Universal principles

of data ethics
12 guidelines for
developing ethics codes
Professional ethics codes serve a surprisingly broad range of
purposes. On their face, ethics codes set out the standard
for acceptable behavior within a profession. However,
the acts of assembling, deliberating, distributing, training
and enforcing ethics codes ultimately result in much
broader impacts. A set of shared norms helps to define the
boundaries of a professional community and identifies the
standards that the public should demand of practitioners
within that profession. It establishes the type of relationship
that professionals have with the rest of society—including
with their clients, research subjects, users of their services,
and governments.
In many professions, learning expectations about responsibility fiduciary, and legal professionals.
about and complying with and liability. In the long run, This is why establishing a shared
ethical obligations is a marker of ethics codes can play a major set of norms is critically important
accountability. Indeed, the act of role in defining a community of for data scientists and practitioners
becoming licensed in professions professionals. (and those making requests of
such as accounting, law and them). It’s good for the profession
Public discussions about the ethics
medicine includes swearing to and good for society.
of “big data” analytics are rapidly
uphold specific ethical norms.
gaining prominence in public This report discusses the dynamics
In some professions, such as
discourse. The insights derived from involved in generating a code
journalism (and arguably science
data already permeate much of our of ethics that could guide the
and engineering), the ethics of
lives, and promise to shape even profession of data science as it
professional practice are the
more of the opportunities, limits, grows and evolves, and immediately
only commonality defining the
and major and minor life decisions help organizations shape their
boundaries of what is an extremely
we encounter moving forward. In own internal guidelines related to
varied group. Though such codes
other words, data professionals will, data. A broad set of principles is
typically lack the force of law,
in all likelihood, play a role in our proposed and intended to inform
they do ultimately shape legal and
lives that’s as intimate as medical, the development of domain-
regulatory dynamics by establishing

2
specific codes of ethics for specific And so the question of whether data such. Furthermore, many familiar
organizations or industries. scientists and practitioners need a ethical controls, such as informed
Developing a code of ethics should special focus on ethics is ultimately consent, occur only at the point of
be a collaborative effort that a question of whether data science collection. But the power and peril
involves all of the stakeholders in represents a distinctly new way of data science is that data is most
a community and builds from the of knowing. valuable when it can be reused and
proposed principles. Additionally, the repurposed in many different
The way data is used today is more
uses of data science are so diverse contexts and in combination with
than just a technical phenomenon.
(and many are still unforeseen) other datasets.
It’s a political, social, and even
that not every scenario can be
mythological phenomenon that has Personal and sensitive data now
accounted for in a code of ethics.
consequences for how we organize travels unpredictably and will be
Nor does “data science” adequately
our lives and express our values.1 reused indefinitely for unforeseeable
capture the many facets of the
Whatever ethical principles are purposes. Because our “data selves”
data ecosystem. There is a diversity
developed in connection with data, are no longer compartmentalized,
of practitioners that utilize the
they should account for dynamics that many different actors can learn
techniques of data science to
extend beyond technical limitations. intimate details about the lives
provide analysis, insights and advice
Data analytics should be viewed as of anybody who leaves a digital
about a breadth of human activities;
a phenomenon with consequences trail. For this reason, the ethical
all of these actors may have specific
beyond technology, and the infrastructures, concepts, and norms
obligations that differ from data
community should demand that data that have been developed to handle
scientists. Nonetheless, these
scientists and practitioners consider compartmentalized data are often
principles are intended to function
those consequences. neither salient nor applicable—how
as a foundation or outline of what a
data moves in time and space is
universal code of ethics for the data Data analytics is an emerging form of
no longer synchronized with our
science field should emphasize. knowledge production that provides
temporally and geographically
the ability to cheaply and easily
Framing data ethics connect and analyze datasets, often
constrained ethical regimes.5 The
language of medical and scientific
The foremost practical question drawn from highly disparate contexts.2
ethics has long emphasized respect for
for data ethics is whether there is The capacity to continually re-analyze
persons and informed consent as core
anything special about data such that and correlate data collected from a values. But it is a daunting proposition
collecting, manipulating, and applying broad range of contexts has proven to explain how such principles can
it requires a distinct code of ethics. challenging to ethically conceptualize
hold when data about individuals is
The history of science and engineering and regulate.3,4 In the past it could be persistently shared, transformed, and
ethics suggests that ethical regimes assumed that data collected in one
aggregated and when future uses
often track new ways of knowing. context—medical, political, genetic,
of datasets are so unknowable that
As new ways to know the world are social, financial, census, behavioral,
“informed consent” is a misnomer at
developed, appropriate rules governing geographic, etc—would stay in that best—and impossible at worst.
those approaches are helpful. context and could be regulated as

