IEEE Technology and Society Magazine - December 2021 - 29

measure-will be whether it is implemented by technology
developers and respected by other stakeholders.
Adoption both supports and signals credibility.
The inability to draw key players into the process
can doom an effort. Even a substantial investment
in creating what seems like a wisely conceived initiative
can end up in the back of the filing cabinet
if those who make the technology do not adopt it or
those who benefit from the initiative do not accept
the credibility of the outcome.
Finally, distributive fairness of the output also
plays a role in an initiative's credibility. Much like
standards, soft law initiatives that take hold tend to
favor or disfavor specific structures, processes, or
behaviors. Like most forms of governance, they create
winners and losers [11], [32]. The distribution
of responsibilities and benefits deepens the importance
of credibility for the success of an initiative.
Even as we seek to mitigate the growing threat of bias
and value-laden decisions made algorithmically, we
need to recognize that the rules surrounding those
algorithms will have distributive effects as well, and
the fairness of that distribution will play an important
role in the perception of credibility [33], [34].
Research on credibility cannot provide a single
checklist or recipe. At the same time, the literature
shows a set of common factors that form the foundation
of credibility necessary for AI soft law: ongoing
relationships across participants, process, and output
with regulative content, adoption, and distributive
fairness in mind.
Centering credibility in soft law development
strategies
Credibility is so fundamental to soft law that it
needs to serve as the strategic " true north " for any
initiative aiming to have impact. Each initiative
will need to earn its own credibility and will also
be affected by the credibility of its ecosystem. This
impacts the entire process of formation, stakeholder
identification, value proposition, and situation
within the wider ecosystem of AI governance, ethics,
and regulatory initiatives.
In many cases, planning for credibility will mean
pulling broad representation across the ecosystem
directly into each part of the development process.
In others, that may mean deliberate dialogue with
specific stakeholders at key development stages.
The specific approach will depend on the initiative,
ecosystem, and actors-but it is critical to plan for
December 2021
late-stage credibility, adoption, and validation even in
the earliest steps of the development process.
As is often the case, it helps to start with the end
in mind. Soft law initiatives take a multistep process
of definition, development, and adoption. Credibility
and adoption are very hard to " bolt-on " to an
initiative in the late stages. Building on the research
identifying participants, process, and output as focal
points, strategic considerations across the soft law
development process follow.
Participants-If credibility is " trustworthiness and
believability in the eyes of other actors, " then the
measurement of a soft law initiative will rest on how
key actors view it. It is that simple and challenging at
the same time. Strategies to build credibility need to
start with identifying those actors. Whose behavior
is the initiative designed to guide or change? Who
needs to adopt, validate, or accept specific practices?
Who is intended to benefit and how? How does the
regulatory ecosystem that the initiative will be a part
of work? Are there specific governments or agencies
who need to be at the table or at least aware and
providing input? What mechanisms are available or
need to be created to establish communication and
build the necessary dialogue? Identifying stakeholders
and meeting their expectations over the course
of the initiative is the essence of building an effective
soft law initiative.
In particular, there is an increasing recognition
that participants in both the creation and governance
of AI will have to reflect far more diverse backgrounds
and views than the current state of the tech
industry [21]-[23]. Recruiting new voices to the
work will be important and, at the same time, doing
so will demand new thinking on the value proposition
for participation-a commitment that those
new voices will be heard and will play a role in outcomes
[22], [23]. In addition, a key advancement
that soft law for AI will need to make, over current
soft law and standards work in the tech industry, is a
much more deliberate focus on how to identify and
engage with a wider group of intended beneficiaries
and provide them with an effective platform to enable
meaningful participation.
Process-The initiative development process
needs to stand up as a credible one against generally
accepted terms and in the eyes of the stakeholder
community over the course of the initiative [27].
This will typically include well-developed processes
for planning goals and output, consensus building,
29
IEEE Technology and Society Magazine - December 2021

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