---
name: thesis-defence-and-viva-preparation
description: >
  Prepares a candidate to defend their own thesis: predicting questions from the
  weak points visible in the document, preparing answers for the standard
  question families, deciding in advance what to defend and what to concede,
  handling questions you cannot answer, building one-minute, five-minute and
  fifteen-minute summaries, running a mock defence, and managing the nerves
  practically. Use for "viva preparation", "thesis defence", "oral examination",
  "what will the examiners ask", "mock viva", "how do I defend my methodology",
  "I am nervous about my viva", "questions on my dissertation".
category: 15 Academic University Research
ref: "15.12"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Thesis Defence and Viva Preparation

## 1. One-line description

A method for preparing to defend a thesis, built on the fact that an examination tests whether the candidate understands and can defend their own work rather than whether the work is perfect, and that the questions are almost entirely predictable from the document itself.

## 2. What this skill is used for

**The research problem it solves.** Most candidates prepare for the wrong examination. They revise the content of their thesis, which they already know, and rehearse a defence of its strengths, which nobody will attack. They do not prepare for the questions that will actually be asked, which come from the weak points, and they do not decide in advance which weaknesses to defend and which to concede. This produces four recognisable failures in the room. The candidate defends an indefensible choice, which converts one weakness into a doubt about their judgement, and examiners will keep pressing precisely because the refusal to concede is more interesting than the original flaw. The candidate concedes everything, including things that were correct, and the examination becomes an inventory of faults with no demonstration of understanding. The candidate cannot state their contribution in one sentence, which reads as not knowing what the thesis is for. Or the candidate is ambushed by a question about a limitation they knew perfectly well was there and had simply hoped nobody would raise. All four are preventable, because the weak points are visible in the document, and examiners find them by reading it.

**Where it sits in the research lifecycle.** After submission and before the examination, typically in a window of two to ten weeks. Some of the preparation is far more useful earlier: the weakness inventory in Step 2, run before submission, is the basis of the examiner-proofing pass, and a candidate who has done it will have pre-empted several questions in the text itself.

**Typical use cases.**
- Working out what the examiners will actually ask, from the thesis.
- Preparing a defensible answer for each known weakness.
- Deciding what to defend and what to concede, in advance and on purpose.
- Building the one-minute, five-minute and fifteen-minute versions of the thesis.
- Rehearsing with a mock defence that is useful rather than reassuring.
- Preparing for the question you cannot answer.
- Handling the practical and emotional side without pretending it does not exist.

**Who uses it.** Masters candidates facing a viva, oral defence or presentation-plus-questions examination; supervisors running mock defences; candidates whose institution's format is unfamiliar to them, including international candidates encountering a different examination convention.

## 3. When to use it

- The thesis has been submitted and an examination is scheduled or expected.
- You do not know what your examiners are likely to ask.
- You know where the thesis is weak and have not decided what to say about it.
- You cannot state your contribution, your method justification or your limitations in a sentence each.
- A mock defence is being arranged and needs to be designed to be useful.
- The examination format is unfamiliar and you need to know what is being assessed.
- You are anxious about the examination and want the anxiety addressed practically.
- You are still writing, and want to pre-empt the questions in the text (run Steps 2 and 3 only).

## 4. When NOT to use it

- **The thesis is not yet submitted and the real task is improving it.** Preparing a defence for a fixable problem is the wrong order of work. Fix what can still be fixed, using **15.10 Data Analysis Chapter Development** or **15.11 Discussion and Contribution Development** as applicable, and defend only what remains. The exception is Steps 2 and 3, which are worth running early precisely so the text can pre-empt the questions.
- **The task is a hostile reading of the thesis as a document rather than preparation of a person.** Systematically reading a thesis as an adversarial examiner would, to find every attackable point before submission, is **15.19 Examiner-Proofing and Defence Strategy**, and it is a doctoral-level activity that this skill draws on in a lighter form. In the other direction: 15.19 audits the document and does not prepare answers, rehearsal, summaries or the candidate's conduct in the room.
- **The examination is a conference presentation or a public talk.** Different audience, different purpose, different time budget, and the questions are curious rather than examining. Use **15.27 Conference Abstract and Presentation** or **11.05 Research Presentation Development**.
- **The thesis has a serious integrity problem.** Where data cannot be produced, where a result cannot be reproduced from the data, or where sources have been cited that were never read, preparation is not the answer and this skill will not help construct one. The route is through the supervisor and the institution's own processes, immediately. `RESEARCHER DECISION REQUIRED` (K5 §2.4).
- **The candidate does not know their own work.** Where a thesis was assembled rather than researched, no amount of rehearsal survives fifteen minutes of questioning, because examiners test understanding by asking about the joins between chapters and about the choices that were not taken. This skill assumes the work is the candidate's and that the gap is in preparation, not in ownership.
- **The examination format has not been established.** Preparing for a two-hour closed viva when the actual format is a twenty-minute presentation with questions wastes most of the preparation. Establish the format first, which takes one email.
- **Your institution prohibits AI assistance for this task.** See §12 item 1. Note that in almost every institution an oral examination must be the candidate's own unaided performance, which makes rehearsed borrowed answers not only prohibited but tactically disastrous.

