, for example, enabling fraud protection, most consumers simply do not know how
extensively their personally identifiable information (PII), property information, and social media use
are being sold for profit.
For example, CoreLogic’s database includes over 147 million records containing property-specific
data for over 99% of all U.S. residential properties. RapLeaf’s data aggregator contains at least one
data point, including PII, for over 80% of all U.S. consumer email addresses.ii
As the FTC report and many well-publicized data leaks brought attention to the risks to consumer
privacy, marketers and data brokers responded to pressure to make terms-of-service agreements
more accessible and detailed. However, many Americans still do not know how much of their data is
being shared nor the extent to which they can opt out of these practices. A study by the Annenberg
School for Communication at the University of Pennsylvania concluded that marketers often frame
data sharing as a tradeoff for the delivery of services or discounts. However, most consumers do not
have the data literacy to make informed decisions to give up their data.48
Experts have suggested remedies based on the E.U.’s data protection directive and similar initiatives.
For example, Alex Pentland of the MIT Media Lab and Data-Pop Alliance calls for a “new deal on
data” to give users ownership of their data and control over its use.49

A dichotomy of Big Data versus Small Data is often delineated especially in relation to
empowerment. While both concepts are significant to discussions of empowerment and
engagement, the demarcation is not without limitations. As a particular type of data, ‘Big
Data’ is actually a bit of a misnomer since the data in question are in fact many little data
points related to people’s behaviors and beliefs to make up very large data streams and sets.
As a field of research and practice, Big Data typically refers to the algorithmic analysis of
large, passively collected, sets and streams to discover patterns and relationships, often not
obvious at the onset of the analysis. The results provide insights into systems that otherwise
wouldn't have been revealed but for the massive collection and automated, computer-
enabled, analysis of data. Small Data, in contrast, centers on active data collection by
engaged, willful participants, with analysis using manual or computer-assisted techniques.
Both practices may use qualitative and quantitative datasets and both can entail structured or
unstructured data.
However, a distinctive characteristic of Big Data is that it always involves structured
quantitative data at some point in the analytics process. For example, in applying machine-
learning, no matter the character or configuration of the source data, it is necessary to
quantify the data in order to perform the necessary operations to give a result.

16
In sum, typical Small and Big Data approaches currently differ along four major aspects:

Big Data Versus Small Data

Aspect Big Data Small Data
Level of awareness by Null or limited Aware and engaged
the subjects of the data
Nature of data used Quantitative and qualitative Qualitative and
quantitative
Technique used for Algorithmic, computer-enabled with room for Manual and computer-
analysis human inputs (for classification purposes in assisted
supervised-machine learning)
‘Traditional’ uses and • Developing and using algorithmic analysis • Collecting data to
requirements techniques; answer questions;
• Understanding what data people generate in • Finding stories in data
their daily lives; that exists;
• Making decisions about the ethical use of • Picking effective
data; techniques for telling
• Creating detailed visualizations to understand data-driven stories.
what the data may represent or indicate.

This dichotomy between Big Data and Small Data should not persist indefinitely. Small Data
will increasingly use and rely on ‘Big Data’ techniques and tools as they become more widely
available, easy to use and eventually adopted (e.g. Google Maps). Big Data, on the other
hand, needs to learn from ‘Small Data’ when it comes to enhancing people’s awareness and
engagement. An additional requirement of all this Big Data, Open Data and data of any kind
is simply the raising of awareness. For users of both approaches, it is important to stress
here the importance of understanding and incorporating context (Box 6).

Box 6: The importance of context in data literacy

Data, design and implementation cycle has yet to begin to fully embrace the benefits of
understanding and incorporating context. This is a vital part of data literacy that is too often limited
or even entirely absent in the way data scientists approach data. To make data truly relevant,
actionable and therefore impactful, analysis should take into account the context and the direct
experience and insights of those about whom or for whom data-driven solutions are provided.
To achieve this potential, the data need to be analyzed with contextual knowledge, the ground truth,
as part of the data scientists’ mindset as they consider what the data may be revealing in terms of
experiences and potential solutions. Context in data literacy should encompass input from civil
society, from industry, from individuals, from local and national governments—from all of the
stakeholders that comprise the data/ info ecosystem.

In the future, as almost every aspect of human life will potentially be subject to data-ification
and data literacy increases (expanding to literacy in the age of data), the boundaries between
Small and Big Data should blur, and the result will become ‘All Data’. ‘All Data’ would then
refer to all the applications and implications of data for societies, and data literacy will be the
means and measure of people’s desire and ability to actively craft that future.

