Faculty FAQ

Short answers to the questions instructors raise most often

These are quick answers to the worries that come up most when an instructor first considers teaching with AI rather than against it. They are deliberately brief; the longer pieces in this folder work each one out in more depth, and they are pointed to where relevant.

Doesn’t allowing AI just make cheating easier? The tools are already in every student’s reach, so a ban mostly moves use out of sight rather than stopping it. The more reliable protection is to grade what a student can do unaided, the judgment and understanding the course is for, rather than the finished product the tool can generate. When assessment measures competence directly, copying an answer earns little, because the answer was never the point. The models in instructors/assessment-models.md describe how to do this without high-pressure exams.

AI is often wrong. Why let students use it at all? Because the unreliability is part of what they need to learn to handle. Students will use these tools throughout their studies and their work, and the skill that matters is checking the output rather than trusting it. A course that has students verify and correct AI work is teaching exactly the judgment that makes the tool’s errors visible. The point is not that AI is dependable; it is that learning to catch its mistakes is now part of the competence.

What about students who cannot afford the paid tools? This is a real fairness concern, and the design answers it in two ways. First, capable free tools exist, and you can build a course that never requires a paid one to succeed. Second, the part you actually grade is unaided competence, which is assessed without any tool in front of the student, so a student with no subscription is judged on the same footing as everyone else. Naming the unaided core, as instructors/semester-redesign.md describes, is what keeps access differences from turning into grade differences.

How do I grade work when I cannot tell what is the student’s and what is the AI’s? Stop trying to attribute the product, and assess the understanding behind it instead. A few minutes asking a student to explain their own submission tells you far more than any attempt to detect a tool’s fingerprints, and it is fairer, because it judges what the student can actually do. The models in instructors/assessment-models.md are built for exactly this, and they scale to large classes by sampling.

Won’t all of this take much more of my time? It takes some redesign at the start, and then less than you might expect to run. The checks are short, a few minutes each, and in a large class you sample rather than reach everyone every week. Starting with one redesigned assignment and one kind of check, as instructors/first-time-guide.md suggests, keeps the cost manageable while you see what works. Much of the upfront effort is one-time, since a redesigned assignment serves for years.

If students lean on AI, won’t they learn less or become dependent? That is the genuine risk, and it is the one the whole approach is built to counter. Passive dependence comes from using the tool to skip the thinking, and the redesigned assignments and the unaided checks are there to make the thinking the graded work, so leaning too hard shows up early and the student can correct course. The book’s concern about students recalling and understanding less when they offload the work is the reason for keeping a verified, unaided core at the center, not a reason to ban the tool.

My department or colleagues ban AI. Why should I do this? Even under a restrictive policy, you can teach students to verify rather than trust, which is the durable skill regardless of what any single course allows. Bans are also hard to enforce and tend to drive use underground, so a clear stance with a reason behind it usually serves students better than a prohibition they route around. The briefings in instructors/briefings.md include syllabus language for a restrictive stance if that is the policy you are working within.

How permissive should my course be? There is no single right answer; it depends on your course and what its unaided core requires. The syllabus briefings in instructors/briefings.md lay out restrictive, default, and permissive stances with the reasoning for each, so you can pick the one that fits and state it plainly. What matters more than the exact setting is that students know the rule and why it exists.