## 5. Required inputs

**Required.**
- **The submitted thesis, as submitted.** Preparation is against the document the examiners have, including its errors. Working from a later corrected version produces answers to questions nobody will ask.
- **The examination format, established from the institution.** Duration, who attends, whether a presentation is required and for how long, whether the supervisor attends and whether they may speak, whether it is open or closed, the possible outcomes, and what happens afterwards. Formats differ enormously between countries, institutions and even departments.
- **The time available before the examination**, so that preparation can be scheduled rather than attempted in one anxious weekend.

**Optional, and what each one adds.**
- **The examiners' names and their published work.** Tells you what they care about, what methods they use, and where they are likely to press. An examiner who has published on a method you rejected will ask about the rejection. Read enough to know their position, and never so much that you construct an imaginary adversary.
- **The marking criteria or examination guidance the examiners work to.** Tells you what they must decide and therefore what they must ask about.
- **The supervisor's own assessment of the thesis's weak points.** A supervisor who has read a hundred theses will name in ten minutes the three things that will be raised.
- **A peer from a different subfield.** The most useful mock examiner available, because they will ask the naive questions that expose unexamined assumptions, which are the questions that actually catch candidates.
- **The examination room or platform, seen in advance.** Removes an entire category of avoidable stress and technical failure.

## 6. Questions to ask before starting

1. **What exactly is this examination assessing, in this institution?** In some systems it verifies authorship and understanding; in others it is a substantive academic debate that can change the outcome; in others it is largely ceremonial with the decision effectively made. Determines how to prepare and how much rests on it. Default if unanswered: prepare as though it is substantive and can change the outcome, since over-preparation is recoverable and under-preparation is not.
2. **What are the possible outcomes and what do they mean?** Pass, pass with minor corrections, major corrections, revise and resubmit, fail. Knowing that corrections are the normal outcome rather than a failure changes how the examination feels. Default: assume corrections are likely, because they usually are, and treat questions as identifying them rather than as attacking you.
3. **Where is the thesis genuinely weak?** Determines the whole preparation. Default: run the inventory in Step 2 honestly, and if the answer is "nowhere", the inventory has not been done properly.
4. **Which weaknesses will you defend and which will you concede?** This must be decided before the room, not in it. Default: concede anything that is genuinely a limitation, defend anything that was a reasoned trade-off, and know which is which.
5. **Can you state the contribution in one sentence?** If not, that is the first thing to fix, since it is the most likely opening question. Default: write it, test it against the thesis, and memorise its shape rather than its wording.
6. **What did you learn that is not in the thesis?** Examiners often ask what the candidate would do differently, and the answer demonstrates researcher development rather than thesis quality. Default: prepare it, because it is asked more often than candidates expect and an unprepared answer sounds evasive.
7. **What is the practical situation on the day?** Format, location, travel, timing, what to bring. Default: settle all of it a week ahead, because logistical stress on the day consumes attention that the examination needs.

## 7. Step-by-step methodology

**Step 1. Establish what the examination actually tests, and reframe accordingly.**
An examination is not a test of whether the thesis is perfect. No thesis is, examiners know this, and a thesis with no limitations would be a thesis with no scope. It tests four things: that the work is the candidate's own; that the candidate understands what they did and why; that they can defend their choices and recognise the limitations of those choices; and, in most systems, that the work meets the standard for the award. This reframing is not reassurance, it is tactical. It means that a question about a weakness is not an accusation, it is the examiner establishing whether you know the weakness is there. It means that "I chose X, for these reasons, and it cost me Y" is a complete and successful answer even where X was not the ideal choice. And it means the failure mode to avoid is not having a flawed thesis but appearing not to understand your own. Write this reframing down, because it changes how every subsequent step is done. *Correct result: a written statement of what your specific examination assesses, taken from institutional guidance rather than from folklore.*

