Wonkhe’s assessment research connects student accounts of feedback and AI use

Updated Sep 07, 2026

Wonkhe’s Trained to stop learning, released on 23 March 2026, examines students’ accounts of assessment, understanding and AI use. Its authors argue that assessment design deserves attention alongside AI rules. The evidence is exploratory and observational; it does not establish that late feedback causes students to use AI. Research announcement.

What was studied

The report describes February–March 2026 focus groups involving student representatives, plus 1,055 survey respondents from 52 UK providers, recruited through Wonkhe SUs subscribers. Survey results were weighted for gender and level of study. This recruitment route and weighting do not establish that the sample represents every UK student. Report, printed page 6.

The report says 38% sometimes submitted work they could not fully explain without returning to their sources. That qualification matters: this is a self-report, not a direct test of understanding or a misconduct finding.

Finding 10 presents accounts of feedback arriving too late to inform subsequent work and associations involving perceived feedback quality, clarity and AI usefulness. The authors recommend reviewing feedback timing and providing task-specific AI guidance. Report, printed pages 8, 27–28 and 36–37.

A local review worth undertaking

Our recommendation is to map the actual sequence of assessment dates, feedback release and the next opportunity to use that feedback. Discuss that sequence with students and teaching teams. A nominal turnaround target alone will not answer whether feedback was available at the point it was needed.

Keep different questions separate: whether instructions were clear, whether feedback arrived, whether it was useful, and what students understood to be permitted AI use. Ask what support they sought before assuming that a reported use of AI was a workaround or a breach of rules.

The report includes accounts of disability-related support and anxiety. These can prompt accessible local follow-up, but should not be generalised to all disabled students, women or institutions. Review support needs with the relevant students and services.

Comment analysis can organise reported experiences; it cannot identify the cause of a behaviour or prove learning loss. Compare themes with operational evidence and avoid interpreting survey associations as demonstrated effects of a proposed intervention. The student comment analysis governance checklist can help document those limits.

Review note, 7 September 2026: corrected the causal headline, restored the qualification to the 38% result, and clarified recruitment, weighting and the observational evidence. The review covered the announcement and relevant report methods, findings and recommendations.