by Mark Hopwood
There is no silver bullet for ChatGPT. Nothing that we can do will make our courses and assignments completely AI-proof, and no AI detection program is completely reliable at identifying AI-generated text. According to James Lang (the author of Small Teaching, Cheating Lessons, and many other influential books on pedagogy), these facts should not cause us to despair. Even if we cannot completely rule out the possibility that students will cheat, there are plenty of relatively simple things we can do to make it much less likely that they will be tempted to do so. Furthermore, the fact that AI can be used to complete a particular assignment does not mean that there is no value in teaching students to do it for themselves. If anything, the advent of large language models only serves to underscore the importance of some basic principles that have always been true of college teaching:
(i) students will do better—and be less tempted to cheat—when we are clear and transparent about what we are asking them to do and why it is important for them to do it, and
(ii) giving students the opportunity to engage in metacognitive reflection on their own learning process not only helps them to retain what they have learned, but also to see the value of doing the work for themselves.
These insights were some of my main takeaways from a webinar delivered by James Lang on “Academic honesty in the age of AI” earlier this semester. (A full recording and slides are available here.) I found the webinar very helpful, both for the measured and realistic discussion of the potential impact of AI on college teaching and also for the concrete suggestions offered for how we can reduce the temptation to cheat while also improving students’ understanding of their own learning. One particularly interesting takeaway was Lang’s discussion of a recent study of a range of “low-effort cheating-reduction methods”, which showed that strategies as simple as reminding students of the AI policy midway through the semester (rather than mentioning it only once at the beginning) and writing transparent assignments that explain why students are being asked to do what they are being asked to do made a significant difference to the incidence of cheating in intro-level computer science classes. Of course, none of these strategies make it impossible for students to cheat, but as Lang noted, students have always cheated, even before the emergence of ChatGPT. Protecting academic integrity is critical, but it is also important for us to remember that our main job is the same as it has always been: helping students to learn. What was perhaps most helpful about the webinar was the way in which Lang demonstrated that promoting student learning and protecting academic integrity need not be in tension with each other. In fact, the very same strategies that help us to achieve the first aim will also help us to achieve the second. Variety in teaching methods, transparency in assignments, and opportunities for metacognitive reflection have always been basic principles of good pedagogical practice, but in the age of AI they have become even more essential.
I highly recommend watching the recording of the webinar and trying out some of the specific strategies Lang recommends in your own classes. If you have tried out any ideas to help your students learn while also reducing the temptation to cheat, let us know: we’d love to feature them in a CfT blog post or session!