**Step 2. Read your own thesis as an examiner, and build the weakness inventory.**
Examiners find weak points by reading the document, so read it the way they will. Read the abstract, introduction and conclusion first and together, because that is what examiners do, and the mismatches between them are the fastest source of questions. Then read chapter by chapter with one question: **where would I attack this?** Look systematically in the places weaknesses concentrate: the sample and what it excludes; the response rate and the non-responders; any instrument that was adapted, unvalidated or borrowed from a different population; every methodological choice that was made for practical rather than principled reasons; assumption checks that were not reported; analyses that appear in the plan and not in the results; any claim in the discussion that exceeds its finding; the causal language that crept in; the limitations that are named nowhere; anything the framework did not explain; the literature published after your review; internal inconsistencies between chapters; and the abstract's claim measured against what the thesis delivered. Write each one down as a numbered inventory item with its page reference. Aim for fifteen to thirty. If you find fewer than ten, you are reading defensively, and the examiner will not. *Correct result: a numbered weakness inventory with page references, ordered by how likely each is to be raised.*

**Step 3. Convert each inventory item into the question an examiner would ask.**
Weaknesses are not asked about as accusations, they are asked about as questions, and predicting the wording helps you recognise the question when it arrives. A convenience sample becomes "how did you recruit your participants, and what does that mean for your findings?" An unreported assumption check becomes "what did you do about the distribution of this variable?" An overclaim in the abstract becomes "you say your study demonstrates X, can you take me through the evidence for that?" A missing recent paper becomes "are you familiar with the work of..." Note that the last of these is usually not a trap: examiners ask about work you have missed to see whether you can engage with it, and "I am not familiar with that specific paper, but if it finds X then it would bear on my second research question in this way" is a strong answer, while a bluff is fatal. Write the predicted question for every inventory item. *Correct result: a question list, one per inventory item, in the wording an examiner would plausibly use.*

**Step 4. Prepare the standard question families, which appear in almost every examination.**
Seven families recur, and each has a shape that a good answer follows.
**Why this topic?** Wants the research problem and its significance, not your biography. Answer with the gap and why it mattered, in three sentences, and only then, briefly, your own route to it if it is relevant.
**Why this method, and why not the alternative?** The most common family and the most predictable. Answer with the research question first, then the method as what the question required, then the named alternative and the specific reason it was rejected, then what your choice cost. This answer already exists in the methodology chapter if it was written properly.
**What is your contribution?** Answer in one sentence with its bounds attached, then expand to three if invited. Never answer this with a description of what you did; a contribution is what is now known, not what you performed.
**What would you do differently?** Wants evidence of researcher development. Answer with a specific, substantive change and the reason, not a trivial one and not "nothing". Saying you would change nothing reads as an inability to evaluate your own work.
**What are the limitations?** Answer with the two or three that actually matter, each with its consequence, and do not recite the list. Volunteering the real one before being asked is a strong move.
**How do you respond to author X, who disagrees with you?** Answer by diagnosing the difference (population, measurement, method, period, analysis) before conceding or defending, and be prepared for the possibility that you are wrong, in which case say so and say what follows.
**What does this finding mean?** Wants interpretation, not restatement. Answer with the reading, the confidence, and the alternative explanation you cannot rule out.
Add two more that are frequently asked and rarely prepared: **where does this sit in the wider literature?**, which tests whether you can see your own work from outside, and **what would you do next?**, which tests whether the work opened anything. *Correct result: a prepared answer for each of the nine families, none longer than ninety seconds spoken.*

**Step 5. Build an answer card for every question, in a fixed four-part shape.**
For each predicted question, prepare: a **direct answer in one or two sentences**, because examiners are testing whether you can locate the point, and a long preamble reads as evasion; the **evidence or reasoning**, with the chapter and page you would point to; the **concession**, which is what you acknowledge the choice cost or the limitation implies; and the **follow-up you expect**, which is what they will ask next if the answer lands. Keep each card to a single page and write it in your own speech, not in written English, because a memorised written sentence is audible as one and it breaks the impression that you are thinking. Do not script whole answers; script the first sentence and the concession, which are the two parts most likely to go wrong under pressure, and let the middle be spoken. *Correct result: a set of thirty to fifty single-page cards, each with the four parts, written in speech.*

