AI has turned college exams into a 'wicked problem' with no obvious fix, researchers warn (www.businessinsider.com)

A new paper in Assessment & Evaluation in Higher Education by four Deakin University researchers argues that generative AI has turned college assessment into a “wicked problem” — messy, interconnected, and without a clean fix. Based on one-hour semi-structured Zoom interviews in late 2024 with 20 unit chairs at a large Australian university, the study documents widespread confusion, burnout, and impossible trade‑offs: attempts to offer both AI-permitted and AI-free assignments doubled workload; stricter, compliance-focused tests risk killing creativity; and scalable alternatives like oral exams are often impractical. Faculty responses range from banning AI after spotting hallucinated citations to embracing it as a productivity tool for lesson planning and grading. For the AI/ML community this matters because it reframes assessment not as a narrow detection problem but as a systemic design challenge with technical and pedagogical trade-offs. Detection and bans are partial and generate unintended consequences; personalized prompts, reflective tasks, handwritten baselines, and live presentations are more robust but costly. The authors urge institutions to stop chasing a silver bullet and instead allow compromise, divergence, and iterative redesign — a call that has implications for tool-makers, educators, and policymakers who must balance model capabilities, hallucination risks, scalability, and the evolving purpose of assessment (including proposals for AI-assisted examiners).