Machine Learning, Sociogeny, and the Substance of Race

Abstract

This thesis examines the relationship between machine learning, sociogeny, and race. Machine learning is understood here as an epistemic process that organises data into discrete spatio- temporal classifications. Following Kara Keeling, the thesis defines spatio-temporal classification as a coordinate that makes the racial imaginary possible, while itself eluding representation. Lastly, sociogeny, as defined by Frantz Fanon, is a ‘sociodiagnostic’ phenomenon that emerges at the ontological limit of the episteme, resulting in the discursive exclusion of the lived reality of racialised individuals. The thesis argues that an immediate parallel can be drawn between the scientific production of knowledge, as established truths, and perceptions of otherness. While historically science and technology have worked to support racial classification, contemporary machine leaning algorithms have been shown, even unwittingly, to replicate existing race relations. The thesis contends that racial classification in machine learning is symptomatic of a wider logics of pathology, as well as enumeration — which, if taken as a spatio-temporal coordinate, arrests self-actualisation. By aligning machine learning with Fanon’s notion of sociogeny, and the spatio-temporal coordinate, the thesis challenges views of a causal relationship between classification, racial position and psychic fragmentation. Instead, the thesis argues for an alternative view of sociogeny as a non-linear process, always-already in excess of racial perception under the conditions of duress. Ultimately, the thesis investigates machine learning as a methodological space to re-articulate the spatio-temporal coordinate and, in turn, affirmative black psychic generation.