More new work! "Data Feminism for AI," with
@kanarinka.bsky.social, which reflects our current thinking about how and why data feminism remains deeply relevant for AI research: arxiv.org/abs/2405.01286
When we wrote DF, we were not yet required to contend with the near-total corporate capture of AI research. Now, examining power (DF principle #1) must also centrally involve examining capitalist systems and the economic power they accrue 2/
@meredithmeredith.bsky.social has been a such leader here, along with theorists like Cedric Robinson and Angela Davis who have given us enduring models of racial and gendered capitalism 3/
Challenging power (DF principle #2) also needs to be recalibrated for AI, since so much LLM/genAI work reproduces the unequal status quo. Here we look to Wendy Chun to leverage biased models to help change the future, and to @wonyoungso.bsky.social to develop models to repair the past 4/
Rethinking binaries and hierarchies takes (principle #3) takes us back to questions of gender and power. The recent Sam Altman / ScarJo fiasco had not yet happened when we wrote this but still, now, always, it's gender and power all the way down. And consent too (more later) 5/
Elevating emotion and embodiment (#4) in AI goes in two directions: listening to those with direct experience of AI harms, as in Joy Buolamwini's "evocative audit" of FRT; and recognizing how data/AI can be an "affect amplifier" (cf.
Helena Suarez Val) going from rage to action 6/
Embracing pluralism (#5) takes up the topic of participatory AI, feat. a feminicide classifier which @kanarinka.bsky.social and Harini Suresh co-designed with data activists, a study of a potential chatbot for pregnant people led by @mariaa.bsky.social, and the caution issued by Mona Sloane 7/