---
name: ethics-clearance-application
description: >
  Prepares a complete institutional research ethics submission: proportionate
  risk assessment, participant information sheet and matching consent form,
  recruitment materials, data management and retention, anonymisation and
  re-identification risk, vulnerable participant protections, distress protocols
  with an escalation route, secondary data consent scope, AI processing
  disclosure and a realistic timeline. Use for "ethics application", "IRB
  submission", "ethics committee", "participant information sheet", "consent
  form", "my ethics application was returned", "do I need ethics approval",
  "data management plan for ethics", "vulnerable participants".
category: 15 Academic University Research
ref: "15.09"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Ethics Clearance Application

## 1. One-line description

A method for preparing an institutional ethics submission that passes first time, by matching the risk assessment to the actual study, writing participant materials that a participant can understand and that describe what will really happen, and closing the gaps of completeness and internal consistency that cause most rejections.

## 2. What this skill is used for

**The research problem it solves.** Most ethics applications are not returned because the research is unethical. They are returned because they are incomplete, internally inconsistent or vague: the consent form promises anonymity the design cannot deliver, the information sheet is written at postgraduate reading level for participants who are not postgraduates, the retention period is left blank, the recruitment poster mentions an incentive the application does not, or the protocol says interviews may cover distressing material and names nobody to refer a distressed participant to. Two further errors are near universal. Risk is misjudged in both directions: a low-risk interview study is written up as though it were a clinical trial, which wastes committee time and invites scrutiny, or a study on a genuinely sensitive topic dismisses risk as negligible, which is the fastest route to a full-committee referral. And the timeline is underestimated. Ethics approval is the most commonly underestimated dependency in a dissertation, and the student who submits in week six and expects to interview in week seven loses a month they did not have.

**Where it sits in the research lifecycle.** After the methodology is settled, because the committee assesses the procedure, and before any contact with any participant, including recruitment and piloting. It reappears whenever the design changes materially, since amendments require approval too.

**Typical use cases.**
- Assembling a first submission from a finished methodology chapter.
- Writing a participant information sheet and a consent form that match each other and match the study.
- Deciding whether a study needs full review, expedited review or an exemption.
- Working out what protections a particular vulnerable group requires.
- Building a distress protocol with a named escalation route for a sensitive-topic study.
- Establishing whether existing consent covers a secondary use of a dataset.
- Disclosing AI processing of participant data appropriately.
- Responding to a committee's requested revisions without starting again.

**Who uses it.** Masters candidates preparing a first ethics submission; supervisors reviewing before submission, since supervisors usually co-sign and carry responsibility; students in disciplines where ethics review is universal and in disciplines where it is newly extended and less familiar.

## 3. When to use it

- The study involves human participants, their identifiable data, or their tissue, in any way.
- The study uses existing data about identifiable people, even where you did not collect it.
- Recruitment is about to begin, including piloting, which requires approval in most institutions.
- You do not know whether your study needs review, or what level of review it needs.
- The topic may cause distress, or the participants may be in a dependent relationship with you.
- The application has been returned with requested revisions.
- The design has changed since approval, and you need to know whether an amendment is required.
- You are planning the dissertation timeline and need to place the approval dependency realistically.

## 4. When NOT to use it

- **The substantive ethical design question is unresolved.** Whether a study should be done at all, how consent should be conceptualised for this population, whether a covert element is ever justifiable, how to balance participant benefit against risk: these are ethical reasoning tasks, not submission tasks. Use **13.05 Research Ethics and Consent Design** for the substantive work, then return here to document it. In the other direction: 13.05 does not know your institution's forms, review levels or committee expectations, and does not produce a submission.
- **The study does not involve human participants or personal data at all.** Purely analytical work on published literature, simulation, non-human data or fully open aggregate statistics may fall outside the review remit entirely. Check rather than assume, since institutions differ on secondary data and on publicly available social media content, but do not manufacture an application for a study that needs none.
- **The methodology is not yet settled.** The committee assesses a specific procedure. An application describing a design still in flux will be returned, and rewriting it costs more than waiting two weeks. Finish **15.08 Research Design and Methodology Chapter** first.
- **A serious ethical problem exists in the design.** Where a study cannot obtain meaningful consent, where risk to participants is not outweighed by any plausible benefit, or where the researcher's relationship with participants makes voluntariness impossible, the answer is to change the design, not to write a better application. `RESEARCHER DECISION REQUIRED` (K5 §2.4). An application that conceals such a problem exposes the student, the supervisor and the participants.
- **The study is in a regulated domain with its own statutory approval route.** Clinical research, research in health or care systems, work with prisoners or children in care, and research involving controlled substances or medical devices typically require external approvals in addition to or instead of the university committee, with their own forms and timelines measured in months. Establish the route before writing anything, because the wrong form is a total loss.
- **Approval has been granted and the design has since changed materially.** That is an amendment, which is a different, shorter submission against the existing approval. Writing a fresh application invites a fresh review of everything already approved.
- **Your institution prohibits AI assistance for this task.** See §12 item 1. Note that some institutions specifically require ethics materials to be the student's own composition.

