IEEE Technology and Society Magazine - March 2022 - 73

Efforts to address embedded values
issues
It is important to observe that there are promising
developments that may help mitigate some of the
problems just described. There is current research
aimed at addressing multiple issues related to data
bias and embedded values. For example, one line
of research aims to make the internal structure of
inscrutable computer models more understandable
to people [21]. Several research groups have developed
computers systems that can analyze " black-box "
computer algorithms and provide basic explanations
for automated decisions that are understandable to
people [15], [18], [21]. Other research programs are
focusing upon detecting bias in data sets, or attempting
to ensure that machine learning models produce
fairer and more proportionate results [1], [29]. Moreover,
note that it is not just computers systems that are
subject to embedded biases and values. Human decision
makers, such as judges, have long been subject
to biases in their judgments that have historically difficult
to detect. Machine-learning research offers promise
in detecting biases in judges and other human
decision makers as well [5].
Observe, however, that this line of inquiry is relatively
new and is currently in its formative stages
outside of law [24]. As such, relatively few of these
bias-mitigating developments appear to have made
their way into the technological systems that are
just now entering widespread use in the legal context.
However, the incorporation of some of these
research findings into legal technological systems
could, in the future, help to mitigate, or render
more transparent, some of these embedded value
problems. The creators of legal technological systems
should pay close attention to developments in
emerging area of bias-mitigating and values-aware
technical research.
technologIcal systems that use artificial intelligence
are increasingly being used in the application
of law. Such systems can contain values subtly
embedded in their technological design. This observation
becomes particularly important in the context
of law, given the significant issues at stake, including
loss of liberty, property, or rights. Legal technological
systems that employ artificial intelligence require
special care and awareness in development, as the
use of artificial intelligence can raise specific issues
of embedded values that may be impactful but hard
March 2022
to observe. New technological developments that
can address some of these issues, as well as vigilance
on the part of developers, may help mitigate certain
problems. However, as previously discussed, it is
important not to frame these topics solely as technical
issues to be solved. In many cases, the legal AI
issues merely reflect deeper underlying social and
structural issues that technical solutions cannot fully
address.

 References
[1] A. Agarwal et al., " A reductions approach to fair
classification, " in Proc. FAT ML, 2017, pp. 60-69.
[2] S. Barocas and A. Selbst, " Big data's disparate
impact, " California Law Rev., vol. 671, p. 104,
Sep. 2016.
[3] J. O. Berger and D. A. Berry, " Statistical analysis and
the illusion of objectivity, " Amer. Sci., vol. 76,
pp. 159-165, Apr. 1988.
[4] B. Joerges, " Do politics have artefacts? " Social Stud.
Sci., vol. 29, pp. 411-431, Jun. 1999.
[5] R. Vunikili et al., " Analysis of vocal implicit bias in
SCOTUS decisions through predictive modelling, "
Proc. Exp. Linguistics, pp. 121-124, Dec. 2018.
[6] D. Citron, " Technological due process, " Washington
Univ. Law Rev., vol. 85, no. 6, p. 1249, 2009.
[7] P. Domingos, " A few useful things to know about
machine learning, " Commun. ACM, vol. 55, pp. 78-87,
Oct. 2012.
[8] I. Goldberg and D. Wagner, " TAZ servers and the
rewebber network: Enabling anonymous publishing on
the world wide web, " First Monday, vol. 3, no. 4, 1998.
[Online]. Available: https://doi.org/10.5210/fm.v3i4.586
[9] W. Hartzog and F. Stutzman, " Obscurity by design, "
Washington Law Rev., vol. 88, p. 385, Jun. 2013.
[10] J. Kleinberg, S. Mullainathan, and M. Raghavan,
" Inherent trade-offs in the fair determination of risk
scores, " in Proc. 8th Innov. Theor. Comput. Sci. Conf.,
2017, pp. 43:1-43:23.
[11] M. Kosinski, D. Stillwell, and T. Graepel, " Private traits
and attributes are predictable from digital records of
human behavior, " Proc. Nat. Acad. Sci. USA, vol. 110,
no. 15, pp. 5802-5805, 2013.
[12] L. Lessig, Code 2.0. New York, NY, USA: Basic Books,
2006.
[13] N. Love and M. Genesereth, " Computational law, " in
Proc. 10th Int. Conf. Artif. Intell. Law, 2005,
pp. 205-209.
[14] J. Kroll et al., " Accountable algorithms, " Univ.
Pensylvania Law Rev., vol. 165, p. 633, Nov. 2016.
73
https://www.doi.org/10.5210/fm.v3i4.586
IEEE Technology and Society Magazine - March 2022

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - March 2022
