[COLLOQUIUM ONLY] Un-captive Listening: From Machine Translation to Machine Transduction in an Era of Expansive AI
**This event is open to those of the anthropology community ONLY**
According to Roderic Crooks (2019), arguments in favor of embedding datafied surveillance regimes in domains as diverse as healthcare and national security tend to espouse a representationalist theory of data: the notion that computational technologies and methods (including those grouped under the kitchen sink category of “AI”) can generate faithful representations of the world or whatever phenomenon a technology is meant to make sensible (487). This justificatory logic takes for granted that “data cannot speak for themselves” and instead “must be made to speak” (485), i.e., that there is always a gap between data and the things they supposedly represent that can be exploited to serve hegemonic ends or bent in the direction of counterhegemonic ones. Building from these provocations, this talk proposes that we reconsider initiatives geared toward seamless machine translation as instances of machine transduction enabled by what I call transductive labor. I draw from ethnographic fieldwork with researchers and human research subjects involved in constructing a class of computational technologies designed to evaluate and track people that pass through the American mental healthcare system based on the sounds of their voices alone. In conversation with linguistic anthropological and sound studies analyses of transduction, I utilize transductive labor to complicate techno-determinist accounts of the expansive inescapability of computational listening devices, revealing the technology development pipeline to be suffuse with small evasion of capture and recalcitrance that suggest otherwise. References: Crooks, Roderic. 2019. “Cat-and-Mouse Games: Dataveillance and Performativity in Urban Schools. Surveillance and Society 17 (3/4): 484-498.
