post-image

From Feedback to Scholarly Judgement: Teaching critical compassion through AI-assisted peer review

Academic Leadership Active Learning Critical Practice Note Learning & Teaching

The provocation

Reviewing is a fundamental academic practice, whether we are evaluating a journal article, critically examining literature, providing feedback on a draft or reviewing a peer’s work. These activities require more than identifying strengths and weaknesses. They involve recognising quality, questioning assumptions, weighing evidence and communicating judgements that are both rigorous and constructive.

Generative AI complicates these practices because it can produce convincing summaries, critical appraisals and apparently well-reasoned feedback. Yet the ability to generate a plausible review is not necessarily evidence of the intellectual judgement involved in producing one. Nor does receiving feedback guarantee that students know how to interpret, challenge or act upon it.

Rather than attempting to make reviewing resistant to AI, perhaps we should reconsider how we teach students to review in the first place. How might we progressively develop the skills of critical judgement, compassionate criticism and scholarly dialogue, while using AI to challenge rather than replace students’ own thinking?

The critical question

How might we help students develop the confidence, capability and responsibility to challenge a judgement—including one offered by AI—while treating the author with critical compassion?

What the research suggests

Three open-access papers provide a foundation. Carless and Boud (2018) describe feedback literacy through four interrelated capacities: appreciating feedback, making judgements, managing affect and taking action. Feedback is not merely information delivered to a learner: it becomes educationally consequential when the learner interprets and uses it. Their account of coached peer review, exemplars and revision supports repeated opportunities to practise both giving and receiving criticism.

Bearman et al. (2024) argue that evaluative judgement depends on an understanding of quality and an ability to appraise work in context. They distinguish judging AI outputs, judging the processes through which AI is used, and asking AI to interrogate students’ own judgements. Yet using AI does not automatically improve judgement: persuasive outputs can encourage uncritical acceptance or cognitive offloading. Comparison with credible disciplinary standards, exemplars and human feedback remains essential.

Corbin, Tai and Flenady (2025) distinguish recognitive human feedback from extra-recognitive feedback generated by AI. Human peer review can involve reciprocal recognition, trust and vulnerability; an AI system can provide useful criticism but does not participate in that relationship in the same way. Their suggestion that AI can operate as a preparatory “sandbox” is especially useful here, provided it supplements rather than displaces meaningful human review. These are conceptual arguments, not direct evidence that the sequence proposed below will work.

An approach to try

These four iterations offer an adaptable progression, not four mandatory workshops or fixed stages of development. Tutors can vary the materials, timing and level of challenge, and decide whether activities are formative or assessed. Each iteration includes a suggested learning output that can be kept, discussed or used in the next activity.

01  ESTABLISH  Learning to give and receive feedback

Begin with empathy: how might an author experience this comment? Move to compassion: how might it support their development? Then introduce critical compassion: how can criticism be rigorous while respecting the author’s capacity to reason, respond and disagree? Using brief examples, students distinguish personal judgement, vague praise and specific, evidence-based, actionable feedback.

Suggested output: A short feedback commentary on an exemplar, identifying a strength, a substantive concern and a constructive recommendation, with a brief reflection on tone and likely reception.

02  CHALLENGE  Learning to give critical feedback

Introduce writing with contestable claims, competing strengths or weaknesses of differing significance. Students interrogate evidence, assumptions and reasoning rather than concentrating on surface corrections. Through comparison with examples and criteria, they prioritise recommendations and articulate the standards of quality underlying their judgements. Compassion is not an excuse to avoid difficult criticism.

Suggested output: A prioritised critical review identifying two or three substantive issues, each supported by textual evidence and an explanation of why it matters.

03  INTERROGATE  AI-assisted review of the review

Students first commit to an independent, evidence-based review. They then invite an AI system to produce the strongest defensible counterarguments, recognising sound observations rather than disagreeing automatically. Against the original work, criteria and credible sources, students decide what to retain, revise or reject. Tutors may offer an example prompt, but should encourage its testing and refinement.

Suggested output: A revised review, accompanied by a concise decision record showing which AI objections were accepted, modified or rejected, and why; optionally note how the prompt was improved.

04  APPLY  Reciprocal scholarly peer review

Students submit a draft and receive two independent human peer reviews. Reviewers may use the third iteration to test their own criticism where institutional AI rules permit. Authors revise their work and respond to reviewers, explaining which recommendations they accepted, adapted or declined. Where both reviews are inadequate, an additional tutor review may be requested. The process resembles scholarly publishing but remains adaptable to local assessment design.

Suggested output: A review-and-revision dossier comprising the initial draft, two peer reviews, the revised work and a point-by-point response to reviewers. This may be formative, part of a portfolio or attached to a final submission.

Where AI fits — and where it does not

A suggested prompt for Stage 3 is: “Act as a rigorous but compassionate critical interlocutor. Examine my review against the supplied work and criteria. Identify what I have judged convincingly, then make the strongest evidence-based counterarguments to weak assumptions, overlooked material or questionable priorities. Do not invent objections for the sake of disagreement. Distinguish major from minor issues; ask me to defend my claims. Do not rewrite my review. Help me decide what to retain, revise or reject, and acknowledge when evidence is insufficient.”

Treat the prompt as a prototype: test it on different examples, observe where it produces superficial agreement or manufactured opposition, and revise it. AI-generated criticism is not an authority in itself; students must evaluate its content and the appropriateness of using the tool. Before uploading another person’s work, tutors should establish institutional requirements around approved platforms, confidentiality, data protection and consent, with an equitable alternative where needed.

Questions for adaptation

  • Where could the four iterations sit within an existing module, and what counts as defensible judgement in this discipline?
  • How might students demonstrate changes in their reasoning and action—not merely produce polished reviews?
  • What safeguards are needed to preserve independent thinking, fair peer review and responsible use of classmates’ work?

Three sources to explore

Carless, D., & Boud, D. (2018). The development of student feedback literacy: Enabling uptake of feedback. Assessment & Evaluation in Higher Education, 43(8), 1315–1325. https://doi.org/10.1080/02602938.2018.1463354

Bearman, M., Tai, J., Dawson, P., Boud, D., & Ajjawi, R. (2024). Developing evaluative judgement for a time of generative artificial intelligence. Assessment & Evaluation in Higher Education, 49(6), 893–905. https://doi.org/10.1080/02602938.2024.2335321

Corbin, T., Tai, J., & Flenady, G. (2025). Understanding the place and value of GenAI feedback: A recognition-based framework. Assessment & Evaluation in Higher Education, 50(5), 718–731. https://doi.org/10.1080/02602938.2025.2459641

Method & provenance

This note was developed through AI-assisted critical enquiry, including a review of research on automated assessment, human–AI grading reliability and institutional guidance on AI-assisted feedback, alongside critical reflection and iterative drafting. The author’s proposed human-led assessment workflow, in which AI interrogates academic judgement rather than independently evaluating student work, is an interpretive contribution; its effectiveness has not been empirically tested. The enquiry draws on published research and institutional guidance, although the three-source, full-text verification process remains to be completed. The initial provocation emerged from the author’s own exploration of AI-assisted assessment and feedback in higher education.

ABOUT THIS SERIES

Short, evidence-informed provocations that move from critical enquiry to adaptable teaching practice. Each note draws on a small, accessible reading list and makes the role of AI-assisted scholarship transparent.

Download Critical Practice Note 1 as a PDF file

Tags:
, , , ,

Leave a Reply

Your email address will not be published. Required fields are marked *