Currently reading Forsyth et al. (2025) on “The impact of AI on inclusive learning communities”

While some people see big potential in GenAI to make learning and teaching more inclusive — by leveling the playing field through providing people with personalized instruction and support that they might otherwise not get — it is of course not as easy as that. GenAI is only as good as the data it is trained on and the training it received, and thus is replicating existing biases and inequalities. Forsyth et al. (2025) give a great introduction to the topic in their chapter “The impact of AI on inclusive learning communities“.

What I found most interesting is their thought experiment of AI giving feedback on, and providing suggestions for, group processes, like checking in on someone who hasn’t been active in the chat (I feel like some app on my phone tried to tell me that I haven’t contacted someone in a while and that I should, but I tend to turn off that kind of notifications as soon as possible, so I don’t remember what that was. But I know Facebook tries to remind you of people’s Birthdays, and LinkedIn of work anniversaries, neither of which I respond to, because the Birthdays I want to remember I put in my calendar myself, thanks a lot), or that a message might be perceived as sexist (my blog actually marks “probably offensive language”, which I usually appreciate [and, interestingly, “sexist” got marked as I am typing this!]). Is this going to make interactions better? AI can only give feedback on what it has access to, so what about offline chats in the corridor? Should they be recorded to give AI access, and what happens with that data? Also, what is the purpose of group work in the first place? Usually it is at least as much about the process as the product, so what happens if we let GenAI intervene in the process?

Also it is important to play with the idea of GenAI making suggestions for inclusive curricula and assessment practices. It might sound like a good idea, and maybe it is to some extent. But all learning and teaching and communication depend so much on the specific situation. The GenAI would have to have access to a huuuge amount of contextual information to even have the chance to do a good job. But do we really want to share that data with a GenAI?

In the end, Forsyth et al. (2025)’s recommendations boil down to these points:

• We need communities for teachers to explore GenAI and make sense of it together
• We also need continuous partnership with students to make sure everybody is ok with the actions and outcomes of change processes to counter distrust
• We need good policies that are formulated in ways that people can understand
• We need to work very hard to understand biases that GenAI propagates to make sure they aren’t fed back into our learning and teaching, and amplify existing inequalities

In the end, I agree with the authors that even though GenAI might be useful for some tasks and of course that should be explored, it is also “reasonable to consider that maybe [learning and teaching] is a human endeavour that is better done by humans, even if the process is messy, inconsistent, and uneven“.

Forsyth, R., Hamshire, C., Olumu, K., & King, E. (2025). The impact of AI on inclusive learning communities. In Handbook on AI in Higher Education. Edward Elgar Publishing.