Qualitative data analysis (QDA)
The systematic interpretation of non-numerical data — interviews, field notes, documents, images — to identify patterns, themes, and meanings. AI tools can assist with coding and thematic analysis but require careful human oversight.
Qualitative data analysis (QDA) involves interpreting non-numerical data — transcribed interviews, observation notes, documents, social media posts, images, video — to understand meanings, experiences, and processes that numerical data can’t capture. It’s central to ethnography, grounded theory, phenomenology, discourse analysis, and many mixed-methods research designs.
Core activities in QDA:
- Coding: assigning labels (codes) to segments of text that share a concept or theme
- Thematic analysis: organizing codes into higher-level themes that address the research question
- Memoing: recording interpretive notes and emerging insights throughout analysis
- Member checking and inter-rater reliability: validating interpretations with participants or co-coders
Where AI can help:
- Transcription: Whisper and similar ASR tools transcribe interview audio reliably and cheaply, removing a major time bottleneck
- Deductive coding assistance: AI tools (in ATLAS.ti, NVivo, or via direct LLM prompting) can apply a predefined codebook to large volumes of text, flagging segments for researcher review rather than replacing the coder
- Thematic suggestion: LLMs can suggest candidate themes from a corpus, which researchers can evaluate and refine
- Named entity recognition: identifying mentions of people, places, organizations, or events across a large document set
Critical limitations and ethical requirements:
AI-assisted QDA requires explicit transparency in methods reporting — “AI was used to assist initial coding; all codes were reviewed and finalized by the researcher” is appropriate framing. Using AI as a substitute for deep interpretive engagement undermines the epistemological foundations of qualitative research. Participant data processed through external AI tools raises consent and confidentiality issues that IRB protocols may not yet address.
Related terms: Topic Modeling, Named Entity Recognition
Related guide: Social Science