17
Designers of data literacy initiatives face a challenging path ahead, fraught with concerns
over appropriateness of activities and ever-changing technology. The Big Data – Small Data
divide is not an easy gap to negotiate. Some will be inclined to focus on Big Data from an
activist/awareness raising point of view, while others will focus on Small Data because it is
more tractable and accessible.
Today, a major imperative and challenge is to make Big Data ‘smaller’, on a scale where most
or many more people are willing and able to be engaged than is the case today. A taxonomy
of potential functions of Big Data has already been put forth that both stresses the need for
and provides an entry point for making Big Data smaller (?).50 These four functions of Big
Data are:
1. Descriptive; i.e. to use of data to produce maps, visualizations, etc.;
2. Predictive; i.e. to make inferences about current conditions and forecasts about future
events;
3. Prescriptive—also referred to as diagnostic—to draw causal inferences with Big Data;
and lastly and critically,
4. Discursive—also referred to as engagement—which “concerns spurring and shaping dialogue
within and between communities and with key stakeholders,” recognizing that “the longer-term
potential of Big Data lies in its capacity to raise citizens’ awareness and empower them to take
action.”51
In other words, Big Data—and, by extension, all data and data approaches—can and must
be leveraged to empower citizens. This requires increasing their levels of data literacy
understood as their desire and ability to argue through and about data.
Consequently, one of the requirements and features of a more data literate society is a
society where citizens demand to have a voice in how and by whom data is used, what it is
used for, and use data to fulfill their goals in an ethical and equitable manner. And so, a data
literate society is a more inclusive society.

3.2 Understanding and designing for data literacy and inclusion using human-
centered approaches
As previously stated, for the much-touted promise of inclusion and empowerment through
data to be realized, individuals and groups in all parts of the data ecosystem have to be data
literate; and this fostering of data literacy is highly contextual (see Box 5).
At the heart of such an approach is the need to empower all members of a data ecosystem –
producers, consumers, movers and users – to expand all participants’ opportunities to make
meaningful and relevant data-informed decisions and actions. There is no ‘one size fits all’
approach. All the participants have different attributes, different needs and challenges, and
these may be distinct depending on the roles of the participants at a given time. While some
of the challenges are clearly evident, many are far more opaque.

Elements of human-centered design

At the root of understanding data literacy and designing for inclusion is an urgent need to
rethink approaches for the design, creation and support of data driven systems, that are
more human-centered and based on inclusion, empathy and responsiveness. Contextual,
human-centered approaches are arguably a critical and currently too often absent element in
the design and development of data-related activities.

18
These methods can identify the complex and nuanced needs, challenges and aspirations of
individuals and groups within a data ecosystem. Central to human-centered approaches are
discovery and learning related to experience. Empathy is a truly powerful and necessary tool
to understand the experiences of others. With mindful attention to the explicit, implicit and
indeed unconscious needs of different individuals and groups, appropriate activities, tools,
supports and communications for data and data-informed actions can be designed and
supported.
A human-centered approach to data literacy would foster:
• Greater inclusiveness: Human-centered design serves to surface complex and nuanced
needs, challenges and aspirations of all individuals and their communities in relation
to understanding, creating, using and communicating data.
• Enhanced community participation: Understanding how to convey data-derived insights
using appropriate, accessible and trusted language, visuals and media etc. will enable
audiences to actively participate in the data ecosystem.
• Prioritization of critical needs: By embracing local context with empathy and
mindfulness, the most pertinent questions to ask of data emerge. Explicit, implicit
and previously unknown benefits and harms can be identified. This serves to
strengthen networks of trust, manage risks, enable effective policies and more fully
value the uniqueness of all individuals.
• Increased resilience: Human-centered methods help all stakeholders listen, learn and
adapt to change and uncertainty. Fluid, open and agnostic, such approaches provide
the means to continually learn and revisit core assumptions that can cloud judgment,
increase risks and drive poor utility and impact.

4 Fostering social inclusion as data inclusion

4.1 Understanding and leveraging the power of words and language(s)
The words we use to describe an object, concept, idea or field are critical to their
understanding. In data literacy programs, this can take many routes. At the most basic level,
using "information" instead of "data" can be significant. The processes employed similarly
need more familiar names. Instead of data analysis, a more enticing term might be story-finding.
Instead of visualization, a more appealing term might be storytelling. Both of these terms
employ the frame of "story" to make the critical connection between data and action.
Story-finding is a scaffolding for data analysis. People can be intimidated by the idea of
‘analyzing data,’ which is not the sole but certainly a key component of being data literate.
The story-finding framing can assuage that intimidation. Story-finding also connects data
analysis with the critical question of "why?" Asking "why" creates a natural link between the
exercise of analyzing data and the changes one hopes to bring about, creating a natural
connection with the ethical considerations discussed above by placing the ends above the
means.
Storytelling allows for more creativity in the approach to data presentation. The term ‘data
visualization’ brings about imagery of fancy graphics or analytics that may create unnecessary
semantic and cognitive barriers. The far more approachable concept of ‘storytelling’
acknowledges there are many ways to present data, not just in complex visualizations that
may appeal to audiences who have capacity in different domains—such as performance,
painting, photography, or drawing. Art has historically been one of the best tools for
engagement and can be a key point of entry to engage people with data.

