A descriptive comparison of results from human and AI-driven Interpretive Phenomenological Analysis

This study explores the concordance and discordance between human researchers and ChatGPT in conducting Interpretative Phenomenological Analysis (IPA) using an interview transcript with a refugee resettling in the United States. Findings revealed both alignment and discrepancies between ChatGPT and human researchers; while ChatGPT effectively identified themes and coding structures with interpretative depth and reflective engagement, it occasionally struggled to capture implicit meanings, particularly in narrative continuity and cultural-contextual analysis. These challenges were evident in its misinterpretation of past and present events, difficulty in recognizing nuanced associations, and tendency to adopt a deficit-based lens in certain analyses. Findings are discussed considering the emergent AI research literature, with attention to culturally responsive and anti-oppressive research practices with historically marginalized populations.