This post is a repost of an IMPACT Insights email originally sent to members of the GW community on September 16, 2025. Reposted with permission.
We’ve been reading and thinking about student assessments. Specifically, ways to help ensure that assessments measure how well students meet the intended outcomes, not how well ChatGPT can.
We’d recommend the podcast, AI, Cheating, and Trusting Students to be Human, an interview with Tricia Bertram Gallant, coauthor of the book The Opposite of Cheating: Teaching for Integrity in the Age of AI (request book through the Consortium Loan Service).
Try the following sections of the podcast:
- Where to start for instructors concerned about academic integrity? (11:35)
- Oral exams: Is it a dereliction of duty to not have them? Plus, a study showing students' appreciation of them. (26:08)
- How to respond if a student has cheated (32:10)
One of the ideas Dr. Bertram Gallant raises is friction, in this case, reducing the temptation to rely on GenAI. Friction is important to learning, and AI reduces friction. The literature we’ve been seeing focuses on two general ways to increase friction for students: visible assessments and process over product.
First, consider carefully: Is this knowledge or a skill students need to be able to demonstrate without any use of GAI? Do students need to remember or memorize this content?
If yes, and a proctored exam is not appropriate or desirable, consider making the assignment a visible one, where you can see students perform the skill, and/or focusing on the process of the assignment more than the product.
Need ideas for this? The Impact team can help with this, whether your course is in-person, hybrid, or online. Contact them at hsptech@gwu.edu.
