Big Data, Communities and Ethical Resilience: A Framework for Action

Big Data, Communities and Ethical Resilience:
A Framework for Action
By 2013 Bellagio/PopTech Fellows Kate Crawford, Gustavo Faleiros, Amy Luers,
Patrick Meier, Claudia Perlich and Jer Thorp
Draft Date: Oct. 24, 2013
In August 2013, a multidisciplinary group gathered at the
Rockefeller Foundation’s Bellagio Center to address the theme
of “Community Resilience Through big data and Technology.”
Creative and critical thinkers were selected from the
technology sector, academia, the arts, humanitarian and
ecological spheres. Over ten days, we explored how data
could be used to help build community resilience in the face of
a range of stresses — environmental, political, social and
economic. Large data collection and analysis may support
communities by providing them with timely feedback loops on
their immediate environment. However, the collection and use
of data can also create new vulnerabilities and risks, by
enabling discriminating against individuals, skewing evidence,
and creating dependencies on centralized infrastructure that
may increase a system’s vulnerability. After analyzing these
risks and opportunities, we developed a framework to help
guide the effective use of data for building community-driven
resilience. In this framework, we propose six domains: ethics,
governance, science, technology, place and sociocultural
context. We believe that by considering all six domains
together, organizations can safeguard against predictable
failures by exposing project weaknesses from the outset rather
than in hindsight.
Big Data, Communities and Ethical Resilience: White paper! 1

Introduction
What is resilience?
The term “resilience” is both complex and shifting, and can be used to mean many
different things. First, we define resilience as the ability to absorb and cope with
unexpected shocks. To this end, resilient communities are those with a particular
capacity to respond to disturbance or hazards. In the popular discourse resilience
often focuses on resources and infrastructure and overlooks issues of power, ethics
and accountability — which recent research suggests are critical.1 Resilience is not
a normative term: systems characterized as resilient may be either desirable or
undesirable. As a result, if ethics is not consciously considered at the inception of
data projects, steps taken to increase community resilience could in fact create
more vulnerability and thus do harm. This explains why our focus is on ethical
resilience – and how we might apply this idea to data-driven community projects.
What is ‘big data’?
“Big data” is a hype-laden concept, but it points to a multi-faceted, emerging
phenomenon.2 Data collection and processing have changed dramatically in the
last five years. Where once data was collected only for a specific purpose, now
massive amounts of data are opportunistically and passively collected and reused in
multiple contexts, over an indefinite time frame, often without informed consent.
Beyond a shift in scale and persistence, data have become more granular and
intrusive, and profoundly interlinked. Using data for predictive modeling can allow us
to “fill in the blanks,” anticipating information about others without even needing to
collect it. big data is thus challenging our assumptions about what can and can’t be
private. And as recent studies have shown, even anonymized data sets can be
used to re-identify people.3 In such a context, even the idea of “privacy” is not
enough — we need to think more broadly to assess how people might be affected
by big data projects. How will these new attributes of data analysis and predictive
1 O'Brien, K., B. Hayward, and F. Berkes. 2009. Rethinking social contracts: building resilience in a changing
climate. Ecology and Society 14(2): 12. [online] URL: http://www.ecologyandsociety.org/vol14/iss2/art12/
Synthesis
2 boyd, d. and K. Crawford. 2011. Six Provocations for big data. A Decade in Internet Time: Symposium on the
Dynamics of the Internet and Society. [online] URL: http://blogs.gartner.com/svetlana-sicular/big-data-is-falling-
into-the-trough-of-disillusionment/
3 Montjoye, Y., C. Hidalgo, M. Verleysen, and V. Blondel. 2012. Unique in the Crowd: The privacy bound of
human mobility. Scientific Reports 3. [online] URL: http://www.nature.com/srep/2013/130325/srep01376/full/
srep01376.html

modeling affect community resilience? How can we enhance the capacities of
communities without doing harm?
Framework
To address these questions and others that arise when conducting data-oriented
community resilience projects, we propose a framework emphasizing six domains:
Ethics, Governance, Science, Technology, Place, and Sociocultural context. This
framework is depicted in Figure 1 below.
Figure 1: The arrangement of the six domains above reﬂects both similarities between domains
(through adjacency) and differences (through oppositions).
Ethics is placed at the top of the hexagon, to emphasize its importance when
designing community-based data projects. It is the primary guiding principle, which
should be considered ﬁrst and then in connection to the other domains. By doing
so, we argue that the most substantial risks can be fully assessed and considered
before data projects commence. What follows is a detailed explanation of each
domain, as well as a discussion of how multiple domains can usefully interact.