19
This story-finding and storytelling framing also stresses a fundamental point that echoes the
aforementioned need to recognize the non-neutral nature of data. Data-driven arguments are
very often used to support opinions—i.e. statements that are, in Bachelard’s terms, “inherently
wrong.”52 Being aware of the propensity of data-driven arguments to pass for objective facts
and seeking to critically engage with and interrogate their validity is a major feature and
benefit of data literacy.
Further adding to this non-neutrality, another obvious and yet largely ignored obstacle to
broad data literacy is the fact that the majority of online content is in English.53 In today’s
and probably even more so tomorrow’s world, not being able to read or understand English
may be an impediment to data literacy. To give a French-speaking person an important
message, a good starting point would probably be to use French. This may seem obvious;
however, where communicating important data is concerned there is little research into what
might be the optimal ‘languages’ for a given message and a given community.
Beyond the boundaries presented by the language that data and information are
communicated in, barriers to entry for data literacy persist, stemming from the various and
rapidly evolving languages that data are captured, manipulated, and analyzed in (Appendix
4). Once inside this community of programmers and data scientists, tools and knowledge are
often easily accessible and highly participatory—with the free and open-source programming
language R and its community being a good case in point. 54 However, currently this
community is isolated from and often esoteric to outsiders. There is a need and potential to
produce accessible, usable tools to enable users of data to verify the information that is
produced from the process of data aggregation and analysis. Research is needed to develop
such tools that reverse-engineer the data path and present this information clearly and
intelligibly to users. And perhaps more importantly, there is a need to train trusted ‘data
translators and connectors’—once called infomediaries — to connect this community to the
rest of the world.
In attempting to teach data literacy, educators are faced with a formidable challenge: the vast
gap between the smaller, more orderly datasets and problems that learners typically work on
in the classroom, and the large, unstructured problems that individuals face in the real world.
Some organizations are attempting to bridge this gap through progressive data education
that transitions from entry-level, pre-defined problems to more complex and uncertain ones,
although these interventions are rare (Box 7).
Journalists and other communicators have access to an ever-burgeoning range of tools and
techniques to manipulate and present data. However, there seems to be relatively little
understanding of which of these approaches may be the most appropriate for a given
audience or type of information.
Internews, Kenya’s data journalism training experience, took place in the context of a virtual
absence of data literacy skills among trainees (Box 8). “Some journalists were hardly numerate; they
didn’t know how to express simple ratios and had a phobia of Excel”, says Dorothy Otieno, Internews
in Kenya’s lead data journalism trainer.55 Otieno describes how in 2011, when Internews
conducted its first data journalism training, journalists had no notion that they could demand
data from policy makers or researchers. The launch of the Kenyan Open Data Initiative
(KODI) in 2012 was a digital leapfrog into data accessibility in a country where journalists
hardly dared ask for data – either because access would be denied or the data would simply
be cumbersome to access. Two years later, Internews has gained valuable insights about the
steps involved in teaching data literacy to content creators, which in turn translates to a more
data literate media audience, able to engage with data and empowered through use of data.

20
Box 7: Progressive Data Education Initiatives
Working with elementary schools, the Oceans of Data Institute, a science education research group
within the Education Development Center (EDC), has developed a model to define the cognitive
stages that learners need to progress through in order to move from pre-defined problems to ill-
defined ones, and from discrete cases to abstract patterns. 56 The ODI’s four stages of learning
progression towards “data scientist” are:
1. Unstructured observation through human senses;
2. Student-collected small data sets;
3. Professionally collected large datasets and well-structured problems;
4. Professionally collected large datasets and ill-structured problems.
Learners develop increasing proficiency as they progress through these stages and must make
significant leaps in learning to transition from one stage to the next. The ODI has also developed a
series of curricula that aim to support students in making these progressions. The EDC Earth
Science curriculum, for example, asks students to compare temperature and precipitation data from
the NOAA’s National Climate Data Center to similar data from their local area, helping them to
transition from analyzing small local data to larger professionally-collected datasets. Another
curriculum, Ocean Tracks, 57 deepens students’ understanding of professionally collected data by
enabling them to explore the migration patterns of large marine species and analyze the relationship
between migration patterns and factors of the ocean environment.
City Digits, an interactive mapping platform and series of high school math curricula developed by
researchers and designers from the MIT Civic Data Design Lab, Brooklyn College, and the Center
for Urban Pedagogy, helps students to bridge the gap between ODI’s stages two and four.58 Students
use data to analyze a local social justice issue from two perspectives: first they collect their own
datasets by conducting interviews in their neighborhood, and then they explore citywide datasets that
illuminate larger-scale patterns. These activities help students compare the small-scale, highly
personal and large-scale, statistical implications of an issue, and to understand the relationship of
individual data points to the larger system of which they are part. The first iteration of the curriculum
focuses on the issue of state lotteries and their impact on low-income communities, and the second
applies the methodology to the topic of pawnshops and “fringe banking.”

The Internews experience suggests some simple tips for addressing data literacy and
empowering data narrative creators and audiences:
• Allow time for discovery and recognize that this is a new field for many;
• Harness the distinct skills of data researchers, coders, developers, designers and
journalists and team these together in collaborative projects;
• Acknowledge that data-derived journalism is time consuming, but that the effort
pays off in the form of unique insights and rewarding opportunities for audience
engagement and crowd projects;
• Apply rigor and discipline, critical thinking and alertness to unreliable data.
By applying these principles, data journalism teams in Kenya have produced stories with
impact, which have transformed the look of mainstream media in Kenya. It is now typical
for data-derived feature stories or investigations to claim double spread space in the
newspaper and for television features with data visualizations to be broadcast in prime time.