3
Professional codes of data ethics
Analyses of professional ethics codes show that the articulation of shared values is often a key stage in the
professionalization of a field: It establishes who is a member of the field and what can be expected of them
by colleagues, clients and society at large.6,7 Mark Frankel offers a taxonomy of professional ethics codes as
aspirational, educational, and regulatory, noting that most codes are an admixture and serve multiple goals. He
argues that the process of establishing a code provides opportunities for critical reflexivity that are perhaps more
important than the final product: “This process of self-criticism, codification, and consciousness-raising reinforces
or redefines the profession’s collective responsibility and is an important learning and maturing experience for
both individual members and the profession.”8

In an analysis conducted for the Council for Big Data, Ethics & Society, Jacob Metcalf identified the
inward—and outward-facing goals of professional ethics codes that may be applicable for data ethics:9

Inward-facing goals: Outward-facing goals:

• Provide guidance when existing implicit norms and • Protect vulnerable populations and individuals who
values are not sufficient; essentially, guidance for a could be harmed by the profession’s activities
novel situation
• Protect and enhance the good reputation of and
• Reduce internal conflicts; strengthen the sense of trust for the profession
common purpose among members of the organization
• Establish the profession as a distinct moral
• Satisfy internal criticism from members of a profession community worthy of autonomy from external
control and regulation
• Create generalized rules for individuals and
organizations that have responsibilities for important • Provide a basis for public expectations and
“human goods” evaluation of the profession

• Establish role-specific guidelines that demarcate • Serve as a basis for adjudicating disputes among
general principles as particular duties members of the profession and disputes between
members and the public
• Establish standards of behavior toward colleagues,
students/trainees, employees, employers, clients • Create institutions that are resilient in the face of
external pressures
• Strengthen the sense of common purpose among
members of an organization • Respond to past harms done by the profession.

• Deter unethical behavior by identifying sanctions and

creating an environment in which the reporting of
unethical behavior is affirmed

• Provide support for individuals who come under

pressure to behave in an unethical manner

4
There are already some ethics codes contains some principles that do
that cover most computing and data still hold up—such as striving to
scientists and engineers. In the US, maintain the integrity of data about
four major computing professional individuals—it lacks the specificity
societies have substantially different that would make the code optimally
codes for their members due to their useful to current and future
different missions.10 The Association generations of data and computing
of Computing Machinery (ACM), the professionals. Other professional
largest professional organization groups that are more closely
for computer scientists and associated with the data revolution
engineers, distributes an ethics code have more recent codes. The recently
for members of its organization.11 founded Data Science Association
However, that code was adopted offers a relatively detailed ethics
in 1992 at the beginning of the code that is notable for detailing
internet age, predating many of how members should adhere
the technologies that define the closely to scientifically sound
ethical conflicts faced by data and statistical methods.12 For example,
computing professionals today. rule 8(d) reads:
Although the ACM’s ethics code
“ If a data scientist reasonably believes
a client is misusing data science to
communicate a false reality or promote an
illusion of understanding, the data scientist
shall take reasonable remedial measures,
including disclosure to the client, and
including, if necessary, disclosure to the
proper authorities. The data scientist shall
take reasonable measures to persuade the
client to use data science appropriately.”
Source: Code of Conduct, Data Science Association