1. Ethics
What is it?
Fairness, Agency, Consent, Social Justice, Participation
Why does it matter?
Ethical practices allow people to participate in how, where and for what
purposes their data is used, which strengthens agency and leads to
stronger communities.
Ethics is currently one of the most overlooked elements of big data science.
Developing a strong set of ethical principles is essential to ensuring that community-
based data projects do no harm. While privacy is important, it is too narrow a
concept to adequately address the risks of what big data projects can do. By
focusing on ethics, we can more broadly consider the human impact — both short
and long term — of how data is being gathered and used. Unfortunately, ethical
considerations are often sidelined because gathering more data is seen as
necessarily better, and concerns about how data might be abused are only
considered after the fact. The result is data sets that are overly intrusive, that collect
personally identifying information without informed consent, and may have serious
unintended consequences, particularly when brought together with other kinds of
personally identifying data sets.
At its core, an ethical framework for data aims to enhance agency: the ability for
individuals and communities to be able to make decisions about how, where, why
and for how long their data is used. By emphasizing agency for community
resilience projects, we are underscoring the responsibility for data projects to allow
both granular control of data collection and retention, and the ability for users to opt
out. However, the burden of responsibility should not be placed solely on individual
users to understand and respond to the privacy guidelines or data retention
practices of a project. Agency means giving people the capacity to engage with
their data and participate, but the onus of responsibility still lies with the project
developers. Ensuring this combination of data control, ethical engagement and
responsibility builds trust and conﬁdence in community data projects. We also think
it makes for better data science.
The negative impacts of data collection and analysis are not evenly distributed.
Some communities may be more exposed than others, and vulnerable individuals
and groups can face serious risks. For example, the use of crowdsourcing for crime

reporting in Mexico resulted in drug cartels tracking and lynching participants.4
Predictive policing models, which use big data to create neighborhood ‘hot spots’
where crime is expected to occur, risk reinforcing existing racial and class
prejudices and institutionalizing differential policing.5 Finally, community activists in
many countries are subject to email and cellphone monitoring.6 The inability to
escape data surveillance affects people’s freedom to express themselves, to
engage in political activism or to trust their governments.
Ethical principles need to be incorporated from the outset when designing or
assessing projects that focus on supporting community resilience. These principles
should include informed consent, data ownership, accountability, transparency,
data protection and data access.7 Informed consent goes beyond merely making
people aware of the terms of service or that data is being gathered about them, but
a clear articulation of how data might be used, whether third parties will get access
to that data, and how people can opt out or limit how much of their data is
gathered. Data ownership, for example, means that a community owns the data it
generates, as well as the learning algorithms and other derivatives from data
projects. Best practices of accountability and transparency would enforce greater
responsibility about how data is used, and allow participants to see where their data
goes and alter or remove their data where possible. Ideally, data protection and
data access should give people more certainty that their data is being stored as
securely as possible, while also giving individuals access to their own data.
While not all projects will be able to achieve these goals, we suggest that they can
serve as a set of best practices. We also suggest the need to consider big data due
process: ensuring fairness when big data is used to make decisions that affect
individuals and communities.8
4 Monroy-Hernandez, A., E. Kiciman, D. Boyd, and S. Counts. 2012. Tweeting the Drug War: Empowerment,
Intimidation, and Regulation in Social Media. HCIC. [online] URL: http://research.microsoft.com/apps/pubs/
default.aspx?id=168809
5 Robertson, J. 2013. How big data Could Help Identify the Next Felon - Or Blame the Wrong Guy. [online] URL:
http://www.bloomberg.com/news/2013-08-14/how-big-data-could-help-identify-the-next-felon-or-blame-the-wrong-
guy.html
6 For example, see Solove, D. 2011 Nothing to Hide: The False Tradeoff Between Privacy and Security. New
Haven: Yale University Press.
7 For example, see the Fair Information Practice Principles: http://www.ftc.gov/reports/privacy3/fairinfo.shtm
8 Crawford, K., and J. Schultz. 2013. big data and Due Process: Towards a Framework to Redress Predictive
Privacy Harms. Boston College Law Review. 55(1). [online] URL: http://papers.ssrn.com/sol3/papers.cfm?
abstract_id=2325784

