Higher education has spent considerable energy debating whether students use artificial intelligence to write papers. I believe an equally consequential question deserves attention: What happens when AI helps produce the findings those papers report? In qualitative research, a polished explanation can sound authoritative before anyone has established whether it represents participants’ experiences. If universities teach students to generate themes without teaching them to defend those themes, they risk confusing production with education.
In my forthcoming textbook, Qualitative Research with Artificial Intelligence: Responsible Methods for Research Design, Analysis, and Teaching, scheduled for release by Routledge in February 2027, I approach this challenge through the lens of human responsibility. AI can support transcription, organization, comparison, and exploratory analysis. Researchers must still explain how their interpretations connect to evidence. That distinction should shape how higher education teaches research, evaluates scholarship, and uses student feedback.
An example from an earlier chapter illustrates the stakes. An AI model might group expressions of fatigue, frustration, and monotony under “negative sentiment.” The researcher must then consider whether those expressions reflect burnout, structural inequity, or cultural forms of resilience. These possibilities carry different implications, and the software’s category cannot settle the question. A convenient label is the beginning of inquiry, not its conclusion.
For educational leaders, that distinction has practical consequences. A department that interprets student frustration primarily as an attitude problem might prescribe motivation workshops when the underlying issue warrants institutional investigation. Before acting, leaders need evidence that supports the explanation they have chosen. Faster categorization offers little value if it accelerates a mistaken response.
The opening chapter introduces a composite doctoral researcher who studies engagement in a digitally mediated course. She initially reads interviews as accounts of motivation and belonging, then recognizes how reminders, chatbot prompts, and analytics dashboards shape students’ participation and perceptions. The lesson extends beyond research methods: institutions must examine how their systems help produce the experiences they later measure. Student engagement cannot be fully understood by separating learners from the environments that universities create.
The teaching response should be concrete. In a later discussion of teaching, I describe sequencing AI assistance after students have closely read and initially coded material themselves, documenting their reasoning before comparing it with machine-generated suggestions. Faculty can ask where interpretations converge, where they diverge, and what the original material supports. Disagreement becomes an opportunity to examine assumptions rather than a problem to eliminate.
That discussion also describes AI reflexivity logs: records of why students consulted a tool, what it produced, and how they responded. These logs can reveal whether a student challenged an appealing summary or accepted it without sufficient examination. Their purpose should be learning, with room for uncertainty and revision. Turning reflection into surveillance would undermine the intellectual honesty the exercise should encourage.
Participant protection belongs in the same conversation. An earlier discussion within the book examines computational participation, asking researchers to consider how participant narratives enter AI systems and what subsequent uses might follow. Consent to an interview should prompt careful examination of whether the proposed computational processing aligns with what participants understood and authorized. Researchers must make these decisions before uploading sensitive material, rather than treating ethics as an afterthought.
These examples explain the educational purpose behind my forthcoming Qualitative Research with Artificial Intelligence. Preparing researchers for AI requires assignments that expose judgment, supervision that welcomes questioning, and institutional expectations that protect participant meaning. The closing chapter describes returning from AI-generated summaries to transcripts, audio, and fieldnotes, keeping computational suggestions open to challenge. That return to evidence is a discipline universities should teach explicitly.
The standard should be demanding: Can students show where an interpretation came from, explain what AI changed, and defend what remains uncertain? The responsibility for that explanation remains with the researcher, regardless of the tool’s sophistication. A fluent answer cannot demonstrate those abilities on its own. Universities must assess the reasoning behind the answer. Otherwise, higher education may produce research more quickly while preparing researchers less thoroughly, an outcome no claim of innovation should excuse.
- AI can generate research themes. Can students defend them? - November 9, 2026
- When AI’s architects ask to slow down, education must speak up - September 28, 2026
- Stop calling every misuse of AI plagiarism - August 19, 2026