5
Some data science sub-disciplines humanities, social science, criminal physicians) and to a more specific
have also produced valuable ethics justice, geography and geospatial set of obligations that apply only to
codes and other types of ethics imaging, manufacturing, social their own members. If data science
guidance for their members. The work, human rights, and many more. continues on its path to ubiquity,
Association of Internet Researchers This poses a major challenge for a then it may be challenging to define
(AoIR) developed an ethics code universal code of data ethics: There a truly universal code that covers its
in 2002, updated in 2012, that may be too few commonalities uses in such a variety of contexts.
addresses the obligations of social across the specific uses of data
One of the quirks of data science
science researchers working in digital science to pull together a single
is that its parent fields have
domains at a macro-level.13 This code. Principles of data ethics that
traditionally fallen outside of the
document is notable for the extensive hold in medicine may not hold in
purview of US federal research
list of questions internet researchers finance because the social roles
ethics regulations. Following a long
should address. The National Center occupied by medical professionals
arc of infamous research scandals
for Education Statistics produced and financiers differ significantly.
in the mid-20th century—ranging
a guide for appropriate use of They have meaningfully different
from Nuremberg to Tuskegee to
educational data in 2010, that mixes obligations to their clients and
the Stanford prison experiment—
core principles with illustrative society, and so it is reasonable to
the 1974 National Research Act
case studies.14 expect that their uses of big data for
empowered federal regulators to
good and ill will similarly vary.
Challenges for a universal identify, define, and enforce ethical
code of data ethics Furthermore, many of these fields standards for human-subjects
already have their own professional research that uses federal funds. The
A unique aspect of today’s ethics codes that may or may not authors of the 1979 Belmont Report
datasets is their sprawling, multi- address the changes introduced commissioned by the Act identified
disciplinary utility—data science is by the data age. Other fields the three primary principles of
arguably closer to a service than have dealt with such problems by bioethics: beneficence (research
a discipline because it is useful in having professional sub-societies should be carefully constructed to
so many industries and disciplines. formulate secondary ethics codes. do good in the world), respect for
The analytical tools developed in For example, the American College persons (research must respect
applied mathematics, statistics, and of Obstetricians and Gynecologists personal values such as autonomy,
computer science are being taken holds its members both to the privacy and dignity) and justice
up by disciplines and sectors such American Medical Association’s (research must further social equity).
as medicine, marketing, finance, the code of ethics (that applies to all

6
These principles subsequently even these regulated standards
informed the rulemaking process do not go far enough. In the
initiated by the Department of humanitarian field, some academics
Health and Human Services that and practitioners are beginning
resulted in the federal regulations to call for higher standards.16
known as the Common Rule. They argue that “demographically
The Common Rule now governs identifiable data”—a broader
(nearly) all human-subjects research classification than Personally
funded by federal agencies. Its Identifiable Information (PII),
most consequential outcome was the gold standard for privacy
establishing Institutional Review professionals—could cause various
Boards (IRBs) as an obligatory harms to entire classes of people.
milestone for most academic
As data science matures as a
research. However, computer
field and increasingly affects the
science and engineering, applied
human condition, there’s a chorus
mathematics, and much quantitative
building among professionals and
sociology research has historically
practitioners to have more guidance
fallen outside of the regulatory
for the ethical decisions they are
definition of “human subjects,”
forced to make—and might be
even when these fields involve
unaware they are making—on a
human lives.15 As a result, most
daily basis. The set of Principles
professionals trained in the parent
proposed below is intended to
fields of data science do not
provide a baseline for those seeking
encounter the primary research
such guidance and those looking
norms and regulatory apparatuses
to develop a group-specific code of
that guide other science and
data ethics.
engineering fields.

The Common Rule and IRBs

dominate conversations about
practical ethics, but in some cases,

7
Principles for Data Ethics
Data science professionals and practitioners should strive to perpetuate these principles:

1. The highest priority is to respect the persons behind the data.

When insights derived from data could impact the human condition, the potential harm to
individuals and communities should be the paramount consideration. Big data can produce
compelling insights about populations, but those same insights can be used to unfairly limit an
individual’s possibilities.

2. Attend to the downstream uses of datasets.

Data professionals should strive to use data in ways that are consistent with the intentions and
understanding of the disclosing party. Many regulations govern datasets on the basis of the status
of the data, such as “public,” “private” or “proprietary.” However, what is done with datasets is
ultimately more consequential to subjects/users than the type of data or the context in which it is
collected. Correlative uses of repurposed data in research and industry represents both the greatest
promise and the greatest risk posed by data analytics.

3. Provenance of the data and analytical tools shapes the consequences of their use.
There is no such thing as raw data—all datasets and accompanying analytic tools carry a history
of human decision-making. As much as possible, that history should be auditable, including
mechanisms for tracking the context of collection, methods of consent, the chain of responsibility,
and assessments of quality and accuracy of the data.

4. Strive to match privacy and security safeguards with privacy and security
expectations.
Data subjects hold a range of expectations about the privacy and security of their data and those
expectations are often context-dependent. Designers and data professionals should give due
consideration to those expectations and align safeguards and expectations as much as possible.