21
Box 8: Case Study - Training Data Journalists in Kenya
In early 2014, five Kenyan media professionals, including two print journalists, a TV journalist, a
developer and a graphic designer, graduated as Internews data journalism fellows. They had
completed a 16-week data journalism training and production that raised awareness about the
misspending, corruption and inequality that plague Kenya’s public healthcare system. Fellows learned
how to access, scrape, analyze and visualize data using digital tools. They also gained an appreciation
of interconnections in data – with the ultimate aim of unearthing stories buried in data through
investigative journalism.59

Such examples point to the power of data-driven investigations to foster a culture of

accountability. Greater investment in these activities is needed to nurture data translators,
able to harness rich data sets in order to reach conclusions that matter to citizens and are
communicated in an understandable manner, in order to spur further audience engagement
with the data.

4.2 Politicizing the (Data) Revolution: towards data inclusion

The 2014 IEAG report emphasizes that “revolutions begin with people, not with reports, and the data
revolution is no different.”60 As we revisit the larger context of the Data Revolution in the light
of data literacy and social inclusion, it becomes clear that if this Data Revolution is to bring
about positive change, it has to be an evolution towards greater social inclusion that goes
beyond the current discourse.
A key problem in the current discourse is the disconnect between the goals of the revolution
(specifically in social inclusion and empowerment) and the ramifications of its current
framing. The current ‘data revolution’ narrative suggests that most of the ills of the world are
due to a lack of data and data skills in the hands of ‘decision-makers,’ often consciously or
unconsciously equated with ‘policy-makers.’ It is indeed common to read and hear that what
‘we’ need is “better data, more timely data, more accurate data, more disaggregated data.” As
it is framed, this “better data” will provide greater information on the desks and briefing
notes of policymakers and their advisers, and, secondarily, better skills in the hands and
brains of average citizens.
This framing focuses attention on a subset of relatively uncontroversial issues, while shying
away from addressing the more complex and controversial barriers. These lie in politics and
power. Since words—that are data, projecting meaning and assumptions—matter, we should
use them with care. What is at play or should be at stake is not a data revolution, a
‘revolution of the data,” but a revolution via data, a social revolution, or a social evolution
led by empowered citizens with, via, and in the age of data. This is data inclusion.
If a ‘business-as-usual’ framing for the Data Revolution continues unabated, the future data-
driven society will fail to realize the aims of the Data Revolution and will reinforce existing
power dynamics that promote social exclusion. This transitional period is the opportune
time to create a path towards empowerment. Data literacy focused on building social
inclusion offers a doorway to understanding, interpreting, and managing data-driven
decisions and arguments for all people. The alternative future we must strive for is one
where people are incentivized and empowered to control their own data and its use. This is
data inclusion.

22
What does politics look like in an age of data inclusion?
Recent political campaigns in the US have been heralded as data-driven successes—using
insights from algorithmic mining to target messages that appeal to particular potential
donors or voters. These methodologies are similar to those of the advertising industry,
casting citizens as consumers who are opting to purchase a particular candidate. While some
might argue this is an apt metaphor, it suggests just one definition of citizenship. The
classical notion of the informed citizen involves choice in governmental representation
based on information they receive from various sources. There are, of course, alternative
definitions of citizenship that provide more opportunities to engage than simply donating to
a campaign or voting for a candidate.
A better alternative for politics in the age of data is the idea of citizens as monitors of
government policy and activity. This presents a future politics centered on the idea of
accountability through data empowerment. Citizens could monitor and collect data about
governmental roles, responsibilities, and services. These crowd-sourced data could be used
to advocate for changes, expose corruption and more. Nascent versions of these types of
tools are springing up across the globe, but only scratch the surface of what is possible when
combined with affordable sensors, mobile phones, and strong community partnerships.
What does education look like in an age of data inclusion?
Current data literacy programs in formal education settings are few and far between. Schools
tend to play catch up with grand societal-level changes, and the data revolution is no
exception to this rule. Most existing programs and curricula focus on numeracy and more
math-related concepts (to be in line with local or federal curriculum guidelines). In fact most
data literacy work in formal schooling is targeted at teachers, helping them understand and
use data about their students performance. These foci ignore the strong potentials to use
data literacy activities to connect schooling to community, action, and citizenship.
A better alternative for education in the age of data is data literacy programs in formal
education that focus on empowering students to collect, work with, analyze, and use data to
create change in their communities. These programs should focus on existing problems in
communities, empower students to collect and analyze data about the problems, and then to
try and effect change.
What does law look like in an age of data inclusion?
In the areas of law and law enforcement, the business-as-usual framing could play out in
terrifying ways over the next decade. Surveillance cameras already cover huge areas of our
main cities. In the United States, police cars are passively collecting license plates as they
drive around without strong rules of data retention and access control. These programs are
more recently beginning to focus on threat modeling and predictive analytics. Certainly there
is a place for data analysis in law and policing, but data divorced from context and ethics
very quickly dissolves into a morass of poor short-term decision making. Forays into
predictive analytics, when combined with insufficient legal frameworks, will not play out well
in the real world.
A better alternative for law in the age of data is strong privacy protection to act as the
anchor of this alternative future. A shift in attitude must be made towards respecting data
ownership and removing passive detection as the norm. This will certainly require major
legislative changes to accomplish, but it is not out of reach.
These visions of data-empowered futures for education, politics, and law are just pieces of
the larger puzzle we must put together – a puzzle with data literacy at its heart. Technical

