A universal rule cannot serve every learning objective
A language model may be a legitimate research accelerator in one course and a substitute for the exact reasoning another course is designed to build. Clear course-level policies turn that distinction into an expectation students can understand before they begin.
The policy should explain permitted tasks, required disclosure, source responsibility, privacy constraints, and how students will demonstrate independent mastery. It should also evolve as evidence about learning changes.
The institution must fund the verification it demands
Hands-on work, oral defense, process records, and redesigned assessments require more instructor time than accepting a polished file. Guidance and communities of practice help, but staffing and pilot funding determine whether the promise becomes routine teaching.
Students also need AI literacy that includes refusal: knowing when automation undermines learning, privacy, collaboration, or professional duty. That judgment is itself a capability higher education should cultivate.
Go to the source
Read the evidence behind this analysis. External links open in a new tab.
MIT — AI and education as a watershed moment MIT — Charge to the committee on AI in teaching, learning, and research training
