Learn · Qualitative methods

The three questions that make qualitative research rigorous

Qualitative work is judged by three questions, not by a checklist: what counts as knowledge, who has the authority to interpret experience, and how interpretation is made credible. They are not history. They decide whether your study is rigorous, and large language models make them more important, not less.

After this guide you will be able to:

State the three questions your method has to answer.
Use theoretical and procedural validation to check your own work.
See why fluent AI output raises the stakes on all three.

The three questions

What counts as knowledge? A theory tested against reality, a meaning co-constructed with participants, or something else? Your answer shapes every later choice.
Who has the authority to interpret experience? The researcher, the participant, the community, or a shared arrangement between them?
How should interpretation be made credible? Not by proof, but by coherence, reflexivity, transparency, and fit between your claims and your material.

How to check your own answers

The Qualifying Qualitative Research Quality framework gives a practical way in. It asks you to make your methodological choices and assumptions explicit, and to consider your own position reflexively. Quality then rests on three checks:

Theoretical validation: does the theory you produced fit the social reality you studied?
Procedural validation: is the research process designed to improve that fit, rather than just assert it?
Transparent, co-constructed claims: are your knowledge claims built with participants and communicated openly, so a reader can trace them?

Read together, these speak straight to the three questions: what you treat as knowledge, whose perspective carries interpretive authority, and how you earn credibility.

Why LLMs raise the stakes

The arrival of large language models does not make these questions obsolete. It sharpens them. Scholars studying LLMs in qualitative research argue that existing quality frameworks are still both applicable and necessary.

A model can produce a fluent analysis, but it is opaque and it carries unknown assumptions absorbed from its training data. Without attention to epistemology, machine-generated patterns get mistaken for genuine understanding, and interpretive authority quietly shifts from participants and researchers to an algorithm. That is exactly what question two warns against. So ethical validation and your own positionality become more critical, not less. You have to ask how the AI system shapes what you see in the data, and how it shapes the way you represent other people's experience.

Expand the questions, do not retire them. The three questions still hold, and new pressures make them wider. Algorithmic bias, data ethics, the environmental cost of large models, and decolonising knowledge production all sit inside "what counts as knowledge, who interprets, and how we make it credible." The questions are old. The work they ask for is current.

Answer the three questions on the page, not just in your head.

A reviewer asks which paradigm your study sits in, and you realise your methods section never actually says. That is not a gap in your thinking. It is a gap between what you decided and what you wrote down, and it is easy to miss when the writing sounds rigorous. MethodVahti checks that your stated epistemic stance, your terminology, and your written claims line up from protocol to publication, and flags combinations that do not belong together. It surfaces the gap. It does not answer the three questions for you. That is yours to answer, and to own.

See MethodVahti

Related guides

See also deductive, inductive, abductive and a short history of qualitative research. More guides are on the Learn page.