Plaintext
ForHumanity
THE RISE OF THE
ETHICS COMMITTEE:
APR 2021 // PREPARED BY RYAN CARRIER
ACKNOWLEDGMENTS
The author would like to thank Jacquie Hughes, and Mark Potkewitz, for their invaluable feedback on
earlier drafts of this report. Special thanks to Tristi Tanaka and the #ethicscurriculum team and the
#children's code team at ForHumanity fo helping to shape and develop these ideas.
PRINT AND ELECTRONIC DISTRIBUTION RIGHTS
© 2021 by ForHumanity. This work is licensed under a Attribution-NonCommercial-NoDerivatives 4.0
International license.
To view a copy of this license, visit: https://creativecommons.org/licenses/by-nc-nd/4.0/
Ryan Carrier, CFA
Ryan is Founder, Executive Director, and Chairman of the Board of Directors of ForHumanity.
Ryan@Forhumanity.center | https://forhumanity.center/
FORHUMANITY PAGE | 01
Introduction
A NECESSARY RESPONSE TO
THE SUSTAINABILITY OF
PROFIT THAT DOESN’T COME
AT THE EXPENSE OF HUMAN
AGENCY.
It is difficult to pinpoint when the trend started. It may be
linked to global concern about climate change — a
problem too existential to handle by any single person,
company, or even nation-state. Or maybe, well-publicized
incidents of racial injustice may have accelerated a sense of
a power and accountability imbalance. Or could it be
related to a growing unease about the impact of pervasive
technology on our individual and collective well-being
played its part. Whatever the root cause, source, or spark,
there is no doubting the gathering demand for
accountability, oversight, and governance of all institutions
and entities who have any kind of domain over our livesThe
goal is worthy — to protect humans and human agency.
However, the “soft” terms being applied, such as well-being,
ethics and good social order, present new challenges for
adherence, compliance, and enforcement.
In an attempt to control and contain rapid and encroaching
technological development, soft laws are being developed
at pace, to aid hard laws in attempting to protect and
sustain human agency. But we should not expect laws to
do all the heavy lifting: company Boardrooms and C-suite
officers - faced with critique and criticism about the impact
of their products and services on an unsuspecting public -
must do their share. At the time of writing, some
companies have dug in their heels while others are
Is beginning to search for viable long-term solutions for the
impact of algorithmic technology being ubiquitously
human agency advanced often at the expense of humans.
slipping away?
PAGE | 02 FORHUMANITY
Companies are deploying systems throughout every aspect of the company, affecting humans at
every turn. Defending humans and human agency in the midst of substantial societal upheaval, and
technological change requires a unique set of skills and those skills cannot be applied post hoc, but
must be applied throughout an algorithm’s lifecycle. A hallmark of recent technological change is the
increased ability of algorithms to nudge, guide, choose, interpret and decide “for us”, and this begs
the question: which elements ‘of us’ are being encoded? Many of us recognize when our shopping
preferences have been encoded into recommendation lists, but what about more subtle examples of
our encoded morality, priorities, well-being? And related to all of those crucial questions, where is
’our agency’ in the design, development and application of these systems?
As company designers and developers build more and more algorithmic systems designed to
replicate/replace portions of human decision-making or physical activity, they tread into areas of
well-being and moral frameworks for which they are ill-equipped. Worse still, they sometimes lack the
awareness of the ethical impact of their work that verges on recklessness. Design teams solving
problems, but also assessing the associated risks of their work is, at best, a conflict-of-interest and
certain to be filled with myriad cognitive biases which prevent a fair or reasonable evaluation of risk.
Failure to manage the risks from diminished human agency and failed application of moral
frameworks can have devastating consequences for people.
Examples of Soft Law
Recently, government and regulatory response have created a few protections to reclaim lost agency
— a legal one, the General Data Protection Regulation (GDPR), is filled with soft law descriptions like
the one found in article 22.1 (amongst others articles, the emphasis is added):
The data subject shall have the right not to be subject to a decision based solely on automated processing,
including profiling, which produces legal effects concerning him or her or similarly significantly affects him
or her.
