Tuesday 14th May 12-1pm

Dr AFM Saiful Islam, Associate Professor (Teaching) in Medical Statistics at UCL

While ChatGPT can be a valuable tool for assisting learning and providing supplementary information, it's important to exercise caution when considering its use in high-stakes assessments like exams. In an educational setting, the focus should be on fostering genuine understanding, critical thinking, and problem-solving skills, which might not be adequately assessed through unsupervised multiple-choice question (MCQ) exams which we adopted in the past. Even when attempting to rewrite MCQ questions with new scenarios, it has been found that, in most cases, AI can pick the correct MCQ answers from reading those scenarios. While including images was a good option in the past, as ChatGPT was unable to identify them in its basic application, it now seems that images are also easily recognizable. We are also experiencing a situation where the exam cannot be conducted under locked/supervised conditions. Thus, in an unsupervised setting, there's a risk that test-takers could misuse ChatGPT to obtain answers without genuine effort or understanding. How can we overcome this challenge?

A copy of the webinar slides are available here.