Saving the Take-Home Essay With Oral Assessments (opinion)

Over the last few years, countless commentators have confidently declared the take-home essay dead. These commentators point out that students are increasingly using large language models to write their essays and that attempts to police this with detection or surveillance have failed.

We agree that policing artificial intelligence use is unworkable. However, we reject the assumption that this is the only way to preserve the integrity of the take-home essay, which still has important pedagogical value. Instead, we advocate pairing take-home essays with short oral exams in which students must demonstrate that they can take full intellectual responsibility for the work they submit.

The appeal of the oral exam is that it doesn’t require the instructor to play cop. If a student cannot adequately explain their own essay, they are simply marked down for their failure of understanding. There is no need for any detective work to figure out whether and how AI was used. Furthermore, students are disincentivized from outsourcing their work to AI, since weak performance on the oral exam carries clear costs.

The rationale is straightforward. But how does it work in practice?

Before the Exam

While the oral exam itself works as a disincentive, there are other things that instructors can do from the beginning of the semester to reduce the likelihood that students will misuse AI.

• First, communicate purpose. Students are more likely to outsource work when they see assignments as nothing more than hoops to jump through. If you assign an essay, explain its value beyond its weight in the course. What intellectual skills is it meant to develop? Why retain a take-home essay rather than replace it with a secured exam? Making this explicit helps students see the assignment as worth doing.

• Second, have a conversation about AI. Once the purpose of the assignment is clear, discuss how AI might undermine, or support, that purpose. Opinions about the appropriate use of AI will likely differ given the lack of agreed-upon best practices among instructors. But, at a minimum, talking to students about their AI usage gets them reflecting critically on it from the outset.
• Third, scaffold the assignment by providing intermediate deadlines and feedback. While scaffolding should be a given in the design of any assignment, it becomes more essential in the face of AI. In its absence, students who lack confidence, foundational skills and social support are more likely to turn to AI for help. For instance, require students to submit a one-page description of their ideas several weeks before the deadline, and provide actionable feedback.

The Oral Exam

Your students have now submitted their essays. What next?

Recall that the essay is still the main assignment, and the oral exam is only meant to safeguard the integrity of the essay, not to test how eloquently students speak or how quickly they can think on their feet.

For students, this implies that as long as they have taken ownership of their work, there should be no further preparation required.

For instructors, it means that you should give the essay a provisional grade before the oral exam and only mark the student down if it is evident they do not have mastery over their own work. While reading each essay, write down questions that enable you to assess whether students have been deeply involved in developing the work. For argumentative essays, you might ask: Why did you frame the problem this way rather than that way? What’s your response to this potential objection? How does your position relate to the broader scholarly conversation? For empirical research papers, you might ask: How did you choose and refine your research question? What is your understanding of this concept, and how is it operationalized in your data? Two or three core questions, plus follow-ups as necessary, should be plenty for each paper.

As for the oral exams, in our experience, 10 to 20 minutes per meeting is sufficient. Rather than being interrogation sessions, we have found these meetings to be genuine intellectual conversations. We take a real interest in understanding our students’ work more deeply, and both instructor and student learn something in the process. Indeed, in a survey conducted by one of us, students had very favorable reviews of the oral assessment. They valued the chance to discuss a project they had invested in, receive verbal feedback and think about their papers in new ways.

Common Concerns

You are probably doing the mental math: Assuming a class of 40 students, this would take between seven and 13 hours. And that’s not counting the time it takes to read and grade the papers. Who can afford that?

For instructors with heavy teaching loads, one idea is to make room for the oral exams in your syllabus. This could mean reducing one week’s worth of content to free up time for conducting the exams. Or it could involve eliminating another assessment component and spending the time that would have been required to grade that assessment on oral exams instead. Essentially, this amounts to trading off the quantity of content or assignments to enhance the quality of the take-home essay.

Beyond what faculty members can do on their own, universities must think about whether larger changes are needed to accommodate the kinds of high-touch approaches to instruction and assessment that the growing presence of AI calls for. Oral exams, in-class discussions, even office hours—all of these would help ensure students are taking intellectual responsibility for their work. But they are only possible under certain structural conditions, such as lighter teaching loads or smaller class sizes.

A second concern is that the student who outsources their essay writing to AI might also use it to anticipate possible oral exam questions and memorize scripted responses. This worry strikes us as far-fetched. First, there are too many possible questions for a student to prepare for all of them. Second, memorized responses sound different from spontaneous thinking. Third, if a student did manage to prepare so thoroughly that they could demonstrate genuine understanding, they would have engaged in exactly the kind of deep learning that the assignment was designed to produce.

The Value of Oral Exams

Oral exams clearly take work. Why take them on, when you could choose the easier alternative of in-class exams? Because they help protect two of the fundamental elements of a university education that are at risk as the presence of AI grows.

First, intellectual effort and ownership. In-class exams mostly reward speed, memory and performance under pressure. In contrast, take-home essays give students the space to think deeply and critically about a topic and the time to draft, iterate and polish. In the process, they develop precisely the kinds of higher-order, independent thinking skills they will need to thrive in a world where many tasks are being outsourced to AI.

Second, human relationships. As chatbot tutors and AI-powered grading tools wend their way into the classroom, they risk squeezing out the moments of connection that motivate students and help them feel seen. Oral exams create space for conversation. They allow instructors to engage directly with students’ ideas and for students to feel that their work is taken seriously. That sense of recognition can motivate effort in ways no technological substitute can.

The take-home essay does not need to be abandoned. With adjustments, it can be adapted to a new technological landscape—one that demands not less, but more, emphasis on understanding, responsibility and dialogue.