Perfect homework, blank stares

Jocelyn Gecker,
MDT/AP Education Writer

The assignment involves no laptop, no chatbot and no technology of any kind. In fact, there’s no pen or paper, either. Instead, students in Chris Schaffer’s biomedical engineering class at Cornell University are required to speak directly to an instructor in what he calls an “oral defense.”

It’s a testing method as old as Socrates and making a comeback in the AI age. A growing number of professors say they are turning to oral exams, and combining old-fashioned and cutting-edge techniques, to address a crisis in higher education.

“You won’t be able to AI your way through an oral exam,” says Schaffer.

Educators are no longer wondering if students will use generative AI to do their homework. The question now is how to determine what students are actually learning.

Across universities, instructors are noticing a troubling pattern. Written assignments are coming back flawless. But when students are asked to explain their work, they can’t. The long-term impact of AI on critical thinking remains uncertain, but many worry students increasingly see the hard work of thinking as optional.

At the University of Pennsylvania, Emily Hammer pairs oral exams with written papers. “It comes across as if we’re trying to prevent cheating,” she says. “That’s not why we’re doing this. We’re doing this because students are losing skills, losing cognitive capacity and creativity.”

Hammer forbids AI use on writing assignments but acknowledges she cannot fully enforce it. If students have not done their own work, defending it face-to-face will likely be stressful. Her class reflects what others describe as a broader shift toward in-person assessments.

Oral exams are not traditionally part of many undergraduate systems, though they are common in parts of Europe, where close discussion between students and tutors is embedded in academic culture. Interest grew during the pandemic amid concerns about online cheating and has intensified since tools like ChatGPT.

Some educators are also experimenting with using AI itself to test students. At New York University, Panos Ipeirotis developed an AI-powered oral exam in which a voice-based chatbot questions students about their work. He describes it as “fighting fire with fire.”

The goal is simple: to verify whether students understand what they submit, or whether they have outsourced the thinking. Ipeirotis plans to pair oral exams with written assignments. “I don’t trust written assignments anymore to be the result of actual thinking,” he says.

Student reactions are mixed. Some find the technology awkward, but many acknowledge a broader reality: there is no version of education in which AI exists and is not used, or misused.

Across disciplines, educators worry that students who bypass the mental struggle required for problem-solving will not develop the skills needed for advanced study or work. That concern is driving renewed interest in formats that require explanation, reasoning and presence.

In Schaffer’s class, students complete written problem sets and then defend them in 20-minute, Socratic-style sessions. He no longer grades the written work, only the oral defenses, incentivizing students to understand what they submit.

Variations are emerging elsewhere: final conversations instead of exams, short mock interviews, or one-on-one discussions. While some argue oral exams can be stressful, instructors say preparation can mitigate those concerns.

In an environment where polished answers are easy to generate, the ability to think aloud, defend ideas and respond under scrutiny is becoming harder to fake. That, increasingly, is the point.