Miatz T'aan — the adaptive viva
A spoken assessment that follows up where the answer was weak, scores against an SME-approved rubric, and always sends a borderline result to a human.
- Follow-up questions go deeper exactly where an answer was thin
- Every plan is approved by a subject-matter expert before it can be used
- The same plan and the same answers always produce the same questions — a disputed viva replays exactly
- A borderline result is never decided by a machine
Questions that follow the answer
A T'aan session works from an approved question plan: openers per blueprint domain, plus deeper probes an author can attach as explicit follow-ups.
- A weak answer drills down in the same area.
- A strong answer moves on to the next uncovered domain, heaviest first.
- No single area may consume the session.
Selection is deterministic. Same plan, same answers, same next question — every time.
The approval gate is real
A plan moves draft → in review → approved. Only an approved plan can run a session, and that is enforced in the database: an author may edit drafts but can neither approve a plan nor edit one that is already approved. An approved plan cannot be deleted by anyone using the app.
The borderline band
Scoring is weighted by the blueprint's own domains, then compared to the cut score. A result within a small margin either side of the line is borderline and goes to a human moderator. So does any criterion the scorer could not grade or flagged for review, and any blueprint domain the transcript never covered. A blueprint with no cut score is always borderline — nothing is certified by default.
What the candidate sees
The question, and whether they are finished. Never the rubric, never the depth map, never a score derived on their own device.
Frequently asked
What makes it adaptive?
The next question is chosen from what has already been answered. A weak answer drills down into the same area — following the explicit follow-up the author wrote, when there is one — while a strong answer moves on to the next blueprint domain that has not been covered. One area can never consume the whole session.
Is it random?
No, and that matters. Selection is fully deterministic: the same approved plan plus the same recorded answers always produce the same next question. A contested viva can be replayed exactly, which is impossible if the questions were sampled at random.
Who approves a plan?
An author drafts the question plan and rubric; a reviewer approves it. Until it is approved it cannot run a session — the database enforces this, not just the interface, and an author cannot approve their own plan or edit one that is already approved.
What is the borderline band?
A weighted result within a small margin either side of the cut score. Viva scoring is judgement, and the closer a result sits to the line the less any automated opinion should be trusted alone — so a borderline result is routed to a human moderator and is not final until they decide. A blueprint with no cut score at all is always treated as borderline.
Can a candidate see the rubric?
No. The rubric is the grading key. It is readable only by assessment staff, in the database as well as the interface, and the portal shows a count of rubric lines rather than their content.
Does it record audio?
Not through this pipeline. The viva works on text turns; any future media capture would be web-only, consent-gated and stored in the private evidence locker behind short-lived signed links — never in an ordinary request.
Related guides
Four deeper ways to establish that someone can actually do the work: an adaptive viva, in-product Journey runs, LTI launches from your LMS, and partner tiers that follow certified headcount.
ReadWhat an exam measures, in what proportion, and the pass mark a panel can defend.
Read