**Step 6. Build the concession register, and learn to concede well.**
This is the step that most distinguishes a strong defence from a weak one. Sort every inventory item into three columns. **Defend**: choices that were reasoned trade-offs, where you would make the same choice again given the same constraints, and where you can state the reasoning and its cost. **Concede**: genuine limitations, errors, and things you would now do differently, which you acknowledge cleanly. **Depends**: things you will defend on the reasoning and concede on the outcome, which is the most common category and needs the most rehearsal.
Then learn the shape of a good concession, because conceding well is a demonstrated strength and not a loss: acknowledge it directly and without hedging, state the specific consequence for the findings, state what you would do differently, and stop. "You are right that the cross-sectional design cannot establish temporal order. That means my second finding is an association and I should not have written it as a process. A follow-up measurement at six months would settle it." That answer is better than any defence of the same point, and examiners respond to it as competence.
What must not happen is the two failure modes. **Defending a real limitation** turns one flaw into a doubt about your judgement, and examiners will press harder precisely because your refusal is more interesting than the flaw. **Conceding a sound choice** because the question sounded critical gives away something you were right about, and it is the more common error among nervous candidates: a question is not an accusation, and "why did you do it that way?" is frequently curiosity. Rehearse holding a defensible position under one round of pushback, since examiners often press once to see whether you know why you did it or merely did it. *Correct result: every inventory item assigned to defend, concede or depends, with the concessions written out in their four-part shape.*

**Step 7. Build the thesis in one minute, five minutes and fifteen.**
Three versions, each with a different job. **One minute**: the problem, what you did, what you found, and what it contributes. Four sentences. This is the answer to "tell us about your thesis", which opens a very large proportion of examinations, and it is the moment the examiners form their impression of whether you understand your own work. **Five minutes**: adds the method and its justification, the principal finding with its evidence, the main limitation, and the contribution with its bounds. This is the typical opening presentation where one is required. **Fifteen minutes**: adds the literature context, the framework, the design in detail, the findings in sequence, the interpretation, the limitations and the future work. This is the full presentation format and it is also the version you will use to structure long answers.
Build them in that order, shortest first, because expanding a one-minute version keeps the argument intact whereas compressing a fifteen-minute version usually preserves the wrong material. Rehearse each aloud and to time. The one-minute version should be near-memorised as a shape, not as words. *Correct result: three timed versions, each rehearsed aloud, the shortest one solid.*

**Step 8. Prepare for the question you cannot answer, because there will be one.**
Every examination contains at least one question you have not prepared for, and how you handle it matters more than whether you know the answer. The sequence that works: pause, which is permitted and looks like thought rather than panic; clarify if you genuinely do not understand the question, since answering the wrong question is worse than admitting confusion; then answer what you can and be explicit about the boundary. "I have not considered that specific comparison. What I can say is..., and I would need to look at whether... before I could answer properly." That is a complete and credible answer. The three things that damage a candidate are bluffing, which experienced examiners identify immediately and then probe; freezing into silence without saying what is happening, which can be defused by saying "let me think about that for a moment"; and arguing about the question's premise without engaging with the concern behind it, since the concern is usually legitimate even when the framing is not. It also helps to know that you may ask for a question to be repeated, may ask for a moment, and may return to a question later in the examination if something occurs to you. Rehearse this explicitly by having a mock examiner ask you three questions you cannot possibly have prepared for. *Correct result: a rehearsed sequence for unknown questions, practised at least three times under pressure.*

**Step 9. Run a mock defence designed to be useful rather than reassuring.**
A mock with your supervisor, who knows and likes the work, is worth far less than one designed adversarially. Design it: at least one examiner who has not read the thesis before this week, so that they read it as an examiner does; at least one from outside your subfield, who will ask the naive questions that expose assumptions; a full-length session in the actual format, including the presentation if there is one; and a brief given to the mock examiners that names the weakness inventory, so they attack the real points rather than the polite ones. Record it if you can, because candidates consistently misjudge their own pace and their own filler. Debrief on four things: which questions you had not predicted, which answers ran long, where you conceded something you should have defended or defended something you should have conceded, and where you did not know the answer. Then revise the cards. One well-designed mock is worth five gentle ones, and two are better than one, spaced a week apart with revision in between. *Correct result: a completed adversarial mock, a list of unpredicted questions now added to the inventory, and a revised card set.*

**Step 10. Prepare the room, the materials and the logistics.**
Know: the format and duration, the location or platform, who will be present, whether you may bring the thesis and notes (usually yes, and a tabbed copy is normal and expected), whether you present and for how long, and what happens at the end, including whether you leave the room while the examiners deliberate. Prepare a tabbed thesis with the tables, figures and key pages marked so you can find anything within seconds, since being unable to locate your own table is a small failure that dents confidence disproportionately. Bring water, a pen and paper for noting multi-part questions, since examiners frequently ask three things at once and answering all three is a visible strength. For online examinations, test the platform, the audio and the screen sharing in advance, and have a fallback route agreed in case of failure. Settle travel and timing so that nothing on the day requires decisions. *Correct result: a checked logistics list and a tabbed thesis, complete at least three days before.*

