Using AI to diagnose conditions like autism is not simply a matter of automating the same kinds of diagnostics used by clinicians. The definition of the condition itself is always contested terrain. AI developers may believe their work is apolitical, but inevitably they become key players in a political struggle over how care is conceived and distributed.

Anti-surveillance makeup that imitates the contrasting stripes of dazzle camouflage is as perennially popular as it is impractical. It plays into a fantasy of an individualistic fashion statement serving as effectual tech resistance, while perpetuating the neoliberal insistence on “personal responsibility” as the response to every crisis.

The U.S. Department of Commerce is developing an approach to defining people’s trust in AI. But rather than define trustworthy systems with respect to their consequences or differential impacts, it focuses on how they are perceived by “users” (not whom they are used on) and does not bother to ask if the systems themselves are worth trusting.

Well-intentioned data science can still reinforce unjust ideas about whose bodies matter

It's not enough to make facial recognition illegal when its infrastructural legacy remains

he scientists preparing and reusing NIST’s data to train facial recognition models may simply be subject to these tendencies and these philosophies — comfortable with screaming, bloodied, and non-consenting research subjects because they are not people to them but abstract sources of abstract data. But there are signs that this isn’t the only thing that is going on.

A reformist approach to facial recognition — making the system more “inclusive” — will not really reduce harm for the people actually harmed. This control and normalization is part of the point of data science. It is required for data science’s logics to work. All reform-based approaches do is make violent systems more efficiently violent, under the guise of ethics and inclusion.