As generative AI becomes ever more ubiquitous, academics from around the University have been experimenting with using it in their teaching. Read on for inspiration as to how you too could adapt your teaching to include GenAI:
Break it to make it
Dr Harish Tayyar Madabushi, Lecturer in the Department of Computer Science: ‘I teach a third-year undergraduate module on natural language processing, which covers language models and generative AI. In an effort to help my students internalise our class discussions, especially regarding the importance of context in preventing errors in language models, I asked them to input the transcript of one of my lectures into a freely available generative AI engine of their choice and then attempt to lead the AI into making a mistake. This exercise was based on insights from research at Bath, demonstrating that providing context can significantly reduce the errors prone in generative AI models. Students consistently discovered that when the AI was given context, such as a lecture transcript, it became almost impossible to induce mistakes, even when they deliberately tried to confuse it. This vividly illustrated the importance of context in the functioning of generative AI systems while simultaneously providing students with practical insights into the use of generative AI. To improve this activity, I would be more directive about the type of questions students ask the AI, for example inputting a continuum from granular questions to more abstract questions, including those requiring inferential reasoning. This would allow them to gain a deeper understanding as to how AI operates and thus be better equipped in using AI as a tool for work and study.’
‘Getting off the blank page’
Kim Watts, Senior Lecturer in the School of Management: ‘On my MSc Marketing course GenAI is optional but encouraged. Evidence shows that students feel anxious about writing and how to start assignments. To support students with this anxiety and drive early engagement with the assignment, I aim to normalize the use of GenAI. Firstly, promoting it in the course handbook; secondly, running a skills session to support confidence in using it for writing and getting off the blank page. The session includes prompt development, bias, spotting hallucinations and so on, and directs them toward a LinkedIn Learning course on GenAI that they can access at any time. I also use Padlet so students have a forum to ask questions anonymously where I again encourage them to use AI. I point out that large language models (LLMs), such as ChatGPT or Bard, can help them structure their work and show them what’s expected in different formats. This is particularly reassuring for neurodiverse students who often struggle with understanding the format required. I also encourage students to explore how AI can help with graphic creation. Photoshop is time consuming, instead I point them towards GenCraft, DallE or PlaygroundAI to create mockups to use in their Marketing Plan Report. To ensure a level playing field, however, next time I run this course I intend to give a closed list of free AI engines they are permitted to use and be more directive about how to reference GenAI. It’s important that the difference in quality of outputs by premium AI engines, particularly image generators, does not become a factor in students marks.’
GenAI: our newest team member
Professor Steve Cayzer, Professor in the Department of Mechanical Engineering: ‘During my engineering master’s class, I use team-based learning to encourage collaboration and enhance student engagement. This year, I introduced GenAI as a tool and encouraged students to use it to improve the performance of their team. For one formative in-class assessment I gave each team a short-time frame to evaluate the sustainability strategy of a company the team had chosen. This prepared the students for the summative team assessment in which teams made strategic recommendations for the University. GenAI was optional for these tasks, and Claude or Perplexity were probably best adapted to it. They could also generate AI images to illustrate their points. Those teams who embraced the use of GenAI were able to produce both satisfactory and comprehensive answers within the time limit, though they lacked some originality. This activity gave students the conditions in which to learn how to use GenAI as a tool for their studies and future careers. The team-based approach, and the plenary session after the task, also enabled students to learn from each other. Before assigning this task, I asked students to complete a GenAI skills course offered by our university’s skills centre. Although I included a reflection on AI as part of the summative assessment, next year I intend to also require submission of the reflective log they create during the GenAI skills course to ensure that students engage with this essential preparation. I will also be more directive in suggesting prompts they could use with AI and in evaluating their use of it afterwards.’