**Step 11. Handle the emotional dimension practically, not dismissively.**
Anxiety before an examination is normal, near-universal, and largely proportionate: this is a high-stakes assessment of work you have spent a year on, and being told to relax is useless. What helps is specific. **Preparation reduces anxiety more than reassurance does**, and the weakness inventory helps most of all, because most examination fear is fear of the unknown question, and the inventory converts unknowns into a finite list. **Know that corrections are normal**, in most systems the majority outcome, and are not a judgement on you. **Rehearse aloud rather than mentally**, because the gap between knowing something and saying it out loud is where candidates are surprised. **Plan the physical basics**: sleep the two nights before rather than the one, eat, arrive early, and do not revise in the final hour, which raises anxiety without adding knowledge. **Have a recovery move for the moment it goes badly**, because it will at some point in the session: a slow breath, a sip of water, and the sentence "let me take that from the beginning" resets it, and examiners do not read a recovered stumble as a failure. **Remember that you know your thesis better than anyone in the room**, which is simply true: the examiners have read it once or twice, you have lived in it. And **have something arranged for afterwards**, since the aftermath, whatever the outcome, is easier with a plan than without. `RESEARCHER REVIEW RECOMMENDED` where anxiety is severe enough to affect functioning, in which case the institution's student support services are the right route and preparation alone is not (K5 §2.4). *Correct result: a plan for the two days before, the day itself, and afterwards, written down rather than intended.*

## 8. Analytical framework

**The weakness-to-answer pipeline.** The whole preparation is one transformation, run on every weak point in the thesis:

    Weakness in the document → Question an examiner would ask →
    Direct answer → Evidence and reasoning → Concession →
    Anticipated follow-up

Applying it: the pipeline is what converts anxiety into work, because each stage is finite and inspectable. A weakness with no predicted question means you have not read as an examiner. A question with no direct answer means you have not decided your position. A position with no concession means you are planning to defend something you may need to give up. And an answer with no anticipated follow-up means you have prepared for the first question only, which is where most candidates stop and where examinations actually begin.

**The defend-concede judgement.** Applied to every item, and the single most consequential distinction in the preparation:

    Was this a reasoned trade-off under real constraints?
      Yes → Defend: state the reasoning, state the cost, hold under one push
      No, it is a genuine limitation → Concede: acknowledge, state the consequence,
          state what you would do differently, stop
      Reasoning sound but outcome poor → Depends: defend the reasoning,
          concede the outcome, in that order

Applying it: the two errors are symmetrical and both are costly. Defending what should be conceded converts a limitation into a doubt about judgement, and examiners press harder when they sense the refusal. Conceding what should be defended gives away a sound choice to a question that was merely curious. Deciding in advance, in writing, is what prevents both, because neither judgement is reliably made under pressure in the room.

## 9. Output format

The output is a preparation pack, and it is for the candidate rather than for anyone else.

**1. Examination brief.** Format, duration, attendees, presentation requirement, possible outcomes, what happens afterwards, logistics.

**2. Weakness inventory.**

| No. | Weakness | Location (chapter, page) | Likelihood of being raised | Defend, concede or depends |
|---|---|---|---|---|

**3. Predicted question list.** One per inventory item, in an examiner's plausible wording, plus the nine standard families.

**4. Answer cards.** One page each:

    Question
    Direct answer (one or two sentences)
    Evidence and reasoning (with page reference)
    Concession
    Anticipated follow-up

**5. Concession register.** Every item sorted into defend, concede or depends, with the concessions written out in their four-part shape.

**6. The three summaries.** One minute, five minutes, fifteen minutes, each timed.

**7. Unknown-question protocol.** The rehearsed sequence, and the phrases you will actually use.

**8. Mock defence record.** Unpredicted questions, answers that ran long, misjudged defend-concede calls, and the revisions made.

**9. Logistics and wellbeing plan.** The two days before, the day, and afterwards.

**When the thesis is weaker than the candidate would like, the format must not produce false confidence (K4 §1).** A weakness inventory that is short is a preparation failure, not a good thesis. An answer card whose concession line is empty has not been thought through. The pack's job is to make the candidate accurate about their own work, and a candidate who walks in knowing exactly where the thesis is weak, and having decided what to say about each point, is in a far stronger position than one who walks in hoping.

## 10. Quality checks

Run a week before the examination.

