Oral assessment questions are most useful when they ask students to explain decisions in their own submitted work. A student can describe why a source was selected, demonstrate a calculation or reconsider a recommendation after a constraint changes. These exchanges give an assessor evidence of understanding that polished prose alone may not provide.

The purpose is to assess learning against published outcomes. An oral follow-up should not become an improvised interrogation triggered by a lecturer’s suspicion about writing style. Plan the format, prepare students and assess comparable evidence across the cohort.

Consider the illustrative case of a master’s student who submits an AI-assisted project proposal. The proposal recommends introducing an online appointment system. The following questions show how an assessor could examine the student’s reasoning without asking them to memorise the report.

Contents  Begin with the outcome being assessed  •  1 Explain one consequential choice  •  2 Locate the supporting evidence  •  3 Reconstruct a calculation  •  4 Change one constraint  •  5 Examine a rejected alternative  •  6 Discuss the AI contribution precisely  •  7 Identify a limitation that changes the conclusion  •  8 Transfer the reasoning to another setting  •  Build a fair plan for oral assessment questions  •  Use follow up prompts consistently  •  Record evidence rather than impressions

Begin with the outcome being assessed

Identify what the student must demonstrate. If the outcome concerns project planning, ask about dependencies, estimates and choices. If it concerns ethical analysis, examine stakeholders, competing duties and consequences. General questions about AI cannot substitute for the subject knowledge the assessment is supposed to measure.

TEQSA’s assessment reform principles provide sector guidance for considering assessment in the presence of generative AI. They are not a prescribed script for an oral examination. The question types below are practical design suggestions to adapt through your institution’s assessment procedures.

A useful preparation step is to annotate the assessment brief with three claims: what the written submission shows, what remains uncertain and what an oral exchange can reasonably clarify. This keeps the follow-up focused and avoids testing unrelated knowledge.

1 Explain one consequential choice

Ask: why did you recommend an appointment system instead of extending reception hours?

The student should explain the criteria used, the evidence available and the reason one option was preferred. A stronger response might acknowledge that extended hours could improve access but would increase staffing costs and would not address duplicate bookings.

Follow with: which piece of evidence most influenced your choice? This checks whether the explanation connects to the submission. It also gives a student who begins with a broad answer an opportunity to become specific.

Avoid treating agreement with the assessor as the standard. Two students may defend different recommendations convincingly when their assumptions are explicit and their reasoning is coherent.

2 Locate the supporting evidence

Ask: show where the report supports the claim that missed appointments are the main source of delay.

Allow the student to use the submitted document. The aim is to evaluate the relationship between evidence and claim, not navigation speed. If the student identifies a survey of staff perceptions, ask what that source can and cannot establish.

An answer that distinguishes perceptions from appointment records demonstrates analytical awareness. A confident claim that the survey proves a causal relationship may reveal a substantive weakness even when the report is fluently written.

Where appropriate, consult Cornell’s guidance on selecting assessment methods for learning outcomes. The format should serve the kind of learning you need to observe.

3 Reconstruct a calculation

Ask: take me through how you estimated the first-year cost.

The student might identify software fees, configuration, staff training and ongoing support. Ask them to distinguish a one-off implementation cost from a recurring annual cost. If the proposal assumes staff time has no cost, invite them to explain that assumption.

Provide a calculator if calculation is relevant and the original task allowed one. Do not turn a project-planning assessment into an unannounced mental arithmetic test. What matters is whether the student understands the quantities, units and relationships.

If the student finds an error during the explanation, assess the demonstrated learning according to the rubric. Recognising and correcting an error may provide useful evidence, but it does not automatically erase weaknesses in the original submission.

4 Change one constraint

Ask: the budget has been reduced by 20 per cent; what would you change first, and why?

This tests whether the student can use the underlying principles in a changed situation. A thoughtful answer may protect essential booking functions while postponing an optional reporting dashboard. Ask what risk the change introduces and whether it alters the expected benefit.

Choose a change that is demanding but manageable within the discussion time. Introducing several new constraints simultaneously can make performance depend more on speed and stress tolerance than on the intended learning.

