AI4Qual: A Comprehensive Field Guide to LLM-Supported Qualitative Research (Tutorial)

Overview

AI4Qual presents a structured tutorial synthesizing empirical findings on large language model integration within qualitative research workflows. The half-day instructional program operationalizes LLM applications across two primary domains: semi-structured interview augmentation and qualitative data coding and thematic analysis. The tutorial derives its pedagogical framework from prior empirical work examining LLM performance in generating contextually appropriate follow-up questions and supporting iterative code development. Delivery includes theoretical grounding, practical implementation guidance, and transferable computational tools applicable to diverse research contexts.

Methods and approach

The tutorial integrates two sequential instructional modules. Part I addresses LLM-supported semi-structured interviewing through examination of design principles for AI-generated follow-up questions, including role assignment strategies, engagement pattern optimization, user perception research, and protocol-aligned integration of LLM prompts during live interview contexts while maintaining informed consent and ethical compliance. Part II operationalizes LLM-supported coding and analysis, translating prior empirical findings on ChatGPT and comparable systems into practice: transitioning from open coding toward categorical and thematic organization, establishing conceptual frameworks for interpreting human-LLM alignment metrics and inter-rater reliability coefficients, and implementing retrieval-augmented generation (RAG) infrastructure linking coded segments to source excerpts. Participants receive instructional materials, exemplar datasets, and a minimal RAG toolkit designed for adaptation to independent research corpora.

Key Findings

The tutorial produces actionable design principles and implementation pathways for two distinct LLM-supported qualitative research workflows. Semi-structured interview augmentation yields empirically grounded guidance on follow-up question generation, engagement optimization, and protocol-compliant LLM integration. Qualitative coding and analysis modules translate prior inter-rater reliability and human-LLM alignment findings into lightweight computational infrastructure, specifically RAG-backed evidence linking that connects thematic codes to supporting source material. Deliverables include presentation materials derived from prior empirical publications, exemplar datasets demonstrating workflow implementation, and a minimal toolkit for evidence path construction adaptable to researcher datasets.

Implications

The tutorial establishes accessible pathways for qualitative researchers to incorporate LLM technologies within established research protocols and ethical frameworks. By grounding instruction in prior empirical work examining LLM performance and human-AI alignment in qualitative contexts, the materials provide evidence-based guidance applicable across disciplinary domains. The RAG-backed infrastructure offers computational support for the iterative interpretation and documentation requirements endemic to qualitative analysis without requiring substantial technical expertise from end users.