1. Is the examination format established from institutional guidance rather than assumed?
2. Does the weakness inventory contain at least fifteen items with page references?
3. Does every inventory item have a predicted question and an answer card?
4. Is every item assigned to defend, concede or depends, in writing?
5. Is every concession written in its four-part shape, ending with a stop rather than a justification?
6. Can you state the contribution in one sentence, with its bounds, without notes?
7. Can you state the method justification, including the rejected alternative, in ninety seconds?
8. Do you have a substantive answer to "what would you do differently", not a trivial one?
9. Are all nine standard question families prepared?
10. Have the one-minute, five-minute and fifteen-minute summaries been rehearsed aloud and to time?
11. Has the unknown-question sequence been practised under pressure at least three times?
12. Has an adversarial mock been run, with at least one examiner from outside the subfield?
13. Were the mock's unpredicted questions added to the inventory and answered?
14. Is the thesis tabbed so any table, figure or key passage can be found in seconds?
15. Are the logistics settled, and is there a plan for the two days before and for afterwards?
16. Have you checked for literature published since submission that an examiner might raise?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Preparing the strengths** | Rehearsal focuses on what the thesis does well | Prepare from the weakness inventory (Step 2) |
| **The defensive read** | An inventory with six items | Read as an attacker, target fifteen to thirty (Step 2) |
| **Defending the indefensible** | Holding a position past the second push | Assign defend or concede in advance (Step 6) |
| **Conceding the sound** | Giving away a reasoned choice to a curious question | Rehearse holding under one round of pushback (Step 6) |
| **The unstatable contribution** | The answer to "what is your contribution" takes two minutes | One sentence with bounds, memorised as shape (Steps 4, 7) |
| **Biographical answers** | "Why this topic" answered with personal history | Lead with the problem and the gap (Step 4) |
| **"I would change nothing"** | Offered as confidence | Prepare a specific substantive change (Step 4) |
| **Reciting the limitations list** | All nine limitations delivered in order | Two or three that matter, with consequences (Step 4) |
| **Bluffing an unread paper** | Claiming familiarity with work you have not read | Say you have not read it and engage with its claim (Steps 3, 8) |
| **Freezing** | Silence with no signal to the examiners | Say "let me think about that for a moment" (Step 8) |
| **The comfortable mock** | The mock examiners are supportive colleagues | Brief them adversarially, include an outsider (Step 9) |
| **Preparing the first question only** | Every card stops at the direct answer | Every card names the anticipated follow-up (Step 5) |
| **Memorised written answers** | Answers that sound like read prose | Script the first sentence and the concession only (Step 5) |
| **Logistics failure** | Platform, room or timing settled on the day | Complete the logistics list three days ahead (Step 10) |
| **AI-scripted defence** | Fluent answers the candidate cannot vary under follow-up | Preparation produces questions and structure, not answers (§12) |

## 12. AI guardrails

Skill-specific. The universal prohibitions in K4 apply in full and are not repeated. The human in the loop for every K5 marker in this skill is the candidate together with their supervisor.

1. **Academic integrity is a condition of use, not a footnote.** These skills assist a researcher's thinking, structure and rigour. They do not produce work to be submitted as the student's own unaided output. The user must comply with their institution's AI use policy and its declaration requirements, which vary by institution and by assessment. Where an institution prohibits AI assistance for a task, this skill must not be used for it. The skill never writes a passage for submission as though the student wrote it; it interrogates, structures, critiques and teaches. Operationally in this skill: it will predict questions, structure answer cards, run the defend-concede sorting, design the mock, and challenge a candidate's draft answers. It will not write the answers the candidate will give, because an oral examination is by definition an unaided performance, and a borrowed answer collapses at the first follow-up question, which is where examinations actually test.

2. **Never supply a defence for a choice the candidate cannot explain in their own terms.** If the candidate does not know why a decision was made, the honest position in the room is that it was made for a practical reason, or on a supervisor's advice, or that in hindsight it was not well justified. A supplied justification the candidate did not arrive at will not survive "and why did you reject the alternative?"

3. **Never characterise an examiner's likely views from their name or field.** Beyond what the candidate has actually read of their work, this is speculation, and it produces an imaginary adversary that distorts preparation. Where the candidate has read a paper, work from the paper.

4. **Never assert what a cited author's position is unless the candidate has read them.** The "how do you respond to author X" family is precisely where an unread characterisation fails, because the examiner has read them (K4 §2.4).

5. **Never help a candidate conceal a weakness.** The correct preparation for a real limitation is a clean concession, not an evasion. Assisting with an evasion sets up the worse outcome, since examiners press hardest exactly where they sense one.

