My humanities department wants to ban detectors and go oral exams only, is that realistic?

I have TA'd for a course that used short orals on a subset of students. It works, but not as "oral for everyone". We sampled: every student knew they might be called, a random selection was, and the conversation was ten minutes on their own paper. Because anyone could be selected, the effect on behaviour was broad without the cost.

The arithmetic matters. Twenty minutes for 120 students is forty hours before any scheduling or marking time. That is the entire problem with "everyone, every essay".

@Priya N. sampling is the piece I did not have. Random selection with everyone aware they might be called gives most of the effect at a fraction of the hours, and it is easier to defend as even-handed. I will take that, plus the separate-motions point from Hana, to the department meeting.

I want to speak for the anxious and the non-native speakers. For me, a spoken exam in English is much harder than writing, and my ideas are better on paper. If this goes ahead, please do not treat hesitation or accent as a signal that I did not write my own essay. And allow a student to bring their notes and annotated draft. Being nervous is not the same as not knowing.

Oral exams are cheat-proof. If you can talk about it, you wrote it. That is the whole point and why universities used them for centuries. Just do it and stop fretting.

On process, the weak point is the record. A written essay is an artefact you can re-read on appeal. A conversation is gone unless you record it, and recording brings data protection and consent questions. If you adopt orals, decide in advance what is kept (a recording, a marker's notes against a rubric), who may see it, and how a student can challenge the result. Without that, you replace an unexplained score with an unrecorded impression, which is arguably worse.

Also the accommodations point Aisha makes is a legal duty in many places, not a courtesy.

Secondary school, so we do this informally all the time: I ask a student to explain a paragraph and it settles most of what a detector cannot. It is not a formal examination, though. It is a conversation to understand the work.

If your department formalises it, I would keep the oral short, about the student's own paper, and use it to confirm understanding rather than to grade the speaking. Those are different skills.

One thing worth separating: banning detectors and adopting orals are two separate decisions, and the first does not depend on the second. You can decide that a score is not admissible evidence, as several institutions have, without committing to a new exam format for everyone. I would put those to your department as separate motions, because the second is the expensive and contested one.

@Lena B. I would not say "cheat-proof". A student can memorise an AI-written essay and rehearse the likely questions, as Tobias already noted, and a fluent speaker can bluff more convincingly than a careful one can explain. A viva tests understanding better than a score does, and it is harder to fake than text, but harder is not impossible. The fair claim is "much stronger evidence", and the design should assume some gaming, for example by following up with unscripted questions on the sources.

@Hana Sato fair. "Cheat-proof" was too strong. Harder to fake and easier to defend is the right bar, and a student can still memorise an essay. The unscripted follow-up on sources is the part that actually catches it.