Brown University probes widespread AI-assisted cheating in economics course

When an economics professor at Brown University noticed unusually high exam scores that defied traditional grading patterns, the discovery unfolded a larger truth about academic integrity in the AI era. The investigation that followed revealed a troubling reliance on generative AI tools among students, particularly in high-pressure courses where performance metrics often overshadow genuine learning.

Unusual exam results spark scrutiny

Roberto Serrano, a blind economics professor teaching an advanced microeconomics course, observed a pattern in student submissions that raised red flags. The exam scores were not just high—they were consistently flawless, with answers that appeared to replicate the precise phrasing found in AI-generated outputs. Serrano, known for his meticulous grading standards, decided to investigate further, suspecting that students were circumventing the learning process by outsourcing their work to large language models.

The professor’s suspicions were not unfounded. In a recent survey of Princeton students, nearly 30 percent admitted to using AI tools on at least one exam or assignment. While the Princeton data provided a broader context, Serrano’s case offered a granular look at how academic dishonesty manifests in real time—within a single classroom, under the radar of traditional detection methods.

The mechanics of AI-assisted cheating

Students leveraging AI for academic work often adopt a systematic approach to bypass plagiarism detection and instructor scrutiny. Common tactics include:

• - Using AI platforms to generate entire essay responses or problem-solving steps, then lightly editing the output to mimic personal writing styles.
• - Inputting exam questions directly into AI tools to receive step-by-step solutions, especially in quantitative subjects like economics.
• - Collaborating in private chat groups to refine AI-generated answers before submission, reducing detectable anomalies in tone or structure.

Serrano’s investigation highlighted how AI tools can produce answers that align perfectly with grading rubrics, making it difficult for instructors to differentiate between original work and machine-generated content. The professor’s determination to uncover the truth underscores a growing challenge for educators: balancing technological innovation with the preservation of academic rigor.

Institutional responses and ethical dilemmas

Brown University’s administration has launched an internal review to assess the scope of the issue and determine appropriate disciplinary actions. The case raises broader questions about how institutions should adapt to the proliferation of AI tools in education. Should universities invest in advanced detection software, revise academic honor codes, or redesign assessments to minimize AI’s influence?

The dilemma extends beyond Brown. Elite institutions, where academic pressure is intense and competition is fierce, are particularly vulnerable to such shortcuts. Students may justify AI use as a pragmatic solution to overwhelming workloads, but the long-term consequences—erosion of critical thinking, diminished skill retention, and compromised personal growth—are becoming impossible to ignore.

A call for proactive measures

Serrano’s decision to expose the issue publicly reflects a growing sentiment among educators: the need for proactive strategies to combat AI-assisted cheating. Some potential solutions include:

• - Implementing in-class, open-book exams with randomized questions to reduce the effectiveness of pre-generated AI responses.
• - Requiring students to submit drafts or outlines alongside final submissions to demonstrate their engagement with the material.
• - Incorporating oral assessments or project-based evaluations that are harder to replicate with AI.
• - Educating students on the ethical implications of AI use and the importance of intellectual integrity.

The scandal at Brown serves as a cautionary tale for higher education. As AI tools become more sophisticated and accessible, the line between assistance and academic dishonesty continues to blur. Institutions must act swiftly to redefine their policies, not just to punish misconduct, but to foster a culture of genuine learning and accountability.