The Rise of the Ethics Committee
FORHUMANITY PAGE | 03
HARD VERSUS
SOFT LAW
. Also, in the UK, the Age-Appropriate Design Code
(hereafter “Children’s Code”), in response to online
harms against children, became law. The Children’s
Code chooses to reinforce the United Nations
Convention on the Rights of the Child (UNCRC) for
those subject to the law. It borrows the definition of
Best Interests of the Child, which is overflowing in soft
law language:
The UNCRC incorporates provisions aimed at supporting
the child’s needs for safety, health, wellbeing, family
relationships, physical, psychological and emotional
development, identity, freedom of expression, privacy,
and agency to form their own views and have them
heard. Put simply, the best interests of the child are
whatever is best for that individual child.
In 2019, after repeated criticisms of a “profit-only”
raison d’etre, the Business Roundtable addressed the
role of corporations in society with this amendment to
the statement of purpose of a corporation:
the release of a new Statement on the Purpose of a
Corporation signed by 181 CEOs who commit to lead
their companies for the benefit of all stakeholders –
customers, employees, suppliers, communities, and
shareholders.
P A GFEO |
R H0U
4MANITY FORHUMANITY
Risks from Soft Law and Artificial Intelligence/Algorithm Risk
Collectively, these changes amount to strong support for the well-being of humans and human
agency. However, figuring out how to facilitate and implement these changes has eluded many
corporate leaders to-date, leaving the world with the impression of “hot air” rather than intentional
change. This paper aims to provide a specific action plan for CEOs who either must comply (when the
soft law is enacted) or desire to comply (when they choose to abide by statements like those of the
Business Roundtable). We hope that a roadmap and process of oversight will enable corporations to
“meet the promise” from the Business Roundtable.
We recognize the theme of this paper flies in the face of a recent news items from Google; ethicists
Timnit Gebru and Meg Mitchell were both let go which dealt a significant blow to the tech company’s
ability to meet the challenges of tackling the risks and soft law implications across their vast product
suite. Instead, Google continues to handle these risks, like most other corporations with the mistaken
belief that their business managers, legal team or designers and developers, without training, can
navigate questions of ethical choice and compliance with soft law. We expect that this belief will prove
costly — far greater than the cost of the wise development and deployment of an empowered,
integrated and well-considered ethics committee.
We believe that Google’s recent actions, including the failed introduction of an ethics panel, will be
looked back at as a watershed moment and highlighted as a key reason for the Rise of the Ethics
Committee. “The Rise” reflects the present state of the ethics committees and our opinion that they
currently remain mostly ineffectual. Today’s ethics committees are the equivalent of “window
dressing” -— not because of the people involved but rather a failed corporate commitment to
adherence and accountability with respect to their own moral frameworks or sometimes worse:
disregard for human agency. Called upon to draw up codes of conduct, codes of ethics, and recently,
codes of data ethics, these teams are cobbled together, usually with a strong concentration in legal
compliance. Some of the employees named in these committees may be well-intentioned and sincere,
but far from being ideal choices, they are far too often the least worst choices. For decades, whatever
these teams decide has been the definition of “ethical” for corporations. These unintegrated teams,
many of whom have little training in Ethics, Applied Ethics, or Philosophy, are treated as afterthoughts
when compared with the primary objective of maximizing shareholder value. Suffice it to say, most
ethics committees would not have to improve much to represent a dramatic “rise” in their materiality.
PAGE | 05 FORHUMANITY
Soft Law Explained
While we use the term “law” to refer to enacted statutes with clear terms and definitions, we consider
soft law to include things like regulatory guidance, commentaries from persuasive authorities like
regulators or government agencies, or guidelines published by official governmental sources. Below
is the most recent definition we could find, it is not current:
Soft law refers to rules that are neither strictly binding in nature nor completely lacking legal significance.
