UC Berkeley

What is it like to live in a world shaped by the axiom that there’s not enough care for all who need it—one in which artificially intelligent caregivers become, accordingly, not only thinkable, but necessary? This dissertation provides an ethnographic account of the turn to artificial intelligence in mental health care via a Silicon Valley startup (“The Startup”) that creates therapeutic chatbots. These bots offer on-demand, interactive emotional support to their human users through text (and in some instances, voice) messages. Though the story of AI in most medical applications is one of outpacing human ability while operating under direct (human) oversight, here the opposite is true: none of my interlocutors regard AI as superior in any way to a human therapist; and while my fieldsite’s team of psychologists monitor aspects of these exchanges, to oversee each interaction would nullify the very justifications that summoned these bots into health care—namely, their rapid scalability, low overhead costs, and 24/7 on-demand, discreet availability. Yet for all that AI isn’t a superior caregiver, it is nevertheless a caregiver— an identity that I explore by recognizing chatbots, startup workers, and end-users alike as ethnographic subjects. This dissertation examines the relationships that arise through and with AI care bots: I argue that the ontological ambiguity of human-bot bonds is not a threat to good care but the means of it; I examine the collaboration between AI and human care workers as one of colleagues who share a job requirement of being paradoxically expendable-yet-essential; and finally, I show how AI caregivers afford a glimpse of suicide prevention care unpaired from the ableism of “cure,” even as they reveal how surveillance and neglect in mental health care become mutually reinforcing. Across this dissertation, I aim to recenter disability in the theorization of human-technology relationships.