I’ve said it before and been proved wrong but I think that we are moving towards an understanding about the impact of AI on education or more precisely on learning. There is a growing understanding that while Generative AI chatbots are good at supporting performance they are not effective in supporting learning. Indeed a series of research studies report that despite impressive performances while using AI for assignments, subsequent similar tasks undertaken without the support of AI display poor learning returns.

In his substack post Ethan Mollick, a long time AI enthusiast of AI in education,illustrates this with an account of an “experiment at a high school in Turkey with about a thousand students learning math. One group used plain ChatGPT, the other had no AI access. The students with ChatGPT did their homework better and thought they were learning more, but at test time, they underperformed their classmates without ChatGPT: Mollick says says this is “because the AI, designed to be a helpful assistant, was really just giving them answers, and actual learning requires mental effort. By short-circuiting effort, you short-circuit learning. That is why the initial results of AI on learning in classrooms can be so worrying.” Mollick is late to the scene with many educationists expressing concerns over lack of agency and of cognitive uploading.

But Mollick provides details of another study which he says did support and enhance student learning. He suggests that AI customised tutors can significantly boost learning when used properly. He claims there is a relatively small difference in how you use AI and yet it leads to big outcome differences. Worse, he continues is that “human nature leads us to make the wrong choices. Learning requires us to face our own ignorance and do hard intellectual work.”

Essentially Mollick is blaming learners for being lazy. But Maha Bali, “a professor of practice at the Center for Learning and Teaching at The American University in Cairo, has a different viewpoint in an account of her students end of term reflections. Maha is no fan of AI. She says “I am critical of the colonialist, capitalist, racist, Eugenicist roots of AI and the way it is overhyped and also has been historically used to harm humans and exacerbate inequality under the illusion of neutral automation. However, I have also believed for a long time that I cannot control or even attempt to control how students use AI. My overall approach is to spend a lot of time nurturing their critical AI literacy so they’re aware of the ethical considerations, hallucinations, biases, and roots of who and which ideologies lie behind AI tools, show them some alternative tools that hallucinate less (like Google Notebook LM and research tools like Keenious and all the Deep Research versions), then encourage them to use their own judgment to see where AI can support them in doing the work, without replacing them or limiting their creativity, their voice, their identity. And I just ask for transparency. But now she feels things are getting better because students themselves are realising the downside of AI for learning. “Students are getting better at using AI, not in the sense of getting better at prompting it to do more work, but in the sense that they are becoming better judges of what appropriate use looks like and what inappropriate use looks like. And able to self-control and take agency over their use of AI when I gave them freedom beside critical AI literacy." 
And Philipa Hardman wants to go beyond the idea of AI tutors: explaining the important difference between AI tutors, which largely improve performance (efficiency), and AI “study mates,” which are designed to help the student become a better learner.

I’ll come back to this issue in a subsequent post.