In the context of international law, soft law refers to guidelines, policy declarations or codes of conduct
which set standards of conduct. However, they are not directly enforceable. Hard law refers to binding laws.
GDPR and the Children’s Code in the EU and UK respectively are binding laws and are directly
enforceable. Therefore if we use the former definition and bring it current, we can manufacture a
more current definition for “soft” law.
Soft law refers to rules that are neither strictly binding in nature nor completely lacking legal significance.
In the context of international law, soft law refers to guidelines, policy declarations or codes of conduct
which set standards of conduct. However, it is not well known or adjudicated on how these guidelines,
policy declarations and codes of conduct will be enforced by the judicial system.
Our estimation is that the legal application of these “soft” law criteria will likely fall into the same
category as “reasonable, fiduciary or industry-standard”.Companies will need to have evidenced-
based documentation to support claims that they are in compliance with the law. In addition,
expertise will be required to satisfy compliance standards as qualified and accountable.
Unfortunately when it comes to the “best interests of the child” or determining the impact of profiling
on a data subject - a GDPR requirement - new skills are needed to meet these challenges and
robustly manage the risks arising from “soft law”.
The Rise of the Ethics Committee
P A GFEO |
R H0U
6MANITY FORHUMANITY
Risks associated with algorithms and soft law are present, but corporate awareness and management
of these risks appear absent. Current risk management solutions rely on strategies such as “hoping to
avoid scrutiny” and “embedding the estimated costs of legal liabilities” into the cost structure of the
product or service. Neither of these solutions offers long-term viability, especially when companies
facing a rising clamor for increased human agency, compliance, and regulation. We suspect that
companies will be forced to adopt risk management techniques rather than burying their heads in the
sand. Ambivalence or neglect for people’s rights and freedoms is not a sustainable business model
and the result will either be consumer backlash or heavy regulatory oversight.
The Rise Starts Now
So how can we explain why the “rise” will begin now? Risk is the answer. At their core, corporations
are profit-making ventures, and any risk to profit becomes a priority. As awareness of these risks
increases, business managers will seek the most qualified experts for managing that risk — in this
case, well-trained ethics officers. Companies are learning firsthand, from new laws, from multi-
stakeholder feedback, and from failure to mitigate significant and monstrous externalities tied to their
products that these risks are real and significant. Examples include: 1) excessive influence on
democratic institutions; 2) dangerous dabblings in constitutional and fundamental human rights; or,
3) more insidiously the embedded bias in our data and artificial intelligence models as highlighted by
director Shalini Kantayya and leading authority Joy Buolamwini in Coded Bias. The risks have become
mission-critical, so the sustainability of profits and shareholder value are at risk. Action will follow.
Governments and regulators have responded first. Now, company boards, CEOs, and risk managers
are beginning to seek remedies for soft law and algorithm risk. ForHumanity, in one example,
identified a mechanism to establish an Infrastructure of Trust for AI and Autonomous systems in the
form of a compliance-by-design solution: Independent Audit of AI Systems. The process has
established audit rules and adaptations of existing law that embed governance, accountability and
oversight across ethics, bias, privacy, trust and cybersecurity for AI and Autonomous Systems.
Compliance with government approved audit rules/criteria necessitate explainability, transparency
and disclosure combined with human agency in risk assessment and feedback loops tied to design
and development.
Ethical Choice - the responsibility of Trained Ethics Officers and the Ethics Committee
ForHumanity designed Independent Audit of AI Systems (IAAIS) to create and enforce accountability,
governance, transparency and oversight throughout the life-cycle of AI and autonomous systems from design
to decommissioning. A key element of this accountability and transparency is the systematic assignment of
instances of Ethical Choice for examination, resolution and disclosure by trained ethics officers wherever
those choices occur in the system life-cycle. Unlike most current, ungoverned processes, IAAIS removes the
temptation for designers and developers to make decisions that they are not trained for and that are filled
with their own conflicts-of-interest, such as failure to ignore sunk-costs and confirmation biases.Turning
instances of ethical choice over to trained ethics officers will reduce the risk of conflicted, concentrated and
myopic choices that often result in negative social impact and reductions in human agency.
