Oral Examination and Viva

Ten minutes of questions about a submitted piece establishes authorship more reliably than any monitoring of the room it was written in. Scheduling and recording sampled vivas can follow the scheduling page as an operations reference, provided performance is judged against the work rather than speaking style.

What it addresses

Contract cheating, which proctoring does not touch at all and which is the dominant integrity problem for coursework.

Understanding rather than production: a student who did not do the work cannot discuss the choices in it.

And identity, incidentally, since the person in the conversation is the person answering.

Where it fits

As a check on a sample of submissions rather than all of them.

As a required component of a final or capstone piece.

As a follow-up where something about a submission raises a question, which is a far better response than a surveillance-based investigation.

And for small cohorts, universally, where the numbers make it feasible.

The cost

Staff time, which is the real constraint: ten to fifteen minutes per student plus scheduling.

For a cohort of three hundred that is a substantial commitment.

Which is why sampling matters: a stated random sample of ten or twenty percent achieves most of the deterrent effect at a fraction of the cost.

Announce that sampling happens and that selection is random.

Fairness considerations

Oral assessment disadvantages some students: anxiety, speech differences, second-language candidates, some neurodivergent patterns.

Which means adjustments apply here as everywhere — extra time, written follow-up, a familiar examiner, questions in advance.

And the criteria must be about the work, not about fluency, which needs stating in the rubric and in examiner training.

Running it well

Questions about specific choices in the submission, not general knowledge.

"Why did you use this method here" rather than "explain this concept".

Two examiners where a case might follow.

Recorded, with consent, where the outcome could be contested.

And a short written record of what was asked and answered.

What it cannot establish alone

Poor performance in a viva is not proof that the work was not the student's.

Nerves, language, and the format itself produce weak performances from honest students.

Which makes a viva a strong indicator and not a finding, and the evidence hierarchy applies here as elsewhere.

The deterrent effect

A cohort that knows a random sample will be examined orally behaves differently at the point of deciding whether to outsource.

This is probably the largest part of the benefit and it requires only that the sampling is real and announced.

What to check

Is any part of your assessment defended orally?

Would a sample be feasible, and has anybody costed it?

Are adjustments defined for oral assessment specifically?

And is it clear to examiners that a nervous performance is not evidence of misconduct?

The point

A cohort that knows a random sample will be examined orally behaves differently at the point of deciding whether to outsource.

That deterrent is probably the larger part of the benefit.

Worth stating

Oral assessment disadvantages some students — anxiety, speech differences, second-language candidates — so adjustments apply here as everywhere, and the criteria must be about the work rather than about fluency..

Also worth knowing

Sample rather than examine everybody: a stated random sample of ten or twenty percent achieves most of the deterrent effect at a fraction of the cost.

Announce that sampling happens and that selection is random.

And finally

Ask about specific choices in the submission rather than general knowledge: why did you use this method here.

Two examiners where a case might follow, and a short written record of what was asked and answered.

Summary

A sampled viva addresses contract cheating, which no monitoring reaches, and most of the benefit comes from the sampling being announced rather than from the examination itself.

In summary

Ten minutes of questions about a submission establishes authorship more reliably than any monitoring of the room it was written in, and it addresses the channel proctoring cannot reach at all.. For wider institutional context, consult the Association for Computing Machinery.