New survey research at the University of Chicago suggests AI usage among students may be difficult to measure accurately, raising stakes for campus policy enforcement. In an anonymous survey of 338 undergraduates, 60% reported personal use of AI tools such as ChatGPT, but 90% said they believed the average student uses AI—creating a 30-point gap. Researchers behind the study suspect “social desirability bias” is driving underreporting, including reluctance to appear unable to complete coursework independently, plus fear of consequences. The researchers also consider overestimation as a factor, where students may infer peer norms from visible classroom and campus behavior. The finding matters for institutional governance: without reliable baseline data on how widely students use AI and for what purposes, colleges risk designing rules based on assumptions rather than evidence—potentially undermining both academic integrity enforcement and student trust.