Whenever oral assessment comes up, someone raises fairness, and they are usually met with enthusiasm about how much better conversation is at measuring understanding. That is not an answer. The objection is specific and it deserves specifics back.
The objection, stated properly
Speaking under evaluation is harder for some people than writing, in ways that have nothing to do with what they know. Anxiety disorders. Stammers and other speech differences. Autistic students who process a question differently or need longer to answer. Students speaking their second or third language. Students with no quiet room at home.
If the format converts any of those into a lower mark, the format is measuring the wrong thing, and it does not matter how good it is at measuring understanding for everyone else.
What design fixes
A surprising amount of it, honestly.
- Time pressure is a choice. Most of the anxiety load in oral exams comes from a clock, not from speaking. Remove the clock for the students who need it removed and much of the gap closes.
- Grounding removes memorisation. An exam about your own submission, with the submission in front of you, is a different psychological event from an exam about the syllabus. There is nothing to blank on.
- A no-stakes first run matters more than anything else. Nearly all of the reported dread is about not knowing what the format feels like. One practice run with nothing riding on it removes most of it.
- Grade content, not delivery. Fluency, accent, hesitation and pace should carry zero weight. This has to be enforced in the rubric and, if software is involved, in the model's instructions, and it has to be checkable.
The question is whether the student understands their work. Not whether they perform well while being asked.
What design does not fix
Some students will still need an alternative, and pretending otherwise is how accessibility policy becomes a paperwork exercise. A written or instructor-run alternative should exist, should be available without a diagnosis in hand for the low-stakes cases, and should not appear in any record as a lesser version of the assessment.
One detail that gets missed: other instructors should not be able to see which accommodation a student has. Knowing that adjustments apply is enough for teaching. Knowing which one is that student's medical information moving around a department.
The comparison that is usually skipped
The fairness discussion tends to compare oral assessment against an idealised written exam. It is worth comparing it against the real one. Timed written exams under invigilation are their own accessibility problem, and unsupervised take-home essays now advantage students who are willing to outsource them, which is a fairness cost too, just one that falls on the honest.
There is no format without a burden. The question is whether the burden is on something you can design around, and for oral assessment, most of it is.
Our own rules for this are public at the Viva standard, including the ones we would be embarrassed to break.