5. Always follow the law, but understand that the law is often a minimum bar.
As digital transformations have become a standard evolutionary path for businesses, governments
and laws have largely failed to keep up with the pace of digital innovation and existing regulations
are often mis-calibrated to present risks. In this context, compliance means complacency. To excel
in data ethics, leaders must define their own compliance frameworks that outperform
legislated requirements.

6. Be wary of collecting data just for the sake of more data.

The power and peril of data analytics is that data collected today will be useful for unpredictable
purposes in the future. Give due consideration to the possibility that less data may result in both
better analysis and less risk.

8
7. Data can be a tool of inclusion and exclusion.
While everyone deserves the social and economic benefits of data, not everyone is equally
impacted by the processes of data collection, correlation, and prediction. Data professionals
should strive to mitigate the disparate impacts of their products and listen to the concerns of
affected communities.

8. As much as possible, explain methods for analysis and marketing to

data disclosers.
Maximizing transparency at the point of data collection can minimize more significant risks as
data travels through the data supply chain.

9. Data scientists and practitioners should accurately represent their qualifications,

limits to their expertise, adhere to professional standards, and strive for peer
accountability.
The long-term success of the field depends on public and client trust. Data professionals should
develop practices for holding themselves and peers accountable to shared standards.

10. Aspire to design practices that incorporate transparency, configurability,

accountability, and auditability.
Not all ethical dilemmas have design solutions, but being aware of design practices can break
down many of the practical barriers that stand in the way of shared, robust ethical standards.
Data ethics is an engineering challenge worthy of the best minds in the field.

11. Products and research practices should be subject to internal, and potentially
external ethical review.
Organizations should prioritize establishing consistent, efficient, and actionable ethics review
practices for new products, services, and research programs. Internal peer-review practices can
mitigate risk, and an external review board can contribute significantly to public trust.

12. Governance practices should be robust, known to all team members and
reviewed regularly.
Data ethics poses organizational challenges that cannot be resolved by familiar compliance
regimes alone. Because the regulatory, social, and engineering terrains are so unsettled,
organizations engaged in data analytics require collaborative, routine and transparent practices
for ethical governance.

9
100/365-day Plans
Over the course of the next year, every organization can be well on its way to leveraging these
12 universal principles to develop a custom-tailored code of data ethics.

In three months, your organization should:

Collect codes of ethics from

these ecosystems, industry
Catalog the ecosystems, groups, and professional
industry groups, and organizations. Aggregate the
professional organizations parts of these codes that deal
where your organization is a with the handling of data and
member or participates in a use them as a minimum bar
meaningful way. for your own code.

Identify the internal and Highlight the relationships

external stakeholders who in the list above where
should contribute to or data is shared or sold.
review your emerging code
of ethics.

10
In one year (and beyond), your organization should strive to:

Share outcomes of
the pilot with all
Circulate an early draft stakeholders and
Once ratified, publish your of your code among notify them when and
code of ethics for public stakeholders and have how they will be held
consumption and consider them indicate existing accountable for being
submitting it to the Center practices that would able to demonstrate
for the Study of Ethics in require modification if the compliance with
the Professions. code were to be ratified. the code.

Note the existing Encourage partners to After incorporating insights from prior
practices that require publicly publish and discussions, publish a code of ethics
modification and commit to abide by this among internal stakeholders and
consult with the process new code of ethics. partners who will be participating in a
owners to understand 12-month pilot of the draft code; once
any impediments to the pilot starts, interview stakeholders
adopting more rigorous and partners every three months
ethical practices. to understand how their work was
impacted. With the insights from the
completed pilot, make a decision to ratify
or update the draft code of data ethics.

