Students' feelings about AI reveal trust and belonging risks universities miss
Updated May 24, 2026
Universities often ask whether students are using Generative AI. The harder question is what emotions the institution is teaching students to feel when they do. At Student Voice AI, we see AI-related comments arrive wrapped in worry, relief, confusion, and judgement, which is why Glenys Oberg et al.'s Higher Education paper, "Feeling AI: Circulating emotions, institutional climates, and moral boundaries in student use of AI", matters. For universities using student voice to understand AI, assessment, and belonging, the study shows that policy signals can shape student feeling before they shape behaviour.
Context and research question
Much of the current debate about AI in higher education still defaults to a narrow frame: are students cheating, or not? That matters, but it is too thin for Student Experience, Quality, and Market Insights teams trying to understand what AI is doing to trust, help-seeking, and academic confidence. If students experience AI use through fear, shame, relief, or moral uncertainty, then a headline usage rate will miss the most important part of the story.
This paper tackles that gap through a national Australian study combining a survey of 8,021 students with qualitative focus groups involving 79 students. The authors use a sociotechnical lens and Sara Ahmed's idea of affective economies to ask how emotions circulate around AI in higher education. For UK universities, the design is useful because it connects large-scale signal with qualitative explanation, and treats emotion as part of the student experience rather than as noise around the real issue.
Key findings
The first headline is ambivalence, not enthusiasm or resistance. The study found strong pairings of optimism with excitement, but those feelings often sat alongside scepticism and worry. That matters for universities collecting AI-related feedback because students are not dividing neatly into supporters and opponents. Many are trying to balance convenience, uncertainty, usefulness, and moral risk at the same time.
Assessment emerged as the main site where AI becomes emotionally charged. Students described an environment in which institutional warnings, plagiarism concerns, and unclear boundaries made AI feel risky even when their use was minor or exploratory. This helps explain why AI-related comments can look harsher than simple adoption data would suggest. Much like earlier evidence that students use Generative AI for feedback, but trust teachers more, the issue is not just access to a tool. It is whether the surrounding context feels safe and trustworthy enough to use it without second-guessing yourself.
Students also treated creativity and authorship as moral boundaries, not just technical ones. The paper shows that many participants saw their own voice, effort, and originality as part of who they were as students. One comment captures that clearly:
"I really like what I write… [AI] kills the writer's voice"
That reaction matters because it shows why some students experience AI use as more than a study aid decision. They experience it as a question about legitimacy, selfhood, and what counts as real academic work.
The institutional climate around AI shapes belonging as well as compliance. When students feel watched, vaguely warned, or unsure where the boundary sits, AI policy can become a trust issue. That sits alongside recent evidence that students judge AI-using teachers by care, not just technical competence. In both cases, students are reading AI through a relational lens: does this make the university feel clearer, fairer, and more supportive, or more suspicious and distant?
Practical implications
For UK higher education teams, the first implication is to stop measuring AI only through usage and misconduct categories. Surveys and pulse checks should ask whether AI feels helpful, risky, relieving, confusing, or unfair, and when those feelings arise. That gives institutions a clearer view of whether a concern is really about academic integrity, or about a lack of confidence, support, or policy clarity. The benefit is sharper diagnosis before sentiment hardens into mistrust.
Second, universities should treat AI guidance as part of the student experience, not just the regulations page. If students are unsure whether small acts of clarification, checking, or drafting support are legitimate, they are more likely to self-police anxiously or hide their behaviour. Clear examples, course-level discussion, and visible staff judgement can reduce that uncertainty, especially where students already turn to AI because support feels private and instant, as seen in recent work on why students seek GenAI when they feel stuck. The benefit is lower anxiety and more honest engagement.
Third, institutions should read AI-related free-text comments as evidence about trust and belonging, not only policy reaction. Open comments can show whether students feel reassured, monitored, embarrassed, or unclear about expectations, and whether those patterns differ by subject, level, or cohort. If that evidence is going to inform policy, it needs a governed workflow such as the student comment analysis governance checklist. Student Voice Analytics fits naturally here because it helps teams separate recurring themes such as trust, fear, fairness, authorship, and support gaps at scale. The benefit is a more defensible evidence base for AI policy and course design.
Finally, universities should remember that trust needs active maintenance once AI enters teaching and assessment. A technically correct policy can still feel punitive if students experience it as suspicion first and guidance second. The most useful student voice work will therefore ask not only "Did students comply?" but "What kind of learning relationship did this policy create?". The benefit is better alignment between AI governance and the student experience institutions say they want to protect.
FAQ
Q: How should universities ask students about AI in a way that produces useful evidence?
A: Ask about specific experiences, not only headline approval or disapproval. Separate questions on usefulness, trust, worry, fairness, and clarity of guidance, then add an open-text prompt such as "What feels helpful or difficult about AI use on your course?". That makes it easier to distinguish policy confusion from deeper concerns about learning, voice, and support.
Q: What should UK teams keep in mind before generalising from this study?
A: This is a national Australian study conducted during an early, high-pressure phase of AI policy development, with a large survey and qualitative focus groups. That gives it strong value as directional evidence, but local context still matters. UK institutions should use the findings to sharpen their own questions, then test whether the same emotional patterns appear in their comments and surveys.
Q: What does this change about student voice practice more broadly?
A: It suggests AI-related feedback should be read as part of the wider student experience, not parked as a niche technology issue. Comments about guilt, relief, trust, or authenticity can reveal how students interpret academic standards, staff support, and institutional care. That makes student voice on AI more diagnostic, and more useful for enhancement, than a simple question about whether students are "for" or "against" the technology.
References
[Paper Source]: Glenys Oberg et al. "Feeling AI: Circulating emotions, institutional climates, and moral boundaries in student use of AI" DOI: 10.1007/s10734-026-01658-6
Request a walkthrough
Book a free Student Voice Analytics demo
See all-comment coverage, sector benchmarks, and reporting designed for OfS quality and NSS requirements.
- All-comment coverage with HE-tuned taxonomy and sentiment.
- Versioned outputs with TEF-ready reporting.
- Benchmarks and BI-ready exports for boards and Senate.
UK-hosted · No public LLM APIs · Same-day turnaround