How to Analyse Interviews: A Documentary Method of Interpretation

Interviews are commonplace in HCI. This paper presents a novel documentary method of interpretation that supports analysis of the topics contained within a collection of transcripts, topics that are endogenous to it and which elaborate participants collective reasoning about issues of relevance to research. We contrast endogenous topic analysis with established qualitative approaches, including content analysis, grounded theory, interpretative phenomenological analysis, and thematic analysis, to draw out the distinctive character of the documentary method of interpretation. Unlike established methods, the DMI does not require that the analyst be proficient in qualitative analysis, or have sound knowledge of underlying theories and methods. The DMI is a members method, not a social science method, that relies on mastery of natural language; a competence most people possess.

💡 Research Summary

The paper addresses the pervasive use of interviews in Human‑Computer Interaction (HCI) and critiques the dominant reliance on coding‑based qualitative methods such as content analysis, grounded theory, interpretative phenomenological analysis, and thematic analysis. The authors argue that coding is an inherently theoretical act: it imposes abstract categories onto raw language that do not exist in the data themselves, requiring analysts to possess substantial methodological expertise and to engage in reflexive theorising. To overcome these constraints, the authors introduce the Documentary Method of Interpretation (DMI), rooted in ethnomethodology, which treats interview transcripts as natural‑language artifacts produced by “members” who are assumed to be competent language users.

DMI discards the coding step entirely. Instead of fragmenting the transcript, the analyst reads the interview as a continuous social phenomenon, seeking the “orderliness” of practical action and reasoning that emerges organically. The method proceeds by (1) viewing the whole interview as a single document, (2) recognizing recurring patterns of reasoning, (3) labeling these patterns as “endogenous topics,” and (4) describing how each topic reflects collective reasoning relevant to the research domain. This process relies on pattern recognition rather than abstract categorisation, thereby eliminating the need for inter‑coder reliability measures, predefined coding frames, or extensive training in qualitative analysis.

The paper contrasts DMI with the four established methods. Content analysis originates from quantitative media studies and builds a coding frame through representative sampling, category definition, and reliability testing. Grounded theory uses theoretical sampling and constant comparative coding to generate a substantive theory, yet still depends on iterative coding and category formation. Interpretative phenomenological analysis and thematic analysis similarly code utterances, group them into themes, and interpret meaning through a researcher‑driven lens. All these approaches share a common abstraction step that the authors deem unnecessary for many HCI interview studies.

To illustrate DMI, the authors revisit a prior study on participants’ post‑experience reflections after a breaching experiment. By applying DMI, they directly identified two endogenous topics—“discomfort with social norm violation” and “reconstruction of self‑identity”—without constructing a coding schema or negotiating agreement among multiple coders. The authors claim that DMI preserves the full conversational context, reduces methodological overhead, and makes interview analysis accessible to practitioners who lack formal qualitative training.

Nevertheless, the authors acknowledge limitations. Because DMI relies on the analyst’s subjective pattern recognition, reproducibility and comparability across studies may be challenging. The absence of explicit codes hampers the creation of standardized data structures for meta‑analysis. To mitigate these issues, the authors suggest incorporating structured peer discussion and, where appropriate, a supplemental coding phase to triangulate findings.

In conclusion, the Documentary Method of Interpretation offers a novel, language‑competence‑based alternative to traditional coding‑heavy qualitative analysis in HCI. It promises lower entry barriers, richer contextual fidelity, and a focus on the collective reasoning embedded in interview discourse. Future work should develop concrete guidelines for transparency, validation, and integration with existing qualitative frameworks to strengthen the method’s reliability and broader adoption.