11
References Contact Us Data Ethics Research Initiative
1 Crawford K, Gray ML and Miltner K (2014) Critiquing Steven C. Tiell Launched by Accenture’s Technology
Big Data: Politics, Ethics, Epistemology. International
Senior Principal—Digital Ethics Vision team, the Data Ethics
Journal of Communication 8(0): 10.
Accenture Labs Research Initiative brings together
2 Mayer-Schönberger V and Cukier K (2013) Big Data: A
leading thinkers and researchers
Revolution that Will Transform how We Live, Work, and [email protected]
Think. Houghton Mifflin Harcourt. from Accenture Labs and over a
3 Zwitter A (2014) Big Data ethics. Big Data & Society dozen external organizations to
1(2). DOI: 10.1177/205395171455925. Jacob Metcalf explore the most pertinent issues of
4 Metcalf J, boyd danah and Keller EF (2016) Ethical Resolve data ethics in the digital economy.
Perspectives on Big Data, Ethics, and Society. Council for
Big Data, Ethics, and Society. (accessed 31 May 2016). [email protected] The goal of this research initiative
5 Metcalf J and Crawford K (2016) Where are
is to outline strategic guidelines
and tactical actions businesses,
human subjects in big data research? The emerging
ethics divide. Big Data & Society 3(1): 1–14. DOI:
About Accenture Labs government agencies, and NGOs
10.1177/2053951716650211. Accenture Labs invents the future can take to adopt ethical practices
6 Metcalf J (2014) Ethics Codes: History, Context, and for Accenture, our clients and the throughout their data supply chains.
Challenges. Council for Big Data, Ethics, and Society.
(accessed 21 October 2015).
market. Focused on solving critical
7 The Illinois Institute of Technology maintains a
business problems with advanced
thorough collection of professional ethics codes as part technology, Accenture Labs brings About Accenture
of their Center for the Study of Ethics in the Professions, fresh insights and innovations to Accenture is a leading global
8 Frankel MS (1989) Professional codes: why, how, and our clients, helping them capitalize professional services company,
with what impact? Journal of business ethics 8(2-3):
109–115.
on dramatic changes in technology, providing a broad range of services
9 Metcalf, J (2014). See also: Frankel MS (1989);
business and society. Our dedicated and solutions in strategy, consulting,
Gaumnitz BR and Lere JC (2002) Contents of Codes of team of technologists and digital, technology and operations.
Ethics of Professional Business Organizations in the researchers work with leaders across Combining unmatched experience
United States. Journal of Business Ethics 35(1): 35–49;
Kaptein M and Wempe J (1998) Twelve Gordian Knots the company to invest in, incubate and specialized skills across more
When Developing an Organizational Code of Ethics. and deliver breakthrough ideas and than 40 industries and all business
Journal of Business Ethics 17(8): 853–869.
solutions that help our clients create functions—underpinned by the
10 Oz E (1993) Ethical standards for computer new sources of business advantage. world’s largest delivery network—
professionals: a comparative analysis of four major
codes. Journal of Business Ethics 12(9): 709–726. Accenture works at the intersection
Accenture Labs is located in six key
11 “ACM Code of Ethics and Professional Conduct.” ACM of business and technology to help
research hubs around the world:
Code of Ethics and Professional Conduct. ACM, Inc. 16 clients improve their performance
October 1992. Web. 31 May 2016. Silicon Valley, CA; Sophia Antipolis,
and create sustainable value for their
12 “Code of Conduct | Data Science Association.” Data France; Arlington, Virginia; Beijing,
stakeholders. With approximately
science code of professional conduct. Data Science China; Bangalore, India, and Dublin,
Association. Accessed 31 May 2016. 373,000 people serving clients in
Ireland. The Labs collaborates
13 http://aoir.org/reports/ethics2.pdf. Accessed 31 May more than 120 countries, Accenture
extensively with Accenture’s
2016. drives innovation to improve the way
network of nearly 400 innovation
14 https://nces.ed.gov/pubsearch/pubsinfo. the world works and lives. Visit us at
asp?pubid=2010801. Accessed 31 May 2016. centers, studios and centers of
www.accenture.com.
15 Metcalf J and Crawford K (2016) Where are human excellence located in 92 cities and
subjects in big data research? The emerging ethics 35 countries globally to deliver
divide. Big Data & Society 3(1): 1–14.
cutting-edge research, insights
16 Karunakara U (2014) Data Sharing in a Humanitarian
Context: The Experience of Médicins Sans Frontières. In:
and solutions to clients where
Moore SA (ed.), Issues in Open Research Data, London: they operate and live. For
Ubiquity Press. more information, please visit
www.accenture.com/labs.
© 2016 Accenture.
All rights reserved. Learn more: www.accenture.com/DataEthics
This work is licensed under the Creative Commons
Attribution 4.0 International License. To view a copy
of this license, visit http://creativecommons.org/ This document makes descriptive reference to trademarks that may be owned by others.
licenses/by/4.0/ or send a letter to Creative Commons,
PO Box 1866, Mountain View, CA 94042, USA. The use of such trademarks herein is not an assertion of ownership of such trademarks by Accenture and is not intended
Accenture, its logo, and High performance. to represent or imply the existence of an association between Accenture and the lawful owners of such trademarks.
Delivered. are trademarks of Accenture.