Enhancing qualitative inquiry: AI-assisted focus group data collection
Huan Chen, Ye Wang, Cheng ChangPurpose
The purpose of this study is to examine how AI-generated summarization, with and without human oversight, influences the quality of data collected in virtual focus groups (VFGs) and to explore the utility of AI-assisted thematic analysis in qualitative research. By comparing human-only moderation with AI-supported conditions, this study evaluates the impact of AI on participant engagement, thematic depth and analytical efficiency. It aims to advance understanding of human-AI collaboration in qualitative inquiry and assess the potential and limitations of integrating large language models into both data collection and analysis processes within socially complex research contexts.
Design/methodology/approach
This study employed a between-subjects experimental design to compare virtual focus groups (VFGs) across three conditions: human-moderated (control), AI-generated summaries without human oversight and AI-generated summaries with human oversight. All groups discussed virtual influencers using identical prompts. A human analyst conducted thematic analysis to assess content quality across conditions (RQ1). For RQ2, we used a three-step AI-assisted thematic analysis: human-generated seed themes, AI-powered semantic quote retrieval via MPNet embeddings and quantitative benchmarking using cosine similarity scores. A human analyst reviewed retrieved quotes for precision and coverage and reflected on the cognitive burden and value of AI-assisted analysis.
Findings
The study found that human-moderated virtual focus groups generated the most emotionally rich and contextually nuanced responses. AI-generated summaries with human oversight produced comparable thematic depth and coherence, suggesting the value of hybrid moderation. However, AI-only summaries resulted in slightly lower engagement and semantic richness. AI-assisted thematic analysis significantly reduced cognitive burden, improved quote organization and enhanced analytical efficiency, though human oversight remained essential for ensuring relevance and depth. Overall, the findings support responsible human–AI collaboration, highlighting the strengths of AI in streamlining analysis while underscoring the irreplaceable interpretive role of human researchers in qualitative inquiry.
Originality/value
This study offers a novel contribution by empirically comparing AI-assisted and human-only approaches in both qualitative data collection and analysis within virtual focus groups. It is among the first to assess the impact of AI-generated summaries with and without human oversight on discussion quality and to implement a hybrid AI–human thematic analysis using semantic similarity benchmarks. The study advances the emerging field of AI-mediated qualitative research by demonstrating how human–AI collaboration can enhance analytical efficiency while preserving interpretive depth, providing practical and ethical insights for integrating generative AI into qualitative inquiry.