I caught my students using AI to cheat in an exam — here’s what universities must do to stamp this out

Utilizing AI for unauthorized assistance during examinations has become an increasingly prevalent issue in academia, as highlighted by an incident at Brown University. During a recent semester, I uncovered evidence of widespread illicit use of artificial intelligence to gain an unfair advantage on a take-home exam for my course on welfare economics and social-choice theory.

Despite raising the stakes with this more challenging assessment, the average score reached an unprecedented 96%, far exceeding the typical range of 65–80% observed in previous iterations of the course. Alarmed by this anomaly, I opted to administer the final examination in person, without the option to refer to any resources. The average score on this in-person, closed-book exam plummeted to 48.6%.

Moreover, several students either withdrew from the course or failed to take the final exam. This situation underscores the urgent need for universities to reassess their approaches to examinations in light of the rapid advancements in generative AI tools. The most troubling aspect of this instance was not the presence of cheating on the take-home exam, but the apparent extent of its occurrence.

Universities have grappled with addressing this challenge effectively. It is imperative that they do not turn a blind eye to this issue, as the integrity of learning and assessment hangs in the balance. Two critical areas demand immediate attention: reviving a culture of academic integrity and adapting educational practices to accommodate the era of AI.

The persistence of cheating at universities, in part, can be attributed to academic leaders' reluctance to confront it head-on. When university administrators fail to unequivocally defend academic integrity, they inadvertently foster an environment where dishonest practices are normalized. My experience and numerous communications with faculty members at other institutions suggest that some administrators respond to evidence of extensive cheating with hesitation or denial.

This hesitation may stem from a desire to protect the institution's reputation and market value. However, this approach overlooks the factors that truly determine a university's standing: high standards, a strong work ethic, and genuine academic excellence. The rise of AI poses significant challenges, including the potential to exacerbate inequality if left unchecked.

To mitigate this risk, it is crucial to amend copyright licenses and reassert human control over AI usage. The future of scholarly peer review must incorporate AI support rather than resorting to AI bans. As conflicts continue to escalate in war zones, protecting education in these volatile environments becomes paramount. Strategies for optimizing the benefits of tech-enhanced sleep, such as improved rest and productivity, must be balanced with concerns over privacy.

An alarming discovery of over 18,000 questionable images found in antibody catalogs from 15 companies further emphasizes the need for vigilance and control in the application of AI technologies.