PAGE | 07 FORHUMANITY
The audit criteria assigns instances of ethical choice to an ethics officer no matter where they are
found in the design, development, input and risk assessment phases of the AI and Autonomous
lifecycle. Again, normalizing the lexicon, we define Ethical Choice as:
Awareness of a set of options to be made In the context of automated and intelligent systems, using a set of
principles and rules concerning moral obligations and regards for the rights of humans and for nature,
which may be specified by a given profession or group. The result, outcome or judgment is made using a
shared moral framework. or set of moral principles based upon the entity’s Code of Ethics.
Examples include Min/Max benchmark (discussed in a forthcoming paper) setting and
explanations/disclosures for dataset bias remediation or working with designers and developers to
establish Key Performance Indicators to assess the stability of an algorithm in accordance with the
initial scope, nature, context and purpose. Our current working list of embedded instances of ethical
choice in the life-cycle of design and development for an algorithm include:
1) Proportionality Assessments
2) Necessity Assessments
3) Risk Analysis
4) Diverse Inputs and Multi Stakeholder feedback
5) Reliability tests
6) Validity tests
7) Explainability
Instances of Ethical Choice occur frequently in the algorithm life-cycle and corporate risk
management will benefit from the expertise of trained algorithm ethics officers. This list comes from
ForHumanity's GDPR Certification Scheme criteria and in no way reflects the comprehensive
instances of ethical choice embedded in the vast array of systems being deployed across the
economy.Therefore, it is reasonable and expected that this list/skillset will grow in breadth and depth
as IAAIS expands to cover all AIs and Autonomous Systems that impact humans.
The Rise of the Ethics Committee
P A GFEO |
R H0U
8MANITY FORHUMANITY
The aforementioned list is further augmented by soft law examples. Ethics officers need expertise
and training to fairly and responsibility consider and evaluate corporate responsibility, priorities and
compliance reacting to the expansion of soft law such as:
1) The health and well-being of the Child
2) Physical, psychological and emotional development.
3) Right to develop their own views and identity
These are but a few soft law examples related directly to the Children’s Code, however they are
representative of the latest evolution in human agency — the further identification and specification
of rights and freedoms drafted into law by legislatures around the world.
These examples pinpoint the expertise needed to manage the risks sourced and identified in the
audit criteria. Further examples, beyond the scope of an algorithm ethics officer but still in the
domain of the ethics committee include areas of societal or existential risk, like ESG (Environmental,
social and governance) responsibilities.
Facilitating Genuine Change - a roadmap for Accountability and Transparency
Somewhere between the inevitable consumer backlash and heavy regulatory oversight lies the middle
ground of robust protections for humans and a system of compliance, accountability and oversight
that will satisfy regulators. We believe that IAAIS, a replica of the system of independent, third-party
governance, accountability and oversight which has served the financial accounting and reporting
world successfully for 50 years, will supply the needed infrastructure of accountability and
transparency to address these risks.
Adoption of certification schemes and audit rules/criteria by regulators and governments creates the
opportunity for proactive compliance with laws and best practices to mitigate algorithm risk. Thus we
can facilitate genuine change by systematizing accountability. To accomplish any human-led change,
accountability is required.
Accountability is the deciding factor between those who talk, wish and hope for change and changemakers
who take action and see the act through to completion
PAGE | 09 FORHUMANITY
It is an unfortunate tenet of human existence that when we are watched or checked-in upon, we are
more likely to get the work done or comply with precise requirements. But clearly accountability is a
concept that Silicon Valley seems loathe to accept. Like proverbial foxes guarding hen houses, tech
companies hire their own ethics panels and require them to apply self-declared standards. To much
fanfare and little surprise, these companies find and declare themselves ethical to the highest
standards they have conjured.