2. Governance
What is it?
Public Policy, Corporate Policy, Social Networks
Why does it matter?
Good governance ensures that communities are placed at the center of
data-driven projects
Governance, in an ethical resilience context, focuses on how public policy and
corporate data policies can better prioritize and serve local communities by
supporting community resilience projects that are genuinely people-centered,
community driven and locally owned.
Data-driven approaches to community resilience run the risk of promoting a
technical discourse that crowds out local participation: emphasizing technology
rather than the multiplicity of approaches and perspectives.9 We are concerned by
deﬁnitions of resilience that emphasize external expertise and technological
solutions as these are often misrepresented as politically neutral. More informed
policies at a local and national level can prioritize community-driven participation
and indigenous knowledge rather than solely emphasizing technology and
infrastructure investments. Policies for ethical resilience can foster a balance
between external assistance and expertise and the pre-existing capacity of local
communities to help themselves. In sum, the responsibility of public policy for
ethical resilience is to foster an environment conducive to respectful reciprocity, self-
governance, improvisation and mutual aid.
The governance domain also encompasses corporate governance, which typically
dictates data management policies such as access, ownership and privacy.
Corporate data collection policies that are formulated without direct input from local
communities may violate community norms and practices. As noted above, digital
intrusion and surveillance erodes trust and threatens community resilience.
Corporate governance should therefore promote respectful data management
policies that enable individuals to determine if, how and for what purposes and for
how long their data can be used.
9 Meier, P. 2013. How to Create Resilience Through Big Data. iRevolution. [online] URL: http://iRevolution.net/
2013/01/11/disaster-resilience-2-0

At best, governance models can facilitate and protect environments conducive to
local and ethical self-organization. This means acting as a mediator between local
and external interests. Good governance can support the transparency and
accountability structures necessary for a just and ethical approach to community
resilience.
3. Science
What is it?
Hypothesis Testing, Purposeful Data Collection, Projections, Theory
Why does it matter? Science allows people to assess causality, as
opposed to simple correlation from big data, and to identify
vulnerabilities and opportunities for resilience in complex systems.
Scientific knowledge is critical for building resilience as it enables society to
understand trends, identify thresholds and anticipate vulnerabilities. However, the
rise of data science, which emphasizes finding meaning in patterns, has begun to
threaten elements of traditional scientific method, which purposely collects data to
test hypotheses about how a system works, and conducts experiments that are
reproducible. Without these characteristics of science we may create vulnerabilities
in our knowledge base.
Big data science is used by government, corporations and international
development organizations to identify patterns to better “serve the people.” Indeed,
correlations made with big data do provide an opportunity for improved decision
making by increasing the speed of analysis, enabling more targeted response and
facilitating greater opportunities for citizen participation in the production of
knowledge. However, taken in isolation outside of the traditional modes of scientific
inquiry, data science can increase inherent biases and lead to erroneous
conclusions when correlation is mistaken for causation.10 There have been notable
advances in data science around reliable causal analysis from observational data
that has to be considered.11
Perhaps the biggest challenge regarding the focus on big data for building
community resilience is that much of this is heavily centered on analysis data
10 Crawford, K. 2013. Hidden Biases in big data. Harvard Business Review [online] http://blogs.hbr.org/2013/04/
the-hidden-biases-in-big-data
11 Van Der Laan, M., and S. Rose. 2011. Targeted Learning: Causal Inference for Observational and
Experimental Data. Springer Series in Statistics

sensed in human systems — search queries, social media, and mobile activities.
Such data sources are inherently linked to the here and now. Without a longer term
understanding of the social and biophysical systems in which these data are
captured, they provide no ability to understand future trends.12 For example,
consider the challenge of managing the risks of increasing shocks from climate
change. While the collection and processing of huge amounts of data has been the
foundation of climate change science, enabling a picture of possible futures, the
collection and analysis of social data over the last decade has provided little
guidance on how social and natural systems are likely to respond to these changes.
When using data science to inform and build community resilience in the face of
uncertainty of future shocks, it will be critical to not take this in isolation from
traditional scientiﬁc method.
4. Technology
What is it?
Modes of Data Collection, Storage, Analytics, Coordination
Why does it matter?
Technology provides new and more opportunistic modes to collect
potentially valuable data, provides tools for coordination but can create
dependencies and introduce new vulnerabilities
The main drivers of the recent big data trend are advances in technology around
data sharing, data collection, predictive modeling and storage. While data and data
analysis has been used for centuries, big data approaches have changed the nature
of how information is stored and analyzed. The size of data is less relevant than the
vastly increased level of granularity and the ability to capture not just information
about entities and people but on the very speciﬁc connections between them. With
this new richness and complexity of big data has come a growing expectation that
technology should play a large part in emergency response and resilience.
In order to meet this expectation, we need to understand that technological
systems are often very fragile. For example, many communications and power
systems come under enormous stress during crisis events, leading to instabilities in
technologies that rely on them. Another consideration is the correlation of the
12 Luers, A. 2013. (Appropriate) big data for Climate Resilience?. Stanford Social Innovation Review. [online]
URL: http://www.ssireview.org/blog/entry/appropriate_big_data_for_climate_resilience

