The role of generative AI in qualitative data analysis : opportunities and limitations in supporting dissertation supervision in academia

Abstract

Generative AI is rapidly reshaping dissertation supervision, particularly in qualitative research where large language models (LLMs) are increasingly used for tasks such as coding, summarisation, and thematic exploration. This chapter examines how LLMs can support ―but not replace― the interpretive work that underpins rigorous qualitative inquiry. Drawing on Cultural-Historical Activity Theory (CHAT) and connectivism as theoretical framework, we show how technical parameters such as tokenisation, context windows, temperature, and platform guardrails function as methodological variables that directly influence analytical outcomes. We identify key supervisory concerns including students’ over-reliance on automated analysis, the erosion of interpretive competencies, documentation and transparency challenges, and inequities in access to advanced AI tools. In response, the chapter advances two guiding commitments: reflexive integration, which positions AI outputs as provisional and subject to triangulation, and digital stewardship, through which supervisors model ethical, transparent, and methodologically coherent AI use. To operationalise these commitments, we introduce the SUPERVISE framework, offering practical strategies for hybrid human–AI workflows, responsible documentation, and equitable supervisory practice. The chapter argues that supervision must shift from transmission to co-navigation, preserving interpretive ownership while leveraging AI’s analytical affordances. In doing so, it aligns responsible AI use with the epistemic values and pedagogical aims of qualitative research.