Use a consistent level of challenge across students. Equivalent scenarios need not be identical, but their demands should be comparable and justified.

5 Examine a rejected alternative

Ask: what is the strongest argument against your recommendation?

A weak response may simply repeat the proposal’s benefits. A stronger response identifies implementation disruption, unequal digital access or dependence on a supplier, then explains how these affect the recommendation.

The follow-up question is: what evidence would make you choose the other option? This reveals whether the student’s position can respond to evidence. It also discourages the idea that assessment rewards defending an initial answer at any cost.

For a related discussion of decision reasoning, see The Case HQ’s executive decision-making guide. Use the principles to sharpen questions, while keeping the actual assessment aligned with the course outcome.

6 Discuss the AI contribution precisely

Ask: identify one suggestion from the AI tool that you changed or rejected, and explain your decision.

A student might show that the tool proposed a three-week implementation schedule without allowing time for procurement. The important evidence is the student’s evaluation, not how sophisticated the original prompt sounds.

Use the disclosure requirements communicated in advance. Do not retrospectively demand a complete chat history if students were only asked to provide a brief use statement. Logs may also contain personal or confidential information that is unnecessary for the assessment.

If AI use was prohibited, follow the institution’s integrity procedures when concerns arise. An assessment conversation and a misconduct investigation have different purposes and protections; avoid blending them informally.

7 Identify a limitation that changes the conclusion

Ask: which limitation in your evidence most affects how confidently the organisation should proceed?

For the appointment-system proposal, the student may note that data came only from weekday users. That matters if evening demand is central to the decision. Ask how the recommendation should be qualified and what further evidence would be proportionate.

Reward explanations of consequences. Simply listing small sample size or limited time does not demonstrate why the limitation matters. The student should connect it to a particular inference, decision or population.

Do not expect every limitation to invalidate the project. Judgement includes recognising which uncertainty can be managed and which requires a different design.

8 Transfer the reasoning to another setting

Ask: how would your proposal change for a service used mainly by people with limited internet access?

The answer should reconsider assumptions, rather than replace one organisation’s name with another. A student may propose assisted booking, telephone access or a staged pilot. Ask which elements of the original analysis remain useful and which need new evidence.

This question is especially valuable when the outcome includes application or evaluation. It provides a clearer test of transfer than asking students to repeat a definition they already included in the report.

The University of Sydney’s explanation of its two-lane assessment approach offers one institutional example of combining different assessment conditions. Adapt principles thoughtfully rather than importing another institution’s rules unchanged.

Build a fair plan for oral assessment questions

For a ten-minute follow-up, select a small set of questions that sample the intended outcomes. For example, allow a brief orientation, examine one decision and its evidence, change one constraint, then ask the student to clarify a limitation. Do not attempt all eight question types in every conversation.

Publish the format, duration, broad question categories and marking criteria beforehand. Give students a low-stakes opportunity to practise. Explain whether they may consult their submission and how technical disruptions will be handled.

Plan reasonable adjustments through established support arrangements. Students’ anxiety, disability, language background or processing speed should not be mistaken for lack of subject understanding. If spoken communication is itself assessed, make that criterion explicit and distinguish it from analytical quality.

Use follow up prompts consistently

Agree a small set of neutral prompts, such as explain that step or identify the evidence you used. These help a student clarify an answer without supplying the missing reasoning. Record substantial prompts so moderation can consider how much assistance was needed. A prepared question bank is useful, but assessors should still listen to the particular explanation rather than rush through a checklist.

Record evidence rather than impressions

Use a compact assessment record with the outcome, question asked, evidence observed and rubric judgement. Notes such as explained recurring costs but omitted staff training are more useful than seemed confident.

Calibrate assessors using sample responses. Agree what counts as an adequate explanation, what follow-up prompts are permissible and how a student can recover from an unclear first answer. Review borderline decisions and unexpected patterns across assessors.

For further development, The Case HQ’s case-based assessment article discusses evaluating applied work. Educators seeking structured professional learning can explore the paid PGCert in AI in Teaching and Learning.

A well-designed oral follow-up leaves a clear record of what the student understands. Start with a small number of outcome-aligned questions, allow explanation and judge the reasoning shown in the exchange.