ReportOctober 11 2023

Blooming in Muddy Waters: DEI at AI Ethics Conferences

Emnet Tafesse
with Meg Young,
Ranjit Singh, and
Jacob Metcalf

Generally understood as an extension of the computer science domain, the field of AI ethics has historically struggled with gender and racial diversity. And even as BIPOC (Black, Indigenous, and people of color) scholars have substantially set the field’s research agenda, they remain underrepresented among participants in its conference spaces. While the commitment of AI ethics spaces to DEI issues have contributed to recognizable progress in the field, thornier and culturally ingrained barriers to equity remain. 

The result of a year-long study by Research Analyst Emnet Tafesse, Blooming in Muddy Waters: DEI at AI Ethics Conferences presents the experiences of BIPOC attendees at three major AI ethics conferences as they relate to DEI efforts — that is, the policies and practices each conference used to promote diversity, equity, and inclusion for attendees, organizers, and presenters. The report focuses on AI ethics conferences at a key moment in their evolution, as they shift from being about matters of statistical fairness in the development of data and technology, to playing an active role in fostering these values where knowledge production happens.  

With coauthors Meg Young, Ranjit Singh, and Jacob Metcalf, Tafesse argues that the current DEI efforts of AI ethics conferences must go further to act on the same social forces that precipitated the need for the field. To address deeper challenges like tokenism and misperceptions of BIPOC colleagues’ experiences, AI ethics conference spaces must commit to including and highlighting a wider range of perspectives, including with funding, programming, and organizing. The report offers pragmatic recommendations for conference organizers, meant to fuel changes that will better acknowledge each person’s individual expertise, experiences, and contributions, and support the field to recognize and further act on AI ethics problems at their source.

This work was supported by the inaugural DEI Scholars grant from the ACM FAccT Conference.

Lead Author and Co-authors

Connected Track