availability of networked technology (resilient technology in particular) with economic
means. In such cases, sensing and response solutions need to consider and
actively pursue approaches that include community members on the lower
economic scale, who may not use smartphones or social media.
One important consideration in developing technology for resilience is the
acknowledgement that the goal has to be to empower resilient characteristics in
communities rather than developing top-down technological solutions. This must
include an integration of local leadership and structure as much as the need to have
local skills and knowledge to sense, interact with, and evaluate data. Technological
approaches should also consider the broader environment. Many large data
projects require data centers, which draw considerable resources like water and
energy. Making these processes more sustainable will be a priority for community
resilience in the coming years.
To facilitate a tight integration of the community and to allow local knowledge
sharing, data solutions should give preference to broadly available and open source
tools (R, hadoop) as well as open standards (XML, JSON) where possible. Not only
are these tools low cost and therefore more accessible to communities, they are
often in the long run more stable as they do not depend on the ﬁnancial well-being
of a company. Their code is also open to inspection and auditing. Human skill
remains the main success factor in data-driven projects. Being able to attract, train,
and maintain that human capital locally is key to developing resilient data-driven
solutions; relying on non-local expertise that may not have strong ties to a region or
community creates new kinds of vulnerability and does not build data skills at a
local level. A more detailed discussion of important technology consideration and
an outline of best practices in non-commercial data is provided here.13 Above all,
the design of data projects for communities should emphasis broad participation,
shared awareness, and strong boundaries for privacy and ethics.
13 ‘Seven Principles for Big Data and Resilience Projects’, http://irevolution.net/2013/09/23/principles-for-big-
data-and-resilience/

5. Place
What is it?
Natural Environment, Built Environment, Infrastructure, Climate, Cities,
Neighborhoods
Why does it matter?
Community resilience is determined by place based peculiarities. Data
generated on the local level strengthens decision making
A consideration of place is critical for building resilience. Both the natural and built
environment define vulnerabilities and determine opportunities for resilience. The
application of data science to increase ethical community resilience may consider
the influence of historical changes made in the landscape and the accumulated
knowledge of locals about their own environment. The relationship of communities
with place can be a source of resilience as well as risks.
Weather patterns, landscape transformation and availability of natural resources
shape local knowledge and define the capacity to absorb shocks. Nevertheless,
climate change and the depletion of forests, water stocks, among other
environmental stresses, are redefining established conditions. Constant
assessments are needed. The access to more location specific data, like high
resolution satellite imagery or geo-tagged media (photos, reports) can help to build
greater situational awareness and can help to improve predictions and response.
Geolocated data can serve as a tool for local decision making.
The ágora, the city, the borough, the prefecture or the council, among other
instances of local governance, should be forces working towards stronger and
more resilient communities. But there can be considerable tensions between local
and global impacts: a hydroelectric dam can be considered a secure source of
energy for a region, but also can act as a disruptive event for local communities that
will be displaced. Meanwhile, an investment in extracting oil may improve the
medium-term economic conditions of one specific location, but undermines the
need to reduce emissions of greenhouse gases at a global level.
Because of these tensions between local and global impacts, hyperlocal data
generated by sensors in many cities may not be effective for strengthening
resilience unless an ethical approach is taken to give individuals and communities a
choice as to whether they will participate or not in sensor data collection and use.
6. Sociocultural Context

