John Warner: AI Can't Stand In for Peer Review in Academia

In his Inside Higher Ed column 'Just Visiting,' John Warner argues that generative AI cannot substitute for peer review in academia, in either of the forms the term covers.

Scholarly publishing is the easier call. 'A large language model is in no sense a peer,' Warner writes, 'so any publication that chooses to use these tools as substitutes for peer review should be transparent that they've done so and drop the pretense that they still participate in a system of peer review.'

Student peer review, which Warner prefers to call peer response, is the subtler argument. The practice, he writes, is one 'where students exchange drafts of a writing assignment with one of their colleagues and then evaluate it according to criteria established by the instructor.' The teaching value, he argues, is not what the writer gets back. 'The core purpose of peer response is not to receive feedback, but to give it.' An LLM can produce comments, but it cannot do the reflective work a student does when reading and judging a classmate's draft.

Warner reaches past feedback quality to a point about pacing. 'Maybe we learn more when feedback is not on demand and we have to sit with a particular problem for a period of time,' he writes, framing the loss as a cost to 'productive struggle.' Automation, he concludes, 'will never be human.'

Four of the AI commentators we follow shared the column shortly after it ran.