23
and social infrastructure must be built to support these changes. Remembering that most
data is simple information about our interactions in the world, we are forced to recognize
that more data will be created each minute. This rate is increasing as more and more of daily
interactions become governed by digital technologies, which lend themselves to easy data
gathering. Most technologies currently being developed lend themselves to these types of
large-scale data gathering exercises, but to return to an earlier theme, we need to ensure that
the small-data efforts are similarly supported.
Our data-empowered future is creative, not consumptive. People will create the datasets they
need to solve problems they are concerned about. People will create powerful stories that
pull the data together in relevant ways. People will create effective presentations of those
stories to bring about change.

Concluding Remarks: The data revolution, data inclusion and

data generations
Data literacy is a “strange thing” that should not be promoted without specifying what is
meant and expected from it. We put forth two main counter proposals with strong historical
and political undertones. One is to talk about—or at least think in terms of— ‘literacy in the
age of data’ as a much more useful concept, simply defined as “the desire and ability to
constructively engage in society through and about data.” The second is to promote it via and for
social inclusion.
To date, the appeal and success of ‘data literacy’ in the public discourse and psyche reflect
and fuel a relatively narrow conceptualization of the ‘Data Revolution,’ itself rooted in a
simplistic diagnostic of the world’s problems and what data can do about them. For too
many champions of data literacy, the main solutions focus on technical capacity gaps that
need to be filled and fixed, so that more people can become better at analyzing data. As we
noted, this is of course, partly accurate: to varying degrees, the vast majority of the world’s
population—including those crafting and implementing public policies or other public
service functions—are ill-equipped to deal with the ‘data deluge’ that is only in its early years.
There are massive and pressing needs to strengthen technical capacities for the positive
transformative power of data to be unleashed in sectors and regions where lack of relevant
and timely information has been a real impediment to social progress. Priority constituencies
will include national statistical officers, elected officials, journalists, down to communities
and individuals. Building those key skills requires significant investment over many years if
they are to remain relevant in the age of data.
However, the current data literacy narrative overlooks many more complex and controversial
questions. History has repeatedly shown how technology could entrench rather than
challenge power structures that perpetuate detrimental outcomes—for instance inequity,
poverty, corruption, and environmental degradation. This is obviously because technology is
often invented and used first and primarily, when not exclusively, by those in power. The
promotion and diffusion of technology to the masses is not necessarily at odds with this
model, as Lévi-Strauss argued about literacy promotion. This is old news, but history has a
tendency to repeat itself as its lessons are forgotten. There is a real risk that the same
processes may play out in the age of data, at a speed and scope commensurable with those of
the spread of data as a social phenomenon.

24
Settling for a medium-based, technical conceptualization of ‘data literacy’ may realize rather
than mitigate this risk—a world where the latest data advances ‘work’ first and foremost to
serve surveillance and commercial ends, with ‘data literacy’ serving the function of a nice
sugarcoat. In a world where fewer than half of governments represented at the United
Nations—all of which supported the SDGs—are not near being democratic and rule over
half of the global population, educating and creating a data literate global citizenry would
mean putting a lot of politicians and members of supporting elites out of power.
Conceptualizing and promoting ‘data literacy’ as ‘literacy in the age of data’ is consistent with
the expansion and deepening of the concept of literacy over time to continuously and
increasingly consider its requirements and metrics in light of its intended purposes—agency,
empowerment, enlightenment, inclusion. Today and tomorrow, being literate ought to be
defined and measured by how individuals are "enabled to achieve their goals, to develop their
knowledge and potential, and to participate fully in their community and wider society.”
We go one step further. We argue that this requires putting social inclusion front and center
of policy and community discussions and initiatives. A data literate society—a literate society
in the age of data—is a more inclusive society. Data as a concept and object is a powerful
means to affect social inclusion positively or negatively; and reciprocally, the future of data
as a concept and object will be determined in great part by how inclusive versus exclusionary
or fragmented our societies are. Spurring literacy in the age of data must advance inclusive
economic and social impact; and vice versa. We call the end of this process data inclusion.
Should kids learn how to code in school? Of course. And outside of school? They will. Any
parent and anyone who interacts with a young child realize from their own recent experience
how quickly and fundamentally technology and the world are changing in tandem, and what
that may mean for their future. A term like the “Snapchat generation” reflects how new
technologies and the increasing volumes of data associated with new devices now define the
experiences, interactions and education and the future of children and teenagers. This
generation is actually the first of the many ‘data generations’ to come.
By the time the children of this ‘data generation’ turn 15, by 2030, a lot of them may be able
to write sophisticated code in Python and R to run analysis on various kinds of data sets and
streams—including some or many about them, that they may collect and use themselves. The
quantified-self movement of today will probably grow in size and significance, and
tomorrow quantified communities will emerge. People may have gained full or partial access
to the rights to data about them; data may be born with a built-in finite life expectancy; legal
and technological systems radically changing informed consent will be in place. The very
definition of individual and group privacy will continue to be challenged and adapted. These
processes will not be solely relevant and confined to the micro-worlds of the US East and
West Coasts, highly developed pockets of Europe, Asia, Oceania, and a few other cities of
the Global South.
It is impossible to predict which power systems and structures will then govern societies in
which their own children will live—except that they will probably look very different from
what past and current generations have known. Will representative governments still be the
norm? Maybe not. One thing is sure—data will be pervasive and infuse almost all aspects of
human life, from the societal to the individual levels. The ethical and political responsibility
of those in positions of power today is to empower people to shape this future themselves.

