IEEE Technology and Society Magazine - December 2021 - 87

The Two Faces of Facial
Recognition Technology
Neil A. Chilson and Taylor D. Barkley
Charles Koch Institute, Arlington, VA 22201 USA
 The face is the most public part of human
bodies. We usually recognize or identify someone
by their face. Our brains are wired to quickly spot
faces-even in arrangements of inanimate objects-
and to instantly identify the emotions they convey.
Thanks to the increased ubiquity of digital cameras
and advances in computer vision algorithms,
computers can now identify, recognize, and interpret
images of faces. This facial recognition technology
(FRT) is a powerful tool that has already proved
its utility in commercial uses. And it has provoked
concerns, including when used by governments,
and especially when used by law enforcement.
When considering how to govern new technologies
such as FRT, there are two categories of alternatives.
The traditional approach is " hard " law. Hard
law is legislation or common law decisions established
and enforced by government. For example,
concerns about the popularity of portable cameras
led U.S. courts to develop common law privacy torts
such as intrusion into seclusion and public disclosure
of private facts [1], [2]. The development of
the telephone and other modern communications
technology prompted state and federal governments
to adopt wiretapping and eavesdropping laws and
drove courts to develop new applications of Fourth
Amendment law [3], [4].
The other broad category is " soft law, " which Arizona
State University law professor Gary Marchant
and his colleagues describe as frameworks that " set
forth substantive expectations but are not directly
enforceable by government and include approaches
such as professional guidelines, private standards,
codes of conduct, and best practices " [5].
Soft law has certain characteristics that separate
it from hard law. It is more agile, flexible, and easier
to adopt because it avoids the procedural and legal
requirements of hard law. Questions of legal authority,
geographic jurisdiction, and conflict of mandatory
laws are not issues for soft law, which can
span jurisdictions and offer multiple overlapping
solutions. In addition, soft law generally encourages
collaborations between stakeholders rather than
adversarial proceedings that typify hard law [6]-[8].
These characteristics make soft law a good candidate
for governance in certain circumstances. Where
an area is rapidly evolving; where the benefits, risks,
and trajectories of change are inherently uncertain,
and where the technology has an extremely wide
scope of applications across many industries-in
these circumstances, soft law is far preferable to
hard law, which is either impossible to adopt or rapidly
outdated [8].
Of course, as Marchant and others note, the very
strengths of soft law in some circumstances are disadvantages
in others. Soft law avoids procedural
requirements at the cost of transparency and inclusiveness.
It is frequently nonbinding, which can
mean uncertainty and a lack of redress for those
affected by the governed tech. It can also lack effectiveness
and credibility with the public. In situations
where there is a narrower application of a technology,
the benefits and risks are easier to assess, questions
of legal authority and geographic jurisdiction
are less important, and the outcomes of potential
abuse are known to be serious and likely, soft law
may not be the right tool [6], [8].
Digital Object Identifier 10.1109/MTS.2021.3123752
Date of current version: 7 December 2021.
December 2021
We argue in this article that soft law and hard
law both have key roles when it comes to governing
FRT. Building on the soft law/hard law framework
described by Marchant, our methodology applies
a descriptive benefit-cost analysis to compare the
0278-0097/21©2021 IEEE
87
IEEE Technology and Society Magazine - December 2021

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