DeTAILS: Deep Thematic Analysis with Iterative LLM Support

TL;DR

DeTAILS is a toolkit that integrates large language models into the thematic analysis process, making qualitative research more scalable, efficient, and interactive while maintaining researcher control.

Contribution

It introduces an interactive workflow combining LLM assistance with established thematic analysis methods, supported by empirical evaluation with qualitative researchers.

Findings

LLM support aligns well with researcher refinements

Reduces analysis workload

Enhances researcher trust and engagement

Abstract

Thematic analysis is widely used in qualitative research but can be difficult to scale because of its iterative, interpretive demands. We introduce DeTAILS, a toolkit that integrates large language model (LLM) assistance into a workflow inspired by Braun and Clarke's thematic analysis framework. DeTAILS supports researchers in generating and refining codes, reviewing clusters, and synthesizing themes through interactive feedback loops designed to preserve analytic agency. We evaluated the system with 18 qualitative researchers analyzing Reddit data. Quantitative results showed strong alignment between LLM-supported outputs and participants' refinements, alongside reduced workload and high perceived usefulness. Qualitatively, participants reported that DeTAILS accelerated analysis, prompted reflexive engagement with AI outputs, and fostered trust through transparency and control. We contribute: (1) an interactive human-LLM workflow for large-scale qualitative analysis, (2) empirical evidence of its feasibility and researcher experience, and (3) design implications for trustworthy AI-assisted qualitative research.