25
Appendices

Appendix 1: “Data science without conscience….”

Emmanuel Letouzé, 2014

a
b
 Appendix 2: Claude Lévi-Strauss on writing and illiteracy programs in the
original
“C'est une étrange chose que l'écriture. Il semblerait que son apparition n'eût pu manquer de déterminer des
changements profonds dans les conditions d'existence de l'humanité; et que ces transformations dussent être surtout de
nature intellectuelle. La possession de l'écriture multiplie prodigieusement l'aptitude des hommes à préserver les
connaissances. On la concevrait volontiers comme une mémoire artificielle, dont le développement devrait s'accompagner
d'une meilleure conscience du passé, donc d'une plus grande capacité à organiser le présent et l'avenir. Après avoir
éliminé tous les critères proposés pour distinguer la barbarie de la civilisation, on aimerait au moins retenir celui-là :
peuples avec ou sans écriture, les uns capables de cumuler les acquisitions anciennes et progressant de plus en plus vite
vers le but qu'ils se sont assigné, tandis que les autres, impuissants à retenir le passé au delà de cette frange que la
mémoire individuelle suffit à fixer, resteraient prisonniers d'une histoire fluctuante à laquelle manqueraient toujours une
origine et la conscience durable du projet.
Pourtant, rien de ce que nous savons de l'écriture et de son rôle dans l'évolution ne justifie une telle conception. Une des
phases les plus créatrices de l'histoire de l'humanité se place pendant l'avènement du néolithique, responsable de
l'agriculture, de la domestication des animaux et d'autres arts. Pour y parvenir, il a fallu que, pendant des millénaires,
de petites collectivités humaines observent, expérimentent et transmettent le fruit de leurs réflexions. Cette immense
entreprise s'est déroulée avec une rigueur et une continuité attestées par le succès, alors que l'écriture était encore
inconnue. Si celle-ci est apparue entre le 4e et le 3e millénaire avant notre ère, on doit voir en elle un résultat déjà
lointain (et sans doute indirect) de la révolution néolithique, mais nullement sa condition. À quelle grande innovation
est-elle liée ? Sur le plan de la technique, on ne peut guère citer que l'architecture. Mais celle des Égyptiens ou des
Sumériens n'était pas supérieure aux ouvrages de certains Américains qui ignoraient l'écriture au moment de la
découverte. Inversement, depuis l'invention de l'écriture jusqu'à la naissance de la science moderne, le monde occidental a
vécu quelque cinq mille années pendant lesquelles ses connaissances ont fluctué plus qu'elles ne se sont accrues. On a
souvent remarqué qu'entre le genre de vie d'un citoyen grec ou romain et celui d'un bourgeois européen du XVIIIe siècle
il n'y avait pas grande différence. Au néolithique, l'humanité a accompli des pas de géant sans le secours de l'écriture ;
avec elle, les civilisations historiques de l'Occident ont longtemps stagné. Sans doute concevrait-on mal l'épanouissement
scientifique du XIXe et du XXe siècle sans écriture. Mais cette condition nécessaire n'est certainement pas suffisante
pour l'expliquer.
Si l'on veut mettre en corrélation l'apparition de l'écriture avec certains traits caractéristiques de la civilisation, il faut
chercher dans une autre direction. Le seul phénomène qui l'ait fidèlement accompagnée est la formation des cités et des
empires, c'est-à-dire l'intégration dans un système politique d'un nombre considérable d'individus et leur hiérarchisation
en castes et en classes. Telle est, en tout cas, l'évolution typique à laquelle on assiste, depuis l'Égypte jusqu'à la Chine,
au moment où l'écriture fait son début : elle paraît favoriser l'exploitation des hommes avant leur illumination. Cette
exploitation, qui permettait de rassembler des milliers de travailleurs pour les astreindre à des tâches exténuantes, rend
mieux compte de la naissance de l'architecture que la relation directe envisagée tout à l'heure. Si mon hypothèse est
exacte, il faut admettre que la fonction primaire de la communication écrite est de faciliter l'asservissement. L'emploi de
l'écriture à des fins désintéressées, en vue de tirer des satisfactions intellectuelles et esthétiques, est un résultat secondaire,
si même il ne se réduit pas le plus souvent à un moyen pour renforcer, justifier ou dissimuler l'autre. […]
Si l'écriture n'a pas suffi à consolider les connaissances, elle était peut-être indispensable pour affermir les dominations.
Regardons plus près de nous : l'action systématique des États européens en faveur de l'instruction obligatoire, qui se
développe au cours du XIXe siècle, va de pair avec l'extension du service militaire et la prolétarisation. La lutte contre
l'analphabétisme se confond ainsi avec le renforcement du contrôle des citoyens par le Pouvoir. Car il faut que tous
sachent lire pour que ce dernier puisse dire : nul n'est censé ignorer la loi.
Du plan national, l'entreprise est passée sur le plan international, grâce à cette complicité qui s'est nouée, entre de
jeunes États - confrontés à des problèmes qui furent les nôtres il y a un ou deux siècles - et une société internationale de
nantis, inquiète de la menace que représentent pour sa stabilité les réactions de peuples mal entraînés par la parole écrite
à penser en formules modifiables à volonté, et à donner prise aux efforts d'édification. En accédant au savoir entassé
dans les bibliothèques, ces peuples se rendent vulnérables aux mensonges que les documents imprimés propagent en
proportion encore plus grande.”
Claude Lévi-Strauss,Tristes tropiques, 1955.

