This eSchool News opinion piece by Nesreen El-Baz, an ESL educator with over 20 years of experience, looks at why oral exams are coming back. It asks what should count as evidence of learning now that generative AI can produce written work.
El-Baz opens with Catherine Hartmann, a University of Wyoming professor who in 2023 asked students in a history of meditation course to try a contemplative practice and write a reflective analysis. One student used AI and left the pasted prompt at the top of the paper. For Hartmann, the problem went beyond one shortcut. Take-home writing could now be generated or heavily shaped by AI in seconds, and teachers risked becoming detectives hunting for machine authorship. She rebuilt her upper-level humanities course around oral assessment, using backward design so students discussed ideas, answered questions and practiced explaining their thinking all term. The final oral exam grew out of that practice. El-Baz notes the shift reaches beyond the humanities, citing University of Pennsylvania mathematics professor Robin Pemantle, who has students work calculus problems at the board while explaining their reasoning aloud.
The author argues that if catching cheaters becomes the main aim, schools would swap one form of surveillance for another. In her view, an essay shows an outcome, while a conversation gives access to the thinking behind it, letting teachers hear misconceptions as they emerge and separate memorized answers from usable knowledge. Students might still use AI to brainstorm or prepare. The question she poses is whether a student can take intellectual ownership of the work.
She also raises cautions. Not every student is equally ready to think aloud under pressure, and for multilingual learners, hesitation, accent, vocabulary retrieval or processing time shouldn't be read as missing subject knowledge. Oral assessment needs low-stakes, scaffolded practice long before exam day. El-Baz says written work keeps its place, and she calls the effort to build AI-proof assessment a race education is unlikely to win. She concludes that AI should prompt educators to widen what counts as evidence of learning.