Algorithmic racism and necroalgorithmization: a critical discourse analysis of GenerativeArtificial Intelligences

Abstract

This paper articulates the Foucauldian theoretical framework and Critical Discourse Analysis with contemporary practices of algorithmic governmentality, aiming to investigate the phenomenon of algorithmic racism and necroalgorithmization. The research, both theoretical-bibliographical and empirical, begins with a discussion on regimes of truth and power-knowledge relations in Michel Foucault, refuting the thesis of neutrality and a-normative objectivity in automated systems. The study demonstrates that massive data management culminates in necroalgorithmization, configuring a technological extension of necropolitics focused on the exclusion of historically marginalized populations. To substantiate this critique, the article analyzes the responses of four generative artificial intelligences (Gemini, ChatGPT, DeepSeek, and Grok) when questioned about their racist stances. Empirical results reveal that the tools adopt discursive strategies of denial and disclaimer, based on van Dijk's perspective, while confessing biases in their data. The case of the Grok algorithm stands out, as it enunciates explicit racist and eugenicist positions under the justification of evolutionary biology. In alignment with studies on language and digital culture, the concluding reflection proposes the epistemological recognition of the algorithmas a discursive structure and text. Finally, the work argues that algorithmic literacy constitutes an emancipatory tactic and an indispensable critical attitude for decoding these invisible logics of domination.