In April, Ståle Ellingsen, Anne Bjune, and Sehoya Cotner (all BIO-UiB) published an essay in Science Norway on teaching with Large Language Models (LLMs) such as chatGPT. Here, we present data from a February 2023 survey indicating that students have been more likely to engage with the tool than instructors have, and that students generally think the impact of these tools on education will be more positive than do their instructors.
Further, we argue that some of the initial dialogue about the use of LLMs has been unconstructive, in the sense that many have focused their attention on the potential use of these tools to cheat on exams and assignments. While cheating is clearly possible with chatGPT etc., it has always been possible, and we urge our colleagues to think constructively and creatively about how best to use these tools in our courses. We suggest that instructors can (1) help our students understand the challenges and affordances of LLMs, in the context of our disciplines, (2) partner with our students to better understand the student perspective, and what needs they have identified, and (3) reconsider our assessments—are they evidence-based, equitable, and educational?
The three of us, together with David Grellscheid in Informatics, will be hosting an August workshop with several MatNat instructors. The aim of this workshop is to discuss how best to incorporate LLMs into our teaching, develop some simple modules for implementing in the Fall term, and assess the utility of these modules. We will certainly provide updates and suggestions!