c
Appendix 3: Literacy throughout history

d
Sources: the National Center for Education Statistics (2008), Netherlands Organization for Scientific Research (2014), OECD (2006),
UNESCO (2008)

e
Appendix 4: The evolution of programming languages

f
Endnotes

1 http://www.post2015hlp.org/the-report/
2 One Size Does Not Fit All: The Shortcomings of the Mainstream Data Scientist Working for Social

Good. A. Albright and Levine, 2015, forthcoming.

3 Hellerstein, Joseph. 2015. 'The Commoditization Of Massive Data Analysis - O'reilly Radar'.

Radar.Oreilly.Com. Accessed September 3 2015. http://radar.oreilly.com/2008/11/the-

commoditization-of-massive.html .
4 https://hbr.org/2012/09/data-is-useless-without-the-skills
5http://data.worldbank.org/sites/default/files/issue-paper-financing-the-data-revolution-data-

literacy-and-use.pdf
6 http://idatosabiertos.org/wp-content/uploads/2014/04/OD4DProposaversionfinal11.pdf
7 “A New Global Partnership: Eradicate Poverty and Transform Economies Through Sustainable

Development”, May 2013 http://www.un.org/sg/management/pdf/HLP_P2015_Report.pdf

8 “A World that Counts” Nov 2014, http://www.undatarevolution.org/report/
9 http://www.wired.com/insights/2014/07/data-new-oil-digital-economy/
10 Technology Services Industry Association, 2015. 'The New Data Refineries: Transforming Big

Data Into Decisions'. Blog.Tsia.Com. Accessed September 3 2015. http://blog.tsia.com/blog/the-

new-data-refineries-transforming-big-data-into-decisions.
11 Rowley, J. 2007. 'The Wisdom Hierarchy: Representations Of The DIKW Hierarchy'. Journal Of

Information Science 33 (2): 163-180. SAGE Publications. doi:10.1177/0165551506070706.

12 Maycotte, H.O. “Data Literacy – What It Is and Why None of Us Have It”, Oct 2014, Forbes,

http://www.forbes.com/sites/homaycotte/2014/10/28/data-literacy-what-it-is-and-why-none-of-
us-have-it/
13 https://en.wikipedia.org/wiki/Godwin%27s_law
14 Kramer, Adam, Guillory and Hancock, “Experimental evidence of massive-scale emotional

contagion through social networks” http://www.pnas.org/content/111/24/8788.full.pdf

15 “ Volkswagen boss quits over diesel emissions scandal”, Reuters, Sept 2015,
http://www.reuters.com/article/2015/09/24/us-usa-volkswagen-idUSKCN0RL0II20150924
16 Leví Strauss, Claude, Triste tropiques, 1955
17 More, Charles. Understanding the industrial revolution. Routledge, 2002.
18 http://www.nationalismproject.org/what/hobsbawm.htm
19 In some primary schools in Brittany at the end of the 19th century, it was forbidden to “spit on the

ground and speak Breton”. (« Il est interdit de parler breton et de cracher par terre »)
20 Deahl, 2014 and Bhargava, 2015
21 Fallows, James. 2011. 'Technology Is Not The Answer'. The Atlantic.
http://www.theatlantic.com/technology/archive/2011/03/technology-is-not-the-answer/73065/.
22 Pitchfork. 2015. 'James Murphy Shares Remixes Made With Tennis Data Album'. Accessed