6. **Never reassure a candidate that a genuine problem is not a problem.** Where the inventory contains something serious, say so and prepare the concession. False reassurance is the least useful thing available and the most tempting.

7. **Never fabricate the thesis's content in order to prepare an answer.** Every answer card is built on what the thesis actually says, at the page it says it. Where the candidate cannot recall what a chapter concluded, the instruction is to open it (K4 §6.4).

8. **Never treat examination anxiety as a problem to be talked away.** Address it with preparation, information about outcomes, rehearsal and logistics. Where it is severe enough to affect functioning, name the institutional support route rather than continuing to coach.

9. **Never predict the outcome of the examination.** You have not read the thesis as an examiner, you do not know the institution's standard, and a prediction in either direction is unhelpful: false comfort reduces preparation, and false alarm damages performance.

## 13. Best-practice principles

- **The examination tests whether you understand your work, not whether the work is flawless.** Every thesis has limitations; only some candidates know theirs.
- **The questions are in the document.** Examiners are not inventing attacks, they are reading and noticing. Read your own thesis the way they will and you will predict most of what is asked.
- **Conceding a real weakness well is a demonstrated strength.** Acknowledge, state the consequence, state what you would do differently, stop. Do not justify after conceding, because the justification undoes it.
- **Defending a reasoned trade-off is also a strength, and it needs to survive one push.** Examiners often press once to distinguish a decision from a default.
- **Decide defend or concede in advance.** Neither judgement is reliably made in the room.
- **Answer the question first, then explain.** A preamble before the answer reads as evasion even when it is thoroughness.
- **Ninety seconds is a long answer.** Candidates consistently run long, and a long answer invites the examiner to interrupt, which is a worse dynamic than a short answer inviting a follow-up.
- **Write cards in speech, not in prose.** A memorised written sentence is audible and it stops you thinking.
- **Bluffing is the only unrecoverable move.** Not knowing something is normal; pretending to know it changes what the examination is about.
- **One adversarial mock beats five supportive ones**, and an examiner from outside your subfield is worth more than one inside it, because the naive question is the one that finds the unexamined assumption.
- **Preparation is the most effective treatment for the anxiety.** Most examination fear is fear of the unpredicted question, and the inventory turns that infinite set into a finite list.
- **You know this thesis better than anyone in the room.** That is not encouragement, it is a fact about how much time you have spent with it.

## 14. Worked example

Generic fictional scenario, academic, in an environmental policy context.

**INPUT**

A masters candidate has submitted a dissertation on household participation in a municipal recycling scheme, based on a survey of 187 residents in two neighbourhoods and six follow-up interviews. The examination is a 60-minute closed oral with two examiners, one internal and one external, no presentation. The candidate is confident about the analysis and anxious about "being asked something I have not thought of".

**PROCESS**

*Step 1.* The institutional guidance is read. The examination assesses understanding, ownership and the standard for the award, and the normal outcomes are pass with minor corrections or pass with no corrections. This alone reduces the candidate's expectation that the session is designed to catch them.

*Step 2.* Reading as an examiner produces twenty-two inventory items. The largest four: the two neighbourhoods were selected because the municipality supplied contact lists for them, which is a convenience selection presented in the thesis as purposive; the response rate was 23 percent and non-responders are not discussed anywhere; the survey measured self-reported recycling behaviour and the thesis occasionally discusses it as behaviour; and the discussion states that the scheme "increased participation", which is a causal claim on cross-sectional data.

*Step 3.* Predicted questions are written. "How did you choose your two neighbourhoods?" "What do you know about the people who did not respond?" "You measured what people said they do. What would you expect the relationship to be with what they actually do?" "You say the scheme increased participation. Can you show me the evidence for that?"

*The judgement call.* The candidate's instinct is to defend all four, and specifically to argue that the neighbourhood selection was purposive because the two areas differ in housing type. Applying the defend-concede test changes three of the four. The neighbourhood selection is a **concede**: the areas were available, the housing difference was noticed afterwards, and the honest answer is that it was a convenience selection with a fortunate variation, which limits the comparison to a descriptive one. The response rate is a **depends**: 23 percent is low but not unusual for this mode, so the reasoning is defensible, while the absence of any non-response discussion is a genuine gap and is conceded, along with the direction of the likely bias, since people who recycle are more likely to answer a survey about recycling, which inflates the participation estimate. The self-report issue is a **defend with concession**: self-report was the only feasible measure and the thesis says so, but the discussion's slippage into behavioural language is conceded and identified as a correction. The causal claim is a straight **concede**: the design cannot support it, the sentence should have read "was associated with", and it is offered as a correction before being asked about, which is the strongest available move.

