Educators worldwide are dealing with the advent of AI, what it means for teaching, and especially what it means for homework. There are many ideas flying around, both positive and negative, but mostly complicated. When I attended the SABER East conference in May it was nice to see someone present a refreshingly simple and fun idea for using AI.
Crystal Uminski from Towson University gave a talk called “Bad AI can be good for teaching” based on a paper she recently published in the Journal of Microbiology and Biology Education. To prepare for the lesson, she asked an AI bot to generate an image of something biological, for instance to “create a diagram of a cell membrane”. While there are reasonable elements to the resulting image, the AI often gets some of the meaning wrong. Uminski then presents the AI generated image to students to prompt thinking and generate peer and classroom discussion.
Dr. Uminski threw out several ideas for how to use these images. In the simplest version, an AI-generated image is shown together with a textbook image, and students are asked to note the differences. Or students may be asked to redraw the image themselves, correcting as they go, and then explain their drawing to a peer. Another idea is to ask students how they might identify that the image is incorrect and possibly AI-generated, and discuss the implications of using an inaccurate image in the real world.
These exercises help students build literacy with the visual representations used in biology, as well as with biology content. Unfortunately for this approach, AI models keep getting better, so with the newest models you might need to force the AI bot to generate a poor image by creating a very ambiguous prompt or explicitly telling it to include errors. Even so, the exercise is useful and fun, is quick to develop and may be used multiple times throughout a course.
SimBio Ecology – Comprehensive ecology education
