IEEE Technology and Society Magazine - December 2021 - 57

Challenges of Artificial
Intelligence Review in a
Soft Law Environment
Sara R. Jordan
Future of Privacy Forum,
Washington, DC 20005 USA
 If artIfIcIal IntellIgence (aI) lives up to the
claims of journalists, futurists, and tech companies,
then AI stands to disrupt the landscape of human
cognition, social order, and political power. Within
a week's span, AI can be discussed in both positive
and utopian terms as well as in negative, even apocalyptic,
terms. On the positive side, AI is described as
an extraordinarily confident and capable F-16 pilot
[2], a panoptic gastroenterologist who misses no
part of a colonoscopy [3], the all-seeing solution to
mapping climate change affected regions [4], and a
smartphone application able to show your body with
12% less fat and give dieting advice to achieve that
goal [5]. Conversely, in a similar one-week period,
AI is described as a tool for the surveillance of a
minority population [6], a technique for identifying
the whereabouts of political protestors or dissidents
[7], a rogue system that delivers arbitrary scores
on high-stakes examinations [8], and a harbinger
of decline for reasoning, resilience, and emotional
intelligence [9]. Each of these assertions rests on the
interpretation of both basic and applied AI research
that may or may not be deployed into the production
environment of normal human lives.
Within the abundance of effervescent positive
reports or in the plethora of dire warnings, there is
a continued call to insert meaningful human control
over the development or deployment of AI. In this article,
I focus on one proposed mechanism: AI review
boards [10]. Review boards (variously referred to
as review committees) are generally designed to
apply ethical principles to specified research problems
to produce actionable recommendations for
Digital Object Identifier 10.1109/MTS.2021.3123743
Date of current version: 7 December 2021.
December 2021
improving the risk-benefit profile of a technology
[11]. Principles that inform review boards come
from a variety of sources and, although many have
been published in the past five years, this does not
necessarily mean that there is an exceptionally large
range of principles or mechanisms proposed as
instruments for control of AI [12]. Instead, there is
significant convergence of ideas across the AI landscape.
Major players include tech companies (e.g.,
Google, Apple, Microsoft, IBM, and Facebook),
nongovernmental organizations (e.g., Future of Life
Institute, IEEE, ACM, and the Partnership on AI), and
national and supra-national governments (e.g., Chinese
government, the European Union's High Level
Working Group on AI, and the U.K. House of Lords
Artificial Intelligence Committee) [13], [15].
The most curious feature of the AI review board
landscape is that few institutions that articulate principles
for AI propose review boards as a mechanism
for control and describe how they would operate
[10], [11]. In effect, the concept of a research review
board is borrowed into the AI setting without the
associated regulatory or even normative authority
that makes human subjects research review boards
an effective control mechanism. Research Ethics
Boards (REBs), Institutional Review Boards (IRBs),
or Research Ethics Committees (RECs) draw upon
established applied philosophical traditions backed
by a complex tangle of authorities whose power is
delegated through law crafted by democratically
elected legislators. For example, in the United States,
IRBs are governed by the " Common Rule " agreed
upon by 22 Federal Agencies [16]. The ambition
embedded in calls for AI review boards is that they
will win the support of law and legal authority. This is
0278-0097/21©2021 IEEE
57
IEEE Technology and Society Magazine - December 2021

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