*Step 4.* The nine families are prepared. The contribution answer is drafted three times before it works: the first version describes what was done, the second is unbounded, and the third states that the study provides the first evidence in this municipality that participation differs by housing type rather than by stated environmental attitude, and identifies bin storage access as a candidate constraint the existing local policy framing does not distinguish from motivation.

*Steps 8 and 9.* An adversarial mock is arranged with a supervisor's colleague who has not read the work and a peer from a different discipline. Four questions arrive that were not on the list, three of them from the outsider, including "why did you not just look at the council's collection data?" This is the best question of the session, because it exposes that the candidate had considered and rejected that route for a good reason that appears nowhere in the thesis. It becomes a prepared answer and, in the event, is asked in the real examination.

*Step 11.* The candidate's stated fear, the unpredicted question, is addressed directly: the inventory now has twenty-six items, the unknown-question sequence has been rehearsed three times, and the framing has shifted from "they will find something" to "they will find these, and I have decided what to say about each".

**OUTPUT**

A preparation pack with a 26-item inventory, 35 answer cards, a concession register with four items in defend, eleven in concede and eleven in depends, three timed summaries, a rehearsed unknown-question protocol, a completed adversarial mock with revisions, and a logistics plan. In the examination, three of the four major weaknesses are raised in close to the predicted wording, the causal overclaim is volunteered by the candidate before it is asked about, and the outcome is a pass with minor corrections, two of which the candidate had already identified.

`RESEARCHER REVIEW RECOMMENDED` on the defend-concede sorting, since judging which choices this department's examiners will regard as reasoned trade-offs benefits from a supervisor who knows them (K5 §2.1).

## 15. Advanced usage

**Run the inventory before submission.** Steps 2 and 3, applied to a final draft, let you pre-empt questions in the text itself: a limitation named in the thesis is a limitation you cannot be caught by. This is the highest-return use of this skill and the one candidates discover too late. See **15.19 Examiner-Proofing and Defence Strategy** for the systematic version.

**Read one paper by each examiner, and stop.** Enough to know their methodological commitments and whether your work touches their published position. Beyond that, reading turns into constructing an adversary who does not exist, and preparation drifts toward an imagined debate rather than your own thesis.

**Prepare the corrections list.** Most outcomes involve corrections, and a candidate who arrives with their own list of known errors and intended fixes is visibly in command of the work. Offering it at the end, if invited, frequently shortens the corrections process.

**The three-question drill.** Have someone ask any question about the thesis, then ask "why?" twice more in sequence. This is what a real examination does, and most candidates have prepared only the first layer. Running it on ten questions exposes exactly where understanding stops.

**Where the standard approach does not fit.** In systems where the defence is public and partly ceremonial, the preparation shifts toward the presentation and toward composure with an audience, and the weakness inventory matters less. In systems where the examination can substantially change the outcome, everything here matters more and the mock should be run twice. Where the examination is conducted in a language that is not the candidate's first, rehearse aloud considerably more, prepare the technical vocabulary explicitly, and know that asking for a question to be repeated or rephrased is entirely normal and carries no penalty. Where a candidate has a disability or condition affecting oral examination, institutional adjustments exist and should be requested well in advance rather than endured.

## 16. Skill chain

**Recommended previous skills:**
- **15.11 Discussion and Contribution Development.** Hands over the contribution statement, the diagnosed disagreements with the literature and the limitations, which supply the answers to three of the standard question families.
- **15.08 Research Design and Methodology Chapter.** Hands over the rejected alternatives, which are the prepared answer to the most commonly asked family.
- **15.10 Data Analysis Chapter Development.** Hands over the reporting completeness, whose gaps are among the most predictable question sources.
- **15.19 Examiner-Proofing and Defence Strategy**, where a full adversarial reading of the document was run before submission.

**Recommended next skills:**
- **15.24 Journal Article Development.** The examination frequently identifies the publishable core, and examiners often say so directly.
- **15.26 Peer Review Response.** The concession discipline learned here transfers directly to responding to reviewers.

**Runs well alongside:**
- **11.05 Research Presentation Development**, where the examination includes a presentation.
- **15.18 Thesis Architecture and Chapter Coherence**, whose coherence checks surface the inconsistencies examiners ask about.
- **13.04 Bias Detection**, for finding the places where the argument was assembled around a preferred answer.

---
A Yazi Supplied Skill and resource.