Screenshot from one of the presentations outlining what ChatGPT is and is not: it is not human, not sentient, and not reliable!
This session on the robot uprising was facilitated by the University of Kent, and in a welcome contrast to some of the other sessions I have been to on AI recently, this was much more positive, focusing on early examples of using ChatGPT to enhance and support teaching and the student experience.
Some highlights were Maha Bali from the American University in Cairo who argued that we need cultural transparency around this technology as people are going to use it regardless of whatever regulations are put in place. This was echoed by some of the other presenters who noted that after graduation, when students enter industry, they will use, and be expected to use, any and all available relevant technologies. Someone else in the chat also noted that if you ban AI writing at university, then one outcome is going to be that students will only use it for cheating. So good luck, Cambridge. On transparent, ethic use, Laura Dumin from the University of Central Oklahoma talked about a new process they have implemented which asks students to declare if they have used AI tools to help with writing, and highlight which text has been AI generated so academics can clearly see this.
Some presenters had suggestions around re-focusing assessments along the lines of what ChatGPT can’t do, but which humans can. Some of these I feel are short term solutions. One person, for example, talked about how ChatGPT is generally better at shorter pieces of writing, so they have changed their assessments from 3x 800 word assessments throughout the year to 1x 2,000. Debbie Kemp at Kent suggested asking students to include infographics. I think these suggestions are going to work for now, but not in the long term. And the long term here isn’t even very long, given the pace of technological developments. By the time you could get changes to assessment through a programme board and in place for students, the technology may well have rendered your changes moot.
I think a better idea is around including more critical reflection from students. Margaret Bearman from Deacon University in Australia made the point that AI is not good at providing complex, context sensitive value judgements, and that I think is going to be a harder barrier for AI to overcome. Neil McGregor at the University of Manchester talked about this in a slightly different form. Instead of having students write critical reflections, they are now generating those with ChatGPT and asking the students to analyse and critique them – identifying what parts of the AI text they agree with, and where are the weaknesses in the arguments presented.
All of these sessions were recorded and are available on YouTube.
[…] about experimenting with ChatGPT to produce a response to one of their assignments and, echoing Margaret Bearman’s point from the teacher led session last month, they found that the result lacked critical analysis, […]