AI Detectors Fail, Universities Turn Away from Reliance on AI-Detection Programs

Academic integrity is facing unprecedented challenges as faculty members struggle to combat the growing use of artificial intelligence (AI) tools by students. The ease with which access to these tools has made large-scale cheating a reality has left many institutions scrambling for solutions. However, recent developments suggest that relying on AI detectors may not be the answer they were hoping for.

In fact, numerous studies have shown that AI detectors exhibit inconsistencies and produce false positives when evaluating human-generated writing. This can lead to accused students filing lawsuits against their universities, as seen in some cases where those false positives resulted in disciplinary action. The issue is further complicated by biases against non-native English writers, which can unfairly penalize students from diverse backgrounds.

The reality of AI detectors’ unreliability has prompted a growing number of institutions to reevaluate their approach to academic integrity. At least 12 universities have disabled Turnitin’s AI detection software due to its high false-positive rate, while others are banning or discouraging faculty reliance on these programs as the sole evidence of alleged cheating.

Yale University is one such institution that has taken a proactive stance against relying solely on AI detectors. The Kelley School of Business at Indiana University also prohibits the use of AI detection tools in favor of designing assignments that encourage process, reasoning, and authentic engagement. This approach recognizes that simply trying to ‘catch’ AI use can be an exercise in futility.

Jennifer Frederick, executive director of Yale’s Poorvu Center for Teaching and Learning, notes that students are increasingly deploying AI to game the detectors. She argues that instead of engaging in a cat-and-mouse game with AI detection tools, universities should focus on teaching students how to use these technologies responsibly and ethically.

The shift away from relying solely on AI detectors is also motivated by concerns about the impact on student learning. Kevin Yee, director of the Faculty Center for Teaching & Learning at the University of Central Florida, notes that in larger classes, there’s less personal connection between students and faculty, making it more difficult to convince students not to take shortcuts.

Yee co-authored a guide with suggestions centered around making students show how they are using AI to support their work. He also recommends feeding assignments into large language models to get ideas on how to make them more AI-resilient. However, redesigning assessments that can’t be outsourced to AI is crucial for institutions looking to prove the value of their course offerings.

Marc Watkins, a writing and composition lecturer at the University of Mississippi, emphasizes that getting faculty on board with changing traditional assessments will be ‘the biggest lift’ for universities responding to rising AI-related academic integrity concerns. He highlights the human labor and resources required to scale AI-resilience tactics such as proctoring oral exams.

Tricia Bertram Gallant, director of the academic integrity office and testing center at UC San Diego, points out that higher education has been slow to adapt to changes in assessment methods over the past 20-plus years. She notes that relying on unsupervised written word as evidence of learning is no longer feasible due to AI’s impact.

Bertram Gallant advises departments to focus on teaching foundational skills required for using AI responsibly and ethically, rather than policing AI use through assessments. She emphasizes the need to give students a distraction-free environment where they can demonstrate their knowledge without relying on AI tools.

The challenges posed by AI-generated content are not limited to written assignments; machine learning jobs and AI tools for businesses also require careful consideration of academic integrity concerns.