At the University of Chicago, researchers surveyed 338 undergraduates and found a large gap between self-reported AI use and beliefs about peer use. While 60% said they personally use AI tools such as ChatGPT, 90% believed the average student uses AI—creating a 30-point difference that researchers suspect is driven by misreporting and shame. The study’s authors say the pattern resembles “social desirability bias,” where students may underreport usage because they worry about being viewed as dishonest or unable to do coursework independently. Researchers also suggest some students may overestimate peer use because AI tools are visible in day-to-day campus life. The findings raise a policy challenge for universities writing AI rules: institutions may be designing enforcement and academic integrity approaches based on assumptions rather than reliable data. The development also underscores assessment pressure—if AI use is difficult to measure through self-reporting, universities are likely to weigh more robust, behavior-based and oral demonstration methods.