Western University professors redesign courses to combat AI cheating

Professors at Western University in Canada are fundamentally redesigning their courses and assessments to counter the rise of generative AI. The shift comes as traditional take-home assignments become increasingly vulnerable to automation by large language models.

To ensure academic integrity, instructors are implementing a variety of restrictive and adaptive strategies. One primary approach is the use of "walled garden" software, specifically a tool called MATCHA (Modern Authoring Tool for Certified Human Authorship). Developed by Professor David Bourget, MATCHA allows students to access necessary word processors and research materials while blocking access to AI tools during the writing process. Other faculty members are returning to traditional pen-and-paper in-class essays and increasing the frequency of oral testing to verify that students possess the conceptual knowledge they claim in their written work.

The failure of detection

This pivot toward structural prevention is driven by the perceived inadequacy of AI-detection software. Mark Daley, the Chief AI Officer at Western University, has highlighted that such detectors are not 100 percent accurate. Because of these reliability issues, the university has moved away from relying on detection tools, which are viewed as problematic due to their potential for false positives.

A crisis in assessment

The rise of models like ChatGPT and Claude has rendered previous "AI-resistant" prompting techniques ineffective, creating a crisis in how academic achievement is measured. For years, the take-home essay served as a primary metric for critical thinking and research capability. However, the ease with which LLMs can synthesize information and mimic academic tone has forced educators to choose between two extremes: locking down the digital environment to guarantee human authorship or radically integrating AI into the curriculum.

The path forward

As the academic world grapples with these tools, the goal for many educators is to maintain basic conceptual understanding in a workforce where AI is ubiquitous. While tools like MATCHA provide a technical stopgap, the long-term solution likely involves a hybrid model of assessment. Observers are now watching whether these "walled garden" approaches can scale across different disciplines or if the university will eventually move toward a fully AI-integrated pedagogy where the focus shifts from the final product to the process of learning.