INTEGRATIONS · CANVAS
Are Oral Exams the AI-Proof Assessment? Running Consistent Vivas in Canvas
Oral is harder to fake. It is not automatically fairer. The hard part is applying the same standard at nine in the morning and at four in the afternoon.
By Eduface · July 2026 · 9 min read
Oral exams are widely being pitched as the AI-proof answer to written work AI can now produce convincingly. The claim is broadly right, but incomplete: an oral exam only holds up as a fair, defensible assessment if it’s consistent, not just spoken. Ask any programme lead who has run vivas at scale about the hardest part, and it’s rarely finding good questions. It’s making sure the same standard applies to the student examined at nine in the morning by an examiner in a good mood and the student examined at four in the afternoon by the same examiner, tired, on their sixth viva of the day. That inconsistency is one of the oldest, most stubborn problems in oral assessment, and it’s worth understanding where the field has already tried to solve it before looking at how Canvas and AI change the picture.
Quick answer
Oral exams are much harder to outsource to AI than written work, but “oral” on its own does not make an assessment fair. Consistency does. Oral Exam mode applies the lecturer’s own criteria to every student identically, drafts a criterion-by-criterion grade, and syncs to the Canvas gradebook only after the lecturer reviews and approves it.
The problem medicine solved decades before anyone else took it seriously
Oral and practical assessment has a genuinely long history of examiner variability undermining its validity, and nowhere was this taken more seriously, earlier, than in medical education. In 1979, Ronald Harden and Fionnuala Gleeson published a paper describing what they called the Objective Structured Clinical Examination, now universally known as the OSCE, designed specifically to address the fact that traditional oral and practical exams were producing wildly inconsistent results depending on which examiner a candidate happened to draw (Harden & Gleeson, 1979). Their solution was structural: standardised stations, a fixed checklist of what counted as a correct response, and the same criteria applied to every candidate regardless of who was assessing them. The OSCE didn’t eliminate human examiners. It eliminated the variability that came from leaving the criteria implicit in each examiner’s head.
That’s worth sitting with, because it’s essentially the same problem business and management programmes running vivas, case interviews, or oral defences face today, more than four decades later, usually without anything like the same structural fix.
What criterion-graded AI oral exams do differently
Eduface’s Oral Exam mode applies an examiner’s own criteria to every student’s response, live, during the exam itself, and grades consistently against those criteria regardless of time of day, examiner fatigue, or which cohort a student happens to be in. That’s the OSCE principle applied to a live, conversational oral exam rather than a series of fixed stations, and it addresses the same underlying weakness Harden and Gleeson identified: variability isn’t eliminated by finding better examiners, it’s eliminated by making the criteria explicit and applying them identically every time.
Every grade is criterion-by-criterion, with an explanation for each one, and, crucially, a lecturer reviews and approves the result before it syncs back to the Canvas gradebook. That review step matters enormously here, and it connects to a point we’ve made in our companion piece on human-in-the-loop marking: the tool isn’t replacing the lecturer’s judgement, it’s making sure the same judgement gets applied evenly to every candidate rather than drifting across a long day of vivas the way even a conscientious human examiner’s attention naturally does.
What this looks like set up in Canvas
The setup follows the same pattern as Eduface’s other Canvas integrations. A lecturer configures the session in Canvas, choosing Oral Exam mode, sets the rubric or question parameters, and links it to the relevant assignment. Eduface pushes the session link directly to students through Canvas, so there’s no separate login and no new platform for students to learn. The student completes the exam in the browser, speaking their answers, typically in fifteen to twenty minutes, and the recording and transcript are captured automatically. Once the lecturer reviews and approves the grade, it syncs to the Canvas gradebook the same way any other grade would.
Because it runs through LTI 1.3 rather than a Canvas-specific plugin, there’s genuinely no IT project required to set it up, and a first session can typically be live within 48 hours.
Why this matters specifically for business and management assessment
Business and management programmes lean on oral formats more than most disciplines, case interviews, client-facing role plays, dissertation vivas, group presentations defended individually afterward, precisely because the professional skills being tested, thinking on your feet, defending a recommendation under pressure, are inherently oral skills that a written exam can’t fully capture. That makes the examiner consistency problem worse, not better, than in a discipline that leans mostly on written assessment, simply because there are more oral touchpoints across a programme where inconsistency can creep in unnoticed.
Frequently asked questions
Are oral exams actually AI-proof?
They’re much harder to outsource to AI than a written submission, which is why interest in them has grown sharply. But “oral” alone doesn’t guarantee fairness. An oral exam still needs the same consistency safeguards, explicit criteria applied identically to every student, that written assessment has always needed, which is what Canvas’s Oral Exam mode is built to provide.
Does AI replace the human examiner in Oral Exam mode?
No. It applies the examiner’s own rubric consistently and drafts a criterion-by-criterion grade, but a lecturer reviews and approves every result before it’s released, exactly the human-in-the-loop model our companion pieces on Ofqual compliance describe.
How is this different from the Practice Viva mode described for Moodle?
Practice Viva is ungraded rehearsal students can repeat as many times as they like, covered in our companion piece on Moodle. Oral Exam mode is the graded assessment itself, applied once, with the result syncing to the gradebook after lecturer approval.
Do we need a Canvas-specific plugin to use this?
No. It connects through LTI 1.3, the same standard used across Eduface’s other Canvas integrations, so there’s no separate installation.
Why does examiner consistency matter more for business and management vivas specifically?
Business and management programmes tend to use oral formats more heavily than most disciplines, which means more opportunities across a programme for examiner variability to affect outcomes if it isn’t addressed structurally.
Sources
Harden, R. M., & Gleeson, F. A. (1979). Assessment of Clinical Competence Using an Objective Structured Clinical Examination (OSCE). Medical Education, 13(1), 41-54.
See a criterion-graded oral exam in Canvas
Book a 30-minute walkthrough, or start a free account and set one up on your own rubric.