Stanford professor returns to handwritten exams…at students’ request

Stanford computer science professor Jure Leskovec made a surprising pivot two years ago, switching from open-book, take-home exams to handwritten, in-person tests in response to the rise of AI tools like GPT-3. The change came at the request of his students and teaching assistants, who wanted a way to genuinely assess knowledge without AI assistance, highlighting the complex challenges educators face as artificial intelligence reshapes academic evaluation.

What happened: Leskovec, a machine learning researcher with nearly three decades of experience, found himself grappling with an “existential crisis” among students when GPT-3 launched publicly.

• Students questioned their role in a world where AI seemed capable of doing research independently, leading to deep conversations about their future careers and contributions.
• The solution emerged from students themselves, particularly teaching assistants, who proposed returning to traditional paper exams.
• “We had a big, I don’t know, existential crisis among students a few years back when it kind of wasn’t clear what our role is in this world,” Leskovec said.

The big picture: The shift represents a broader debate in higher education about how to adapt assessment methods in the age of AI.

• Many colleges are either banning AI outright or returning to traditional testing methods, with some professors suggesting “medieval” approaches like oral examinations.
• Leskovec’s 400-person classes now require significantly more grading work, but he insists on hand-grading rather than using AI assistance.
• “No, no, no, we hand grade,” he emphasized, despite the increased workload.

The calculator comparison: Leskovec frames AI as a powerful but imperfect tool that requires careful integration into education.

• “Are you worried about students cheating with calculators? It’s like if you allow a calculator in your math exam, and you will have a different exam if you say calculators are disallowed,” he said.
• He believes educators need to test both students’ ability to use AI tools and their capacity to think independently without them.
• The challenge lies in distinguishing between human skills and AI skills, and understanding where they intersect.

Labor market implications: The AI skills gap is creating complex dynamics in hiring and workforce development.

• MIT studies show 95% of generative AI pilots are failing, while Stanford research indicates a collapse in entry-level hiring for AI-exposed jobs.
• Upwork’s hiring report reveals 40% growth in demand for AI and machine learning skills among small and medium businesses, alongside increased need for human fact-checkers.
• “We’re actually seeing the human skills coming into premium,” said Kelly Monahan, managing director of the Upwork Research Institute.

What experts are saying: Domain expertise is becoming increasingly valuable as AI tools require human oversight and verification.

• “I think we almost need to re-skill the workforce. Human expertise matters much more than it ever did [before],” Leskovec said.
• Monahan noted that without advanced domain knowledge, “it’s easy to be fooled” by AI-generated content, making specialized human skills more crucial.
• The entry-level challenge remains: how young workers can develop the expertise needed to effectively collaborate with AI.

Looking ahead: Leskovec believes we’re still in the early stages of learning how to integrate AI into education and work.

• “I think we’re in the ‘coming-up-with-solutions phase,'” he said, suggesting current approaches like hand-graded exams are experimental responses to unprecedented challenges.
• The focus is shifting toward reskilling curricula and finding new ways to validate human knowledge alongside AI capabilities.
• Organizations must balance leveraging AI tools while ensuring they’re training the next generation of skilled workers.