PREDICTIONS OF THE SELF: AI AND THE POLITICAL ECONOMY OF SUBJECTIVATION

Resumen

The recent widespread availability of Artificial Intelligence (AI) technology and the extensive records of human activities and behaviour in digital format present serious challenges related to how individuals construct their own identities and social relations. AI systems datafy our body and our sense of self, producing a new cartography of biopower (Foucault, 1982) and a new form of the political economy of subjectivation (Langlois & Elmer, 2019) that treats individuals as objects from which raw material is extracted to produce predictive models that act as our data doubles (Haggerty & Ericson, 2000). Issues such as algorithmic social biases (Bolukbasi et al., 2016), the idealized and pragmatic economic uses of AI (Srnicek, 2017), and the consequent reproduction of already existing power structures by predictive models (Crawford, 2021) have been problematized in the literature. This paper asks what kinds of data and labour mobilization occur in and around the production of predictive models: What political economy and socio-technical conditions are involved in the production of AI? How do these conditions produce predictive models that shape our sense of self and identity? Focusing on Kaggle, a platform for crowdsourcing AI development, I use digital methods and a software studies approach to examine the practices of the data science community on three high-profile machine learning projects and conclude by arguing that machine learning has been thought of and developed as a prediction of the self in order to prescribe individual behaviour to fulfill specific economic conditions.