Accountability is effective and meaningful when it is independent, such as abiding by a set of third-
party rules rather than making your own and inviting third-party audits to assure compliance rather
than grading your own homework assignments. When companies demonstrate humility to be
compliant with representative societal values, consensus-driven best practices, and application of the
law, then will they meet this higher standard of accountability. The ethics committee is the agent of
oversight, and management of the moral frameworks found in their own code of ethics. But instead
of empty words on a page, now, when faced with third-party independent audits, companies will have
to submit to compliance audits for societal assurance, and we all will benefit; even the company
being audited as their operations will have greater sustainability.
Accountability is further enhanced when the process requires and fosters greater transparency. Now,
not only is someone watching, and checking, but the whole world is able to see the work, to review
and critique crucial elements of disclosure and compliance. Transparency and disclosure create a
feedback loop. Once information is disclosed, critics can provide constructive critique into the
process and facilitate future improvements. Accountability combined with transparency creates a
virtuous circle of improvement and development.
A fair examination of the accumulation of risks from soft law and instances of ethical choice in our
AIs and Autonomous Systems combined with the aspirational appetite of corporations to move
beyond prioritizing shareholders at the expense of the humans reaches one conclusion -
empowering the ethics committee. Companies will make this choice, not for public relations or ethics
washing, but for genuine application of ethical choice, adherence to a moral framework publicly
declared in the code of ethics, and integrated into corporate risk management because it will result
in a greater likelihood of sustainable profit. Humans benefit with corporate entities that are more
attuned to human agency.
The Rise of the Ethics Committee
P A GFEO |
R H1U
0MANITY FORHUMANITY
At some point for sustainability - companies must figure out that they exist because people/humans
collectively allow them to exist and choose to use their services.They have rights because people grant them
limited rights in our socioeconomic system. These rights are not inherent and unalienable. If corporations
continue to negatively influence, alter, degrade or otherwise subvert human agency, those granted rights
will be rescinded. The risk of being unsustainable is manifest now.
Sustainability happens when companies accept and embrace The Rise of the Ethics Committee. The
alternative, for these entities, appears to be heavy regulatory oversight, and complete transparency at
the expense of intellectual property and trade secrets or, worse, antitrust actions which are founded
in the explicit recovery of human agency. The right choice seems so obvious — choose human
agency, choose humility, choose to empower people to think beyond profit and consider your own
declarations in the Business Roundtable to lead companies for the benefit of all stakeholders —
customers, employees, suppliers, communities, and shareholders. Elevating and integrating the Ethics
Committee is the key to keeping that promise.
ForHumanity
ForHumanity is a 501(c)(3) tax-exempt public charity formed to examine and analyze the downside
risks associated with the ubiquitous advance of AI and automation. To this end, we engage in risk
control and mitigation and deploy the lens and filter of Ethics, Bias, Privacy, Trust, and Cybersecurity
to ensure the optimal outcome…ForHumanity.
ForHumanity is an interdisciplinary group of dedicated expert volunteers, with over 300 contributors
and 32 Fellows, Its collective expertise spans the AI field, ranging from ethics to algorithmic risk and
to security. Our team is drawn from the academic, legal, policy, corporate, and public sectors of over
40 countries around the world. Our mission is to help create an ‘infrastructure of trust’ for all
autonomous systems that directly impact humans.
ForHumanity drafts comprehensive, pragmatic and implementable audit rules and standards for
autonomous systems in every corner of the economy. Our experts collaborate with industry
practitioners to ensure these audits achieve our mission of mitigating AI risk to humans. This system
of audit rules and standards - adapted to local jurisdictional laws and regulations - is called
Independent Audit of AI Systems (IAAIS).
FORHUMANITY.CENTER | RYAN@FORHUMANITY.CENTER