This case study asks how new epistemologies and methodologies for AI systems design could be developed based on critical insights from feminist and postcolonial science and technology studies.

Visual AI systems exhibit well documented racial and gender bias 123. Furthermore, the gendered, raced, classed and colonial aspects both of AI development and of the effects of AI design and deployment have been pointed out 45678, alongside calls for feminist, intersectional, postcolonial, decolonial and indigenous computing and AI 9101112. Proliferation of such calls signals the need not only for better understanding of how specific categorial and cultural biases appear in the AI pipeline and where, but also for development of epistemologies and AI design methodologies that are explicitly oriented towards social justice and intersectional diversity.

This case study thus examines how “minor histories” (i.e. less known or canonized historical narratives) of AI and critical intersectional epistemologies can provide a basis for new approaches to AI design and development, and expand the utility of the AI forensics toolkit beyond single individuals or institutions to include disenfranchised communities and social justice work. Technically, the case study reflects and builds on work that predicts criminality based on faces, and psychological studies suggesting we infer social class based on faces 13 which inform other decisions downstream. However, no deep learning papers have looked at social class biases in face recognition datasets.

References

  1. R. Benjamin, Race After Technology: Abolitionist Tools for the New Jim Code. Polity, 2019.
  2. J. Buolamwini and T. Gebru, Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification, In Proc. Proceedings of Machine Learning Research Conference on Fairness, Accountability, and Transparency, 2018
  3. O. Keyes, The Misgendering Machines: Trans/HCI Implications of Automatic Gender Recognition, In Proc. Proceedings of the ACM on Human-Computer Interaction, 2018, vol. 2. doi:10.1145/3274357
  4. C. Draude, Computing Bodies: Gender Codes and Anthropomorphic Design at the Human-Computer Interface. Springer VS Research, 2017. [Online]. Available: https://doi.org/10.1007/978-3-658-18660-9
  5. J. Zou and L. Schiebinger, AI Can Be Sexist and Racist - It’s Time to Make It Fair, Nature, vol. 559, no. 7714, p. 324–326, 2018. doi:10.1038/d41586-018-05707-8
  6. J. Thatcher, D. O’Sullivan, and D. Mahmoudi, Data Colonialism through Accumulation by Dispossession: New Metaphors for Daily Data, Environment and Planning D: Society and Space, vol. 34, no. 6, p. 990–1006, 2016. doi:10.1177/0263775816633195
  7. M. Kwet, Digital Colonialism: US Empire and the New Imperialism in the Global South, Race & Class, vol. 60, no. 4, p. 3–26, 2019. doi:10.1177/0306396818823172
  8. C. O’Neil, Weapons of math destruction: How big data increases inequality and threatens democracy. B/D/W/Y Broadway Books, 2017.
  9. N. Kumar and N. Karusala, Intersectional Computing, Interactions, vol. XXVI, no. 2, p. 50, 2019. doi:10.1145/3305360
  10. A. Schlesinger, W. Edwards, and R. Grinter, Intersectional HCI: Engaging Identity through Gender, Race, and Class, In Proc. Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems - CHI '17, 2017, p. 5412–5427. doi:10.1145/3025453.3025766
  11. S. Mohamed, M. Png, and W. Isaac, Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence, Philosophy & Technology, vol. 33, no. 4, pp. 659–684, 2020. doi:10.1007/s13347-020-00405-8
  12. A. Abdilla, N. Arista, K. Baker, S. Benesiinaabandan, M. Brown, M. Cheung, M. Coleman, A. Cordes, J. Davison, K. Duncan, S. Garzon, D. Harrell, P. Jones, K. Kealiikanakaoleohaililani, M. Kelleher, S. Kite, O. Lagon, J. Leigh, M. Levesque, J. Lewis, K. Mahelona, C. Moses, I. Nahuewai, K. Noe, D. Olson, ‘. Parker Jones, C. Running Wolf, M. Running Wolf, M. Silva, S. Fragnito, and H. Whaanga, Indigenous Protocol and Artificial Intelligence Position Paper, 2020. doi:10.11573/SPECTRUM.LIBRARY.CONCORDIA.CA.00986506
  13. R. Bjornsdottir and N. Rule, The Visibility of Social Class from Facial Cues, Journal of Personality and Social Psychology, vol. 113, no. 4, 2017. doi:10.1037/pspa0000091