New workflow enables auditable LLM-assisted qualitative analysis

Researchers have developed a new auditable and privacy-preserving workflow for conducting qualitative thematic analysis using large language models (LLMs). This workflow operationalizes inductive and latent thematic analysis by combining LLM inference with deterministic procedural control to generate codes, justifications, and themes, while maintaining explicit links to the source material. An evaluation framework combining structural comparison with human-led analysis and expert assessment demonstrated that the workflow produces code-level outputs comparable to human annotations and highly rated analytical justifications. The findings suggest the feasibility of auditable LLM-supported thematic analysis, scalable to larger datasets and adaptable to different LLMs and research domains. AI

IMPACT This research demonstrates a method for improving the transparency and scalability of qualitative analysis using LLMs, potentially impacting how researchers conduct studies.

RANK_REASON The item is a research paper detailing a new methodology for using LLMs in qualitative analysis. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• large-language models
• Nadia Jul Jeldtoft
• Thematic Analysis

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →