Chapter 15: Peer Review and AI as a Critical Friend
⚡️ Evaluating feedback, planning revision, and practicing epistemic caution
A complete draft creates the conditions for meaningful feedback. This week students exchange literature reviews and use the course rubric to evaluate structure, argument, synthesis, source use, and B2+ academic language. The purpose of peer review is not proofreading alone; it is to test whether a real reader can follow the reasoning and whether the evidence supports the claims. Students then prepare a prioritized revision plan. GenAI may also be used as an editor and critical friend to identify possible issues in cohesion, register, or academic clichés. However, all tool feedback must be evaluated rather than automatically accepted. Strategic prompting, source verification, and transparent documentation demonstrate epistemic caution—the ability to recognize the limits of both human and machine feedback.
Essential Questions
-
How can feedback improve a literature review without transferring authorship away from the writer?
-
What does epistemic caution require when evaluating peer or AI suggestions?
Guiding Questions
-
What rubric-based evidence makes peer feedback actionable?
-
How should revision priorities be ordered?
-
Which prompts invite useful AI analysis without requesting replacement prose?
-
What reasons justify accepting, modifying, or rejecting a suggestion?
Learning outcomes
-
Provide rubric-based peer feedback on a complete literature-review draft.
-
Evaluate structure, argument, synthesis, source use, and B2+ academic language.
-
Create a prioritized and specific revision plan.
-
Use GenAI strategically to analyze cohesion, register, or clichés without surrendering authorship.
-
Document and justify accepted, modified, and rejected feedback.
Key concepts and vocabulary
| Term | Working definition |
|---|---|
| Peer review | Structured feedback from a reader who applies shared criteria to a draft. |
| Actionable feedback | Specific feedback that identifies a problem, explains its effect, and supports a revision decision. |
| Rubric | A set of criteria and performance descriptions used to evaluate work. |
| Revision plan | A prioritized list of changes, reasons, and intended actions. |
| Global revision | Change to purpose, argument, organization, evidence, or section development. |
| Local revision | Change to sentence-level clarity, grammar, vocabulary, punctuation, or formatting. |
| Critical friend | A constructive reviewer or tool that challenges the work while respecting the writer’s decision-making. |
| Strategic prompting | Designing a prompt to request a limited, purposeful form of analysis. |
| Epistemic caution | Careful judgment about the certainty, source, and limits of knowledge claims. |
| Academic cliché | An overused expression that sounds formal but adds little precise meaning. |
Main chapter content
Reading as a reviewer, not a co-author
A peer reviewer’s task is to describe the reading experience and apply shared criteria, not to rewrite the paper. The reviewer should identify the central claim, the organizing principle, and the function of each section. If these cannot be identified, the feedback should explain where the difficulty begins. Comments should refer to specific passages and effects: a missing transition, an unsupported generalization, a paragraph dominated by summary, or a conclusion that introduces a new claim.
Reviewers should also recognize strengths. Identifying an effective synthesis paragraph or clear analytical move helps the writer understand what to repeat elsewhere. Balanced feedback is not praise added to criticism; it is evidence about what already works and what needs change. The rubric provides common language so that comments remain connected to course outcomes rather than personal preference.
Evaluating the five major dimensions
Structure concerns alignment among introduction, body, conclusion, and headings. Argument concerns the clarity and development of the central claim. Synthesis concerns relationships across sources and the writer’s analytical contribution. Source use concerns accuracy, relevance, attribution, citation, and balance. B2+ academic language concerns comprehensibility, register, grammar, vocabulary, cohesion, and the writer’s ability to express complex ideas.
These dimensions interact. A language problem may make an argument difficult to understand; a structural problem may cause repeated source information; weak synthesis may appear as a paragraph-level organization issue. Reviewers should identify the underlying problem rather than mark every visible symptom. Two or three high-priority comments can be more useful than dozens of corrections.
Creating a prioritized revision plan
After receiving feedback, the writer should group comments by level: global, section, paragraph, sentence, and formatting. Global changes come first because they may affect later text. The plan should state the problem, evidence from feedback, intended action, and reason. “Improve organization” is too general. “Move the methodological limitations paragraph to the end of Theme 2 so it qualifies the comparison before the section conclusion” is actionable.
Not all comments must be accepted. The writer can compare peer observations with the rubric, personal intentions, and source evidence. When several readers identify the same difficulty, the issue deserves attention even if the proposed solutions differ. When comments conflict, the writer should clarify the intended meaning and choose the revision that best serves the argument.
Prompting AI for analysis rather than replacement
Approved GenAI use should be narrow and diagnostic. A student might ask the tool to identify places where transitions do not clearly express the relationship, locate repeated academic clichés, or classify sentences as source reporting versus analysis. The prompt can request explanations and questions rather than rewritten paragraphs. This keeps the writer engaged in the actual revision.
The tool should not be treated as an authority on sources or academic truth. It may misread context, invent a problem, remove a necessary hedge, or standardize language in ways that erase the student’s voice. Any feedback about content must be checked against the original sources. Strategic prompting includes specifying the task, limiting the scope, protecting source text, and asking the tool to state uncertainty.
Documenting epistemic caution
Epistemic caution means recognizing that feedback is fallible. Peer reviewers may misunderstand; AI may generate plausible but inaccurate advice; the writer may also be attached to a weak choice. Responsible revision therefore involves comparison, verification, and explanation. The AI log and revision trail should record the prompt, suggestion, decision, and reason. Modified acceptance is often appropriate: the student may use the feedback to identify a problem but create a different solution.
This week’s deliverables include the completed peer review, the revision plan, and documented evaluation of permitted AI feedback. Students should be able to point to examples of accepted, modified, and rejected suggestions. The final decision remains with the writer, who must be able to explain how each major change improves accuracy, coherence, synthesis, or register. Feedback supports authorship when it strengthens judgment; it undermines authorship when it is copied without understanding.
From concepts to this week’s deliverable
Peer review is most useful when it produces a limited set of high-impact priorities. Reviewers should read once for the overall argument before annotating local issues. They can then evaluate five dimensions: alignment with the research problem, thematic organization, source synthesis, quality of analysis, and clarity/APA presentation. Comments should identify evidence and reader effect—for example, “This paragraph reports three studies but does not explain how their differences affect the theme”—and should avoid rewriting the author’s sentences. The writer groups comments by importance and creates a revision plan that addresses global structure before paragraph and sentence concerns.
GenAI can function as a critical friend only when the writer controls the question, checks the response, and retains responsibility for decisions. Productive prompts may ask the tool to identify places where a transition is unsupported, locate repeated academic clichés, or compare a paragraph with a rubric criterion. The student should not ask it to rewrite the review or generate new scholarly claims. AI feedback is evaluated against the manuscript, sources, course criteria, and peer comments. The updated revision trail records accepted, modified, and rejected suggestions with reasons, while the AI log preserves prompts and relevant outputs. Epistemic caution is visible when the writer can explain what the tool missed, oversimplified, or treated with unjustified certainty.
Suggested weekly learning sequence
A productive learning sequence for this chapter begins with a diagnostic response to the essential question, “How can feedback improve a literature review without transferring authorship away from the writer?” Students should first record an individual answer and then compare it with examples, course criteria, and peer interpretations. A short concept-sorting activity using Peer review, Actionable feedback, Rubric, and Revision plan can make important distinctions visible before students work on a full draft. The instructor can then model one decision aloud, showing not only the completed example but also the questions, alternatives, and evidence that led to it. Students apply the same process to a small portion of their own project and receive formative feedback before extending the work independently.
The second part of the week should move from practice to production. Students use the guiding questions—“What rubric-based evidence makes peer feedback actionable?” and “How should revision priorities be ordered?”—as checkpoints while working toward the stated outcomes, especially the ability to provide rubric-based peer feedback on a complete literature-review draft and to evaluate structure, argument, synthesis, source use, and B2+ academic language. A brief peer conference should focus on one high-impact feature rather than attempting to correct everything at once. Students then revise, annotate the evidence of their decision-making, and complete the week’s deliverable: Rubric-based peer review; prioritized revision plan; evaluated GenAI feedback; updated revision trail and AI log. The final five minutes of class can be used for an exit reflection identifying one decision the writer can now justify and one question that must be carried into the next chapter.