Universities Recast Assessment for AI Era as 120-Student Study Flags Retention Risks
Updated
Updated · The Washington Post · Aug 9
Universities Recast Assessment for AI Era as 120-Student Study Flags Retention Risks
3 articles · Updated · The Washington Post · Aug 9
Summary
Universities are being urged to separate independent competence from AI-assisted performance, rather than treating polished submitted work as reliable evidence of learning.
Generative AI can produce strong prose without the reasoning behind it, while a randomized study of 120 undergraduates found unrestricted ChatGPT use during studying led to lower scores on a surprise retention test 45 days later.
That does not amount to a blanket case against AI: a separate randomized undergraduate physics trial found a purpose-built AI tutor delivered stronger immediate learning gains than an active-learning class.
The proposed response goes beyond reviving blue books and proctored exams, adding oral defenses, drafts and revisions, and assignments that explicitly state permitted AI use and the skill being measured.
The broader stakes reach beyond campus, as automation may hollow out entry-level training and degrees must still signal to employers and the public that graduates can exercise judgment without overdelegating to machines.