AI Does Not Judge: Epistemic Archaeology and Governance of Meaning in the Humanities and Social Sciences

The adoption of GenAI in research is reshaping the epistemic conditions of knowledge production. While LLMs are often perceived as neutral agents, this paper argues that the humanities and social sciences (HSS) serve as critical infrastructures for governing AI-mediated inquiry. Drawing on sociotechnical pragmatism, the article introduces epistemic archaeology: an interpretive orientation that treats AI outputs as stratified artifacts shaped by cultural and statistical histories. Rather than seeking validation in convergence, this approach interprets model divergence as signals of training data and epistemic constraints. By foregrounding interpretation, contextualization, and epistemic literacy, the work establishes the foundational coordinates for AI integration. This conceptual positioning preserves human responsibility and judgment as the primary locus of meaning in human–AI interaction. While serving as the theoretical base for future applications, the paper defines the ethical landscape for responsible academic research.