What is it?
Customs, Language, Ideologies, Art, Media, Journalism
Why does it matter?
As data becomes culture, and vice versa, a deep understanding of
sociocultural conditions becomes critical.
Sociocultural context encompasses concerns related to a specific population, and
the behaviors that define them. This domain includes a variety of concepts,
including social customs, language, and art. It also includes a number of broadly
important considerations for communities and individuals, such as gender, race and
class, as well as values, ideologies, and norms. The sociocultural domain also
addresses media and journalism, religious practice, and cultural output such as
literature, performance or film.
In the frame of data and community resilience, there are many ways in which
sociocultural context is important. Norms around usage of personal data can vary
drastically from one culture to another. Technologies or methods that do not infringe
on an individual depending on their gender, race and class may nonetheless be
dangerous to a different individual within the same community (for example, New
York City’s controversial “stop and frisk” policy).14 Thus, an understanding of
sociocultural diversity must be built into ethical community resilience thinking.
Sociocultural context should not only be seen as a set of conditions which should
be considered in thinking around data and resilient communities. For example, the
arts can play an active role in promoting and facilitating data literacy. By moving
data discussions from restricted pockets of academia, corporations and
governments to the public sphere, data practices can become part of the broader
cultural discourse. As a result, communities as a whole can be more informed when
making decisions based on their own data and that of their fellow citizens. This in
turn builds a system of trust, leading to more resilient communities.
An example of such an approach is Nuage Vert, a public art project by Finnish
artists Helen Evans and Heiko Hansen.15 In this project, real-time data about
electricity consumption in Helsinki is projected onto vapor emissions from a local
14 Editorial Board. 2013. Racial Discrimination in Stop-and-Frisk. New York Times. [online] URL: http://
www.nytimes.com/2013/08/13/opinion/racial-discrimination-in-stop-and-frisk.html
15 Andrews, K. 2009. Nuage Vert “Green Cloud” Illuminates Emissions. [online] URL: http://inhabitat.com/green-
cloud-hehe-helsinki-environmental-art/

smokestack. Visible for miles, the project makes consumption data a part of the
built environment. In doing so, it generates discussion amongst the community, and
ultimately seeks to persuade people to change patterns of consumption.
An important additional factor to consider in the context of culture is narrative. The
way that resilience and data are framed within a society can greatly influence the
trust and cooperation generated within the community. The fostering of community-
specific narratives may strongly influence the overall success of resilience initiatives.
Projects that may originate from outside communities need to consider that local
narratives may differ significantly from those of the originating group.
Conclusion
This document offers a framework for thinking and action for big data community
resilience projects. In our conversations, we focused on how future projects might
avoid the failures of previous data-oriented community undertakings. Of primary
importance is to focus on people-centered, community-driven approaches. The
discourse of big data and community resilience often excludes local participation by
less powerful or technically literate populations. As a result, external experts may
reduce complex social problems like community resilience to terms that are suited
to technological solutions. This crowds out local knowledge, participation and
agency, which undermines trust, social connectedness and resilience. Clear public
policy and corporate governance frameworks are needed to foster a generative and
inclusive environment that is conducive to local communities participating in their
own data projects.
Human skill is a key element of effective long-term data-driven projects. This makes
local human collaboration an essential component of data projects for resilience. To
enable greater local skills development and engagement, we suggest a
combination of using open standards for data, common tools for the analytical
components and educational workshops that develop data skills in the local
community. To ensure the effectiveness of more community-oriented data projects,
a set of ethical guidelines should be drafted at the beginning - rather than
considering ethics and privacy merely as a “compliance process” at the end of a
project. We hope this framework will serve as a provocation to consider the many
possible human impacts of data projects much earlier in their design.
Lastly, while community-based projects are typically memorialized by their
successes, there must be openness around and communication about failures as
well. By embracing the possibility of failure, risks and challenges that comes with
the process of embracing new technological approaches, we can all better learn
from each other.

We are at a critical time in the emergence of big data science and its application to
enhance the resilience of communities. We need to ask searching questions about
how best to do this in a way that enhances the agency and capacity of
communities without exposing them to new risks. By considering ethics as a
primary guiding principle, and then the connections with governance, science,
technology, place and the wider sociocultural context, we can produce
groundbreaking data projects that can truly enhance the ability of communities to
thrive.

Big Data, Communities and Ethical Resilience: A Framework for Action

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Big Data, Communities and Ethical Resilience: A Framework for Action (20)

More from The Rockefeller Foundation (20)

Recently uploaded (20)

Big Data, Communities and Ethical Resilience: A Framework for Action