Bring a brief and someone to review it.
Oral assessment can show how learners explain evidence and respond to questions. Scaling it responsibly requires clear criteria, consistent assessor preparation, accessibility, and human review for consequential decisions; an automated score should not silently become the final judgement.
- A clear assessment purpose
- Observable criteria and sample responses
- Accessibility, privacy and appeal arrangements
5 moves to practise
Define what the assessment measures
Separate subject knowledge, reasoning, communication, and other constructs. Make sure each prompt gives learners a fair opportunity to demonstrate the stated outcome.
Use observable criteria and anchor examples
Describe evidence assessors can hear or see. Prepare sample responses for calibration and review differences across assessors before relying on results.
Keep humans accountable
If a tool offers feedback or a provisional score, explain its role and limitations. Provide a way to challenge errors and keep a qualified person responsible for grades, progression, or other consequential decisions.
Plan accessible and private conditions
Tell learners what is recorded, who can access it, how long it is retained, and how it affects assessment. Consider adjustments and equivalent formats. Do not infer competence from accent or eye contact alone.
Evaluate the assessment process
Review completion, assessor agreement, accessibility issues, appeals, and feedback quality. Combine oral evidence with other appropriate work and report limits transparently.
Try it. Review it. Change one thing.
Your practice task
Have two qualified reviewers assess the same sample, discuss disagreements and revise unclear criteria.
What to look for
Check whether the task measures the intended knowledge or reasoning, and document how consequential decisions are reviewed.
Common questions
Does oral assessment make an exam AI-proof?
No assessment method can support an absolute AI-proof claim. Design tasks that let learners explain reasoning and use appropriate human moderation.
Should an AI score be used as the final grade?
Do not assume so. Review validity, policy, accessibility, and appeals, and keep an accountable human involved in consequential decisions.
How can departments check fairness?
Use clear criteria, assessor calibration, accessible conditions, review routes, and appropriate analysis of outcomes across learner groups.
Research and further reading
Use these primary resources to explore the method and its context. References describe their publishers’ work; they do not imply endorsement or certification of Teachrity.
- Cornell: designing oral assessments
Practical assessment-design considerations from Cornell’s Center for Teaching Innovation.
- AACSB: assurance of learning
Evidence and improvement within AACSB’s own accreditation framework; using software does not confer accreditation.
Explore related questions
Continue with practical questions for your course, role or learning goal.
Why might essay-only assessment become less informative as students use AI tools?
When a tool can help produce a polished submission, the final text may reveal less about the learner's own reasoning process. Clarify permitted assistance and collect additional evidence through explanation, revision or application to changed facts. The response should improve assessment design rather than presume every student is dishonest.
Explore the full topicHow can we assess understanding without claiming an assessment is AI-proof?
Choose tasks that require students to explain their reasoning, apply it to unfamiliar conditions and respond to questions. Use several evidence sources and make assistance rules explicit. Acknowledge residual risks, provide accessible alternatives and keep accountable human judgment over consequential outcomes instead of promising perfect protection.
Explore the full topicHow should universities teach students to question AI-generated feedback?
Ask students to compare feedback with the rubric and the evidence in their response. They should identify an unclear point, test a suggested improvement and explain disagreement respectfully. Provide a review route and discuss examples where automated feedback is helpful, incomplete or mistaken.
Explore the full topicAssessment requirements vary by programme and jurisdiction. Obtain academic, accessibility, privacy, and legal review for consequential uses.