IEEE Technology and Society Magazine - September 2021 - 83

understanding that they are not simply disinterested.
They are reflexively strategic and should always
require public scrutiny.
Third, we need to create schemas of contestability
embedded in problem formation before the implementation
of algorithms. One way to do so and avoid
harm is by engaging with both domain experts and
the lived experience of those impacted [4].
At this level, we need to ask who is doing the contestation
and for what purpose. After we clarify the
subject of contestability, we have to establish the
mode under which it takes place. Is there a place
for spontaneous or unregulated practices of contestation?
If so, how much do they weigh in shaping
the initial problem formulation. Finally, what
purpose does contestability serve? In other words,
does contestability fulfill a purely adaptive and normalizing
function of various forms of resistance and/
or is focused on empowering historically vulnerable
beings? What tradeoffs are in play between practices
of normalization of contestability and the empowerment
of historically vulnerable stakeholders?
we cOnclude that problematization and problem
formulation emerge both as practices of critical
judgment in contingent situations and as historical
frameworks for analyzing how problems came to
be in relation to each other within specific domains
of inquiry. Precisely, problem formulation is: 1) a
practice where ML practitioners can engage in interpretive
frameworks in which they can ethically question
their modes of knowing in relation to their own
assumptions, values, and strategic reasoning and 2)
a lens through which critical technology scholars
can examine the historical transformation of past
instances of problem formulation against yet to be
invented standards of ethical reasoning.

 References
[1] H. Arendt, Lectures on Kant's Political Philosophy.
Chicago, IL, USA: Univ. Chicago Press, 1992.
[2] Aristotle, Metaphysics, and J. Barnes, The Complete
Works of Aristotle: The Revised Oxford Translation, vol.
2. 1984.
[3] R. Benjamin, Race After Technology: Abolitionist Tools
for the New Jim Code. Cambridge, U.K.: Polity, 2019.
[4] D. Martin, Jr., V. Prabhakaran, J. Kuhlberg, A. Smart,
and W. S. Isaac, " Participatory problem formulation
for fairer machine learning through community based
system dynamics, " 2020, arXiv:2005.07572. [Online].
Available: https://arxiv.org/abs/2005.07572
September 2021
[5] F.-F. Li. (Dec. 2020). Fei-Fei Li's Remark at AI Debate
2. [Online]. Available: https://www.youtube.com/
watch?v=XY1VTLRIsNo&t=59s
[6] S. Passi and S. Barocas, " Problem formulation
and fairness, " in Proc. Conf. Fairness,
Accountability, Transparency (FAT), 2019, pp.
39-48. [Online]. Available: https://dl.acm.org/
doi/10.1145/3287560.3287567
[7] R. Rosales and S. Charrington. (2018). Romer
Rosales Interview-Problem Formulation for
Machine Learning. The TWIML AI Podcast With Sam
Charrington.[Online]. Available: https://www.youtube.
com/watch?reload=9&v=1EnNo_LZMUc
[8] R. Thomas and D. Uminsky, " The problem with
metrics is a fundamental problem for {AI}, " CoRR, vol.
abs/2002.08512, 2020. [Online]. Available: https://arxiv.
org/abs/2002.08512
[9] M. Zuckerberg. (2005). CS50 Lecture by Mark
Zuckerberg-7 December 2005. [Online]. Available:
https://www.youtube.com/watch?v=xFFs9UgOAlE
Razvan Amironesei is a Visiting Researcher
with the Ethical AI Team, Google, Mountain View,
CA, USA, where he is currently working on the
techno-social formation of offensiveness and on
methodological questions related to the constitution
of machine learning datasets.
Emily Denton is a Senior Research Scientist with
the Ethical AI Team, Google, New York, NY, USA,
studying the norms, values, and work practices
that structure the development and use of machine
learning datasets.
Alex Hanna is a Sociologist and a Senior
Research Scientist with the Ethical AI Team, Google,
Mountain View, CA, USA. Her research centers
on origins of the training data which form the
informational infrastructure of AI and algorithmic
fairness frameworks, and the way these datasets
exacerbate racial, gender, and class inequality.
 Direct questions and comments about this article
to Razvan Amironesei; amironesei@gmail.com.
83

https://www.youtube.com/watch?v=XY1VTLRIsNo&t=59s https://www.youtube.com/watch?v=XY1VTLRIsNo&t=59s https://dl.acm.org/doi/10.1145/3287560.3287567 https://dl.acm.org/doi/10.1145/3287560.3287567 https://www.youtube.com/watch?reload=9&v=1EnNo_LZMUc https://www.youtube.com/watch?reload=9&v=1EnNo_LZMUc https://arxiv.org/abs/2002.08512 https://arxiv.org/abs/2002.08512 https://www.youtube.com/watch?v=xFFs9UgOAlE https://www.arxiv.org/abs/2005.07572

IEEE Technology and Society Magazine - September 2021

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