Meeting standards: (Re)colonial and subversive potential of AI modification

Abstract

AI potential to recolonise language practices by reproducing existing marginalisations in novel ways has already instilled fears of a ‘contemporary dystopia’ (Miras et al., 2022) — a space of cultural and linguistic erasure. Accents represent a distinctive aspect of language practice associated with one’s sociocultural, and ethno-racial characteristics. They account for one’s social identity, status, and proficiency (De Klerk & Bosch, 1995). This makes practices of artificially modifying accents particularly concerning, since they play into the ‘zero’ accent ideology. As a result, any deviation from the norm is marked as abnormal or deficient, and in need of, artificial correction. Using AI accent generators, therefore, has the capacity to further aggravate power inequalities between the linguistically privileged and underprivileged, and to encourage changes in self-representation towards what is perceived as the normative Standard. Artificial modification of self to match a desired representation is not new, given the long-standing discussions on digital image alterations and their negative relationships to self-perceived attractiveness (Ozimek et al., 2023). This conceptual paper explores the (re)colonial and subversive linguistic potential of AI accent generators through the lens of the social tendency of individuals to strive to meet a given Standard. Using the notion of ‘technologies of the self ’ (Foucault, 1988), we draw a parallel between self-perceived attractiveness of bodies and accents, to explain how artificial modifications do not straightforwardly support diversities, but instead encourage ‘self-corrections’ in line with those standardized sets of features which seem to promise a ‘better’ socioeconomic and educational standing within neoliberal societies.