September 3 2015. http://pitchfork.com/news/57887-james-murphy-shares-remixes-made-with-

tennis-data-album/.
23 http://www.un.org/en/documents/udhr/
24 We take concise definitions here to broadly outline the major aspects of each of these terms largely

characterized by their literatures, acknowledging the rich and complex histories of each of these
terms.
25 http://www.nap.edu/openbook.php?record_id=4962
26 Livingstone, Sonia, Elizabeth Van Couvering, and N. Thumin. "Converging traditions of research

on media and information literacies." Handbook of research on new literacies (2008): 103-132.
27 Steen, Lynn Arthur. "Numeracy: The new literacy for a data-drenched society." Educational

Leadership 57 (1999): 8-13.

g
28 Barr, Valerie, and Chris Stephenson. "Bringing computational thinking to K-12: what is Involved
and what is the role of the computer science education community?." ACM Inroads 2, no. 1 (2011):
48-54.
29 https://digitalliteracy.cornell.edu
30 Statistic is “i.e. the science that deals with the collection, classification, analysis, and interpretation of numerical facts

or data, and that, by use of mathematical theories of probability, imposes order and regularity on aggregates of more or
less disparate elements” and the result of this practice http://dictionary.reference.com/browse/statistics
31 Crawford, Kate. "Think Again: Big Data." Foreign Policy 9 (2013).

http://www.foreignpolicy.com/articles/2013/05/09/think_again_big_data.
32 We are grateful to Alex Pentland for these comments.
33 Oxford Dictionaries. Oxford Dictionaries Language Matters’ definition of “data”. Accessed

September 2015. http://www.oxforddictionaries.com/us/definition/american_english/data

34 Hilbert, Martin. “How Big is Big Data?” Input to the World Bank 2016 World Development

Report. June 2015: 2.

35 The American Statistical Association uses M. Davidian and T.A. Louis’ definition of “statistics” as

“the science of learning from data, and of measuring, controlling, and communicating uncertainty;
and it thereby provides the navigation essential for controlling the course of scientific and societal
advances.” Science. 2012 Apr 6;336(6077):12. doi: 10.1126/science.1218685.
36 Aldrich, F. and Sheppard, L. 'Graphicacy': the fourth 'R'?, Primary Science Review, 64 (2000): 8-11.
37 Manovich, Lev. "What is visualisation?." Visual Studies 26, no. 1 (2011): 36-49.
38 Deahl 2014
39 https://en.wikipedia.org/wiki/Planned_shrinkage
40 For a fuller discussion see boyd and Crawford, 2010-2011 Letouzé, 2012
41 For a fuller discussion see boyd and Crawford, 2010-2011 Letouzé, 2012
42 Crawford 2013
43 Deahl 2014
44 UNESCO Education Sector. "The plurality of literacy and its implications for policies and

programs: Position paper." Paris: United National Educational, Scientific and Cultural
Organization (2004): 13.
45 boyd, danah and Kate Crawford. (2012). “Critical Questions for Big Data: Provocations for a

Cultural, Technological, and Scholarly Phenomenon.”Information, Communication, & Society 15:5,

p. 662-679. http://www.danah.org/papers/2012/BigData-ICS-Draft.pdf
46 Approaches to Building Big Data Literacy, forthcoming
47 “Data Brokers: A Call for Transparency and Accountability.” Federal Trade Commission, 2014.

https://www.ftc.gov/system/files/documents/reports/data-brokers-call-transparency-
accountability-report-federal-trade-commission-may-2014/140527databrokerreport.pdf
48 “The Network Society: From knowledge to policy.” Center for Transatlantic Relations, 2005.

http://www.umass.edu/digitalcenter/research/pdfs/JF_NetworkSociety.pdf
49 “The Global Education Technology Report 2008-2009: Mobility in a Networked World.” World

Economic Forum, 2009. http://hd.media.mit.edu/wef_globalit.pdf

50 Big Data for Climate Change Resilience Report, Data-Pop Alliance, September 2015.
51 “Big data for Resilience” World Humanitarian Summit
https://www.worldhumanitariansummit.org/file/504310/.../549514
52https://books.google.com/books?id=8eTeAAAAQBAJ&pg=PA5&lpg=PA5&dq=bachelard+opi

nions+wrong&source=bl&ots=j-
W1PyE4vj&sig=WvifaTprDoJQAbvHaKLkcdrfxDg&hl=en&sa=X&ved=0CCYQ6AEwAmoVCh
MI2rvB4eiYyAIVgho-Ch3EIQBt#v=onepage&q=bachelard%20opinions%20wrong&f=false
53 Note: this paper will be translated in French and Spanish
54 https://www.r-project.org
55 Interview with Dorothy Otieno on 24 September 2014

For more information and examples of data journalism, see “Data Dredger,” a Kenyan site on data
journalism: http://internewskenya.org/dataportal/

h
56 Internews in Kenya website. https://edc.org/newsroom/media_coverage/big_data_classroom
57 Internews in Kenya website. https://edc.org/newsroom/media_coverage/big_data_classroom
58 City Digits website. http://www.citydigits.org/
59 “Kenya Data Journalism Fellows Shed Light on Complex Health Issues,” Internews website.

http://www.internews.org/our-stories/news/kenyan-data-journalism-fellows-shed-light-complex-
health-issues
60 “A World that Counts” Nov 2014, http://www.undatarevolution.org/report/