AI has turned college exams into a 'wicked problem' with no obvious fix, researchers warn

In a rapidly evolving educational landscape, the emergence of generative AI tools like ChatGPT has dramatically altered the way universities approach examinations. A recent study published in the journal "Assessment & Evaluation in Higher Education" highlights the complexity of this issue, labeling it a "wicked problem"—a term used to describe challenges that are intricate and resistant to straightforward solutions. Researchers Thomas Corbin, David Boud, Margaret Bearman, and Phillip Dawson from Deakin University conducted interviews with 20 unit chairs at a large Australian university. Their findings revealed widespread confusion among educators, coupled with overwhelming workloads and no clear strategies to create assessments that are resistant to AI interference. The participants expressed varied opinions on AI; while some regarded it as a necessary tool for students, others feared it could lead to academic dishonesty and undermine genuine learning. The educators faced difficult dilemmas as they attempted to develop assessments that balanced the need for creativity and authenticity with the challenge of making them AI-proof. One instructor described their experience of trying to implement both AI-allowed and AI-restricted assignments as a "nightmare", only adding to their workload. Concerns were also raised about the effectiveness of stricter assessments, which might prioritize compliance over innovation. Furthermore, while oral exams may offer some resistance to AI influence, their scalability for large student groups remains a significant barrier. In an article for The Conversation, the Deakin professors emphasized that wicked problems are interconnected and messy, with no absolute solutions. They cautioned that every proposed fix could lead to new challenges and unintended consequences. Their recommendation for universities is to abandon the pursuit of a singular solution and instead allow educators the freedom to experiment and adapt their assessment methods continuously. The paper concluded with a stark warning: institutions that insist on finding the perfect answer to AI's role in assessments risk exhausting their teachers and ultimately failing their students. In practice, many educators are blending traditional methods with AI-aware strategies, such as incorporating handwritten tasks, live presentations, and personalized prompts that are difficult for AI to replicate. Some are leveraging AI to alleviate repetitive tasks, allowing them to focus more on student engagement, while others have opted to restrict AI usage in introductory courses due to issues like fabricated citations. Prominent voices in the educational community, including economist Tyler Cowen, argue that the implications of AI extend well beyond mere cheating. He suggests that the reliance on traditional homework and easily graded tests is becoming outdated. Similarly, LinkedIn cofounder Reid Hoffman advocates for more robust testing methods that are less susceptible to manipulation, proposing oral exams or integrating AI assessors into the examination process.