Taxonomy of epistemic violence risks in ai systems

This paper proposes a taxonomy of the risks of epistemic violence in artificial intelligence (AI) systems and addresses how these systems can perpetuate disparities in knowledge access. Using critical social epistemology and algorithmic ethics as a foundation, six main categories were identified: epistemic invisibilization, conceptual appropriation, algorithmic silencing, hermeneutic exclusion, asymmetry of representation in data, and naturalization of historical biases. This study sheds light on how these phenomena can perpetuate the marginalization of social groups, distort minority knowledge, and automate discrimination. The proposed taxonomy serves as an analytical tool to guide ethical impact assessments and the development of guidelines for the responsible creation of AI tools. The study emphasizes the importance of critical approaches to ensure technology contributes to an inclusive, equitable digital environment.