IEEE Technology and Society Magazine - December 2021 - 18

Soft Law Governance of Artificial Intelligence
exercise that happens on the side, but should be
core to determining the direction that a soft law
takes when dealing with the ethical implications
of AI [24], [25]. The most critical aspect for successfully
involving citizens is ensuring that their
participation will actively shape the governance
and giving visible evidence of how. Engaging citizens
cannot be an afterthought, lip service, or a
tickbox exercise.
One of the core purposes of AI soft laws focused
on ethics are to provide the guardrails or compass to
help navigate the different interests and values inherent
in technology development. The help of citizens
can improve this type of ethical governance, identify
more unintended consequences in advance, and
improve its relevance.
Trusted environments for defining the
public interest
A significant barrier to the effectiveness of defining
the public interest with all stakeholders can be
the lack of trust and respect between various stakeholders.
This can undermine the effectiveness of the
dialogues needed for sharing knowledge and exploring
complex ethical challenges such as those posed
by AI-particularly where there may be clashes of
values or incentives for different groups.
As a Brookings Institute report on AI and public
opinion shows, " the public does not agree on the
proper regulation of AI applications. Indeed, population-level
support of an AI application may belie
opposition by some subpopulations. " As an illustration,
56% of American adults trust law enforcement
using facial recognition technology, according to a
survey by the Pew Research Center. Nevertheless,
support is lower among those aged 18-29, Black
Americans, and those who identify as Democrats
or lean Democrats [26]. The experiences of these
groups may largely dictate their perceptions. For certain
communities, lack of trust in law enforcement,
rooted in historic lack of trustworthiness, underpins
their view. It also raises clear questions about the
effectiveness and appropriateness of the technology
in use.
While useful for identifying where to start in building
soft law, trust metrics focused on broad " market
research style " approaches are limited. Blunt metrics
cannot shape governance without further context
and direct engagement with those who will most
likely be impacted by a specific technology. The
18
reason for this is that blunt metrics can give a distorted
picture of the reality behind the opinions represented
and embed biases against marginalized or
hard to reach groups. They may be under-weighting
groups who would be disproportionately affected by
a technology because their views are compared to
a wider populations and general statistics to gauge
public trust or concern. These metrics should be
treated with extreme caution if they are the only data
informing AI design or governance.
It is important to incorporate greater nuance into
the core purpose of trust research and processes to
understand and evaluate the full breadth and depth
of issues and concerns. Naturally, there will be many
disagreements and clashes of values and incentives
in these topics. How these clashes are approached
and the methodologies to either resolve or reflect
these in a soft law will be fundamental.
This concept of a " trusted environment " [27]
explicitly designed with these issues in mind is at
the heart of effective multistakeholder cocreation.
A trusted environment has two distinctions: a relentless
focus on the public interest and trustworthy processes
derived from the trust drivers.
The TIGTech initiative has identified important
considerations and building blocks for earning trust
in the process of collaborating with stakeholders
that can be used in defining the public interest of
AI innovations with citizens and other stakeholders.
The building blocks are as follows.
Trust others first
Taking a proactive step to trust first, with the hope
and belief-though no guarantee-that you will be
trusted back, significantly increases the likelihood of
being trusted [28]-[30]. An approach which starts
with respect to, and an inclination to, trust is more
likely to generate trusted behavior in return.
The convenors of a process to define the public
interest for an AI soft law should make a conscious
commitment to trust participants to have good
intent, be cooperative, and participate in good faith.
Seek out diversity of perspective, opinion, and
thought process-and be respectful of that
diversity
Distrust of governance processes and outcomes is
often a result of alternative views and dissenting perspectives
being ignored or silenced [31]. This reluctance
to take seriously or incorporate the concerns
IEEE Technology and Society Magazine
IEEE Technology and Society Magazine - December 2021

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2021
