The evaporation of context: Epistemic agency and the problematic of 'understanding' in social work scholarship in the age of generative AI

Abstract

The growing use of generative artificial intelligence (AI) in academic writing raises questions about authorship, reflexivity, and epistemic responsibility in social work scholarship. This study examines whether generative AI can reproduce not only the structural features of scholarly discourse but also the reflexive and contextually embedded knowledge practices central to critical social work. A qualitative comparative case design was employed, comparing a human-authored article on multicultural social work (2010, Turkish context) with an AI-generated text produced under controlled conditions using identical references and thematic structure. Analysis focused on three dimensions: theoretical integration, reflexivity and epistemic positioning, and contextual embeddedness. While the AI-generated text demonstrated strong structural coherence and citation alignment, differences emerged in dialectical engagement, explicit normative positioning, and national contextual grounding. The findings suggest that generative AI can simulate disciplinary discourse but does not inherently enact accountable epistemic agency, underscoring the continuing importance of reflexive human authorship.

This study compares a Turkish social work article written by a human author in 2010 with a text generated by Google Gemini using the same references and thematic headings. Social work requires understanding people's lives within their social, cultural, and political settings. The study asks whether artificial intelligence (AI)-generated writing preserves this context and the ethical judgement involved in producing knowledge. The AI text organized information clearly, but gave less attention to local circumstances, competing theoretical arguments, and the author's ethical position. These findings come from one comparison and do not establish how all AI systems perform. AI can help organize and summarize information. Human authors remain responsible for interpreting evidence, questioning assumptions, and connecting knowledge to particular communities. Critical review and accountable human judgement are therefore essential when using AI in social work research and education.