AIsplaining

Abstract

The growing implementation of Generative AI (GenAI) in education has implications on the representation of knowledge and identity across languages. In a context where content biases have been reported in AI-generated content, it becomes relevant to interrogate the ways in which AI technologies represent different linguistic identities. This article conducts a systematic analysis of AI-generated content to identify the potential discursive strategies that can contribute to the perpetuation of existing sociolinguistic hierarchies. Data for this study consist of a set of GenAI explanations of assorted linguistic identities comprising dominant and non-dominant languages. The method combines specialization codes from the sociology of knowledge with discourse analysis. Specialization codes are composed of two axes with a differing degree of emphasis (+/−) on epistemic and social relations (ER/SR). This tool is useful for understanding explanations because it focuses on what sort of information is considered legitimate knowledge and what kinds of knowers are considered valid. The analysis of epistemic and social relations reveals a sociolinguistic hierarchy articulated across three definitory aspects of identity: relationship to the world, structuring across time and space, and possibilities for the future.