The rapid rise of generative artificial intelligence (AI) tools has revealed serious weaknesses in traditional academic assessment. The issue became clear to me after I found evidence that many students had used AI without authorization during an examination for my welfare economics and social-choice theory course at Brown University in Providence, Rhode Island, last semester.
Because some students were hesitant to return to campus after a shooting earlier in the academic year, I offered the midterm examination as a take-home test for the first time in almost 20 years of teaching the course. Although the exam was more challenging than previous assessments, the average score reached 96% — significantly higher than the course’s historical average of approximately 65–80%.
Concerned that students had used AI to cheat, I administered the final exam in person under closed-book conditions. The average score fell sharply to 48.6%. Several students also withdrew from the course or failed to attend the final examination.
Why universities need to radically rethink exams in the age of AI
The most troubling part of this experience was not simply that cheating may have taken place during a take-home exam, but the apparent scale of unauthorized AI use. Universities have struggled to develop effective responses to generative AI in education. However, they can no longer avoid the issue: the credibility of academic learning, assessment and university qualifications is at stake.
Two priorities require immediate action: rebuilding a strong culture of academic integrity and redesigning teaching, learning and assessment for the age of AI.
Academic misconduct remains widespread at universities partly because institutional leaders are often reluctant to address it decisively. When administrators fail to defend academic integrity clearly and consistently, they risk creating an environment in which cheating becomes normalized. My experience — along with numerous messages from faculty at other universities — suggests that administrators may respond to evidence of widespread AI-assisted cheating with hesitation or denial.
In some cases, this reluctance appears to reflect concerns about protecting an institution’s reputation and “market value”. But that approach overlooks what truly determines a university’s long-term standing: rigorous academic standards, a culture of hard work and authentic educational achievement. Universities that want to preserve trust must confront AI-enabled cheating openly and develop assessment systems that reward genuine learning.
AI can supercharge inequality — unless the public learns to control it
Source: www.nature.com