The launch of Chat GPT and emergence of other Large Language Models (LLMs) in late 2022 put higher education in a defensive position. Alarms were raised about the potential for academic misconduct, the inaccuracy and bias in the information generated by Generative AI (GAI), and the amount of personal information being shared by users. Having been forced only two years prior to rethink how university programmes were taught and assessed, with a shift away from exams towards open book and coursework assessments, academics in the US and around with world declared the essay to be dead. Universities scrambled to review assessments, with secure exams mooted by some as the only solution to the ‘threat’ of GAI.
In 2023, the conversation has moved on from ‘how can we stop students using GAI to cheat?’ to ‘how can we use GAI to better prepare students for the future?’ Russell Group institutions have developed a set of five AI principles aiming to support AI literacy in students and staff, embed GAI in teaching and assessment to support equal access, encourage collaboration and the sharing of best practice across the sector and ensure academic rigour and integrity is upheld. Some universities, such as King’s College London, have adopted the PAIR (Problem, AI, Interaction, Reflection) framework in order to integrate AI into their curricula and have developed extensive guidelines for students, teaching staff, and programme managers on the incorporation and fair use of GAI in learning, teaching and assessment.
In February 2023, the University of Kent’s popular Digitally Enhanced Education webinar series delivered a session on Chat GPT. The session exceeded its maximum participant capacity. However, of the many submissions for Advance HE’s Assessment and Feedback Symposium in November 2023, only four related directly to generative AI. Does this indicate that institutions feel they have ‘solved’ the GAI issue, or that they have made the changes they are able to for the current academic year due to restrictions on policy and process but have not yet effectively addressed or embraced GAI?
