Few of us set a review to find out whether students can summarise. We set it to see whether they can judge other people’s work. Many of our rubrics never said so, and summarising is exactly what an AI tool does best. A review that rewards coverage now rewards the AI tool.
Two further problems follow. AI tools invent plausible references: Walters and Wilder (2023) found that 18 to 55 per cent of chatbot-generated references were fabricated, so a student’s reference list is no longer a reliable record of reading. And peer review, which builds judgement better than most of us credit (Double et al., 2020), can be quietly outsourced. Banihashem et al. (2024) found that AI-generated feedback described more, while students were better at pinpointing the problem. Pinpointing the problem is the judgement.
So stop paying for summaries. Supply sources that disagree and mark the comparison, make the source trail assessable, cut the word limit so choosing becomes the skill, and mark the peer comments themselves, each one quoting the passage it refers to. An education studies vignette shows the redesign at work, and why it was worth the extra marking.
Reviews
There are no reviews yet.