## 5. Required inputs

**Required.**
- **Your institution's actual ethics form, guidance notes and submission deadlines.** Forms vary enormously in structure, terminology and what they ask. Writing to a generic template and transferring it later loses content and introduces inconsistency. If the form has not been obtained, obtain it before starting.
- **The settled methodology: population, recruitment route, procedure, instrument, data types, storage and analysis.** The application is a restatement of these in the committee's terms, and it must match the methodology chapter exactly.
- **The supervisor's involvement.** In almost all institutions the supervisor co-signs and is formally responsible. A submission the supervisor has not read is not ready.

**Optional, and what each one adds.**
- **A recently approved application from the same department.** The single most useful input available. It shows the committee's expectations, the level of detail that satisfies them, and the phrasing that has already worked.
- **The committee's stated review levels and criteria.** Lets you target the right level and pre-empt the specific things this committee checks.
- **Existing validated participant information sheet templates from the institution.** Removes an entire category of formatting rejection.
- **A named referral or support service appropriate to the participant group.** Turns a distress protocol from a promise into a procedure. Without one, the protocol is unusable.
- **The data protection or information governance guidance applying at your institution.** Determines retention periods, storage locations and lawful basis language, which committees increasingly check closely.
- **Documentation of the original consent, for secondary data.** Determines whether your intended use is covered, which cannot be inferred from the dataset itself.

## 6. Questions to ask before starting

1. **What review level does this study need, and who decides?** Full committee, expedited, departmental or exempt. Determines the timeline and the depth of everything. Default if unanswered: prepare to the standard of the next level up, since over-preparing costs hours and under-preparing costs a review cycle.
2. **When does the committee meet, and what is the submission deadline before each meeting?** This is the whole timeline question and it is answerable in five minutes. Default: assume one full cycle for the first submission plus one for revisions, and count backwards from the fieldwork start date.
3. **Are any participants in a dependent or unequal relationship with the researcher?** Students, employees, patients, clients, colleagues and people the researcher has power over cannot give freely voluntary consent without specific safeguards. Determines a whole section. Default: examine the recruitment route explicitly for this, because it is frequently present and unnoticed.
4. **Could any participant be identified from the reported findings, given the sample?** Small samples, specialist roles and organisational studies routinely make people identifiable even without names. Default: assume identifiability is possible in any sample under about 30 within a defined organisation or role, and design the reporting protections accordingly.
5. **Could the topic or the process cause distress, and what happens if it does?** Determines whether a distress protocol is required. Default: assume it is required wherever the topic touches on health, bereavement, work stress, discrimination, family, finances, safety or past adverse experience, which is more studies than students expect.
6. **What data will exist, in what form, where, for how long, and who can access it?** Determines the data management section, which is where most incompleteness rejections happen. Default: enumerate every artefact including recordings, transcripts, consent forms, contact lists and the linking key, since students routinely account for the transcripts and forget the recordings.
7. **Will any AI or third-party processing touch participant data?** Determines a disclosure that committees now expect and that some institutions restrict. Default: assume it must be disclosed, and assume that transcription, translation, coding assistance and analysis support all count.

## 7. Step-by-step methodology

**Step 1. Establish the route and build the timeline backwards.**
Find the committee, its form, its guidance, its meeting dates and its submission deadlines. Establish the review level your study requires, using the committee's own criteria rather than your judgement of how risky the study feels. Then build the timeline backwards from the date fieldwork must start: fieldwork start, minus contingency for a second revision cycle, minus the interval between submission and decision, minus the supervisor's review time, minus the time to write the application. In most institutions this produces a submission date four to ten weeks before fieldwork, and many students discover at this step that their plan was never possible. Fix the plan now rather than in week nine. Note also the practices that require approval before they happen: piloting, recruitment advertising, and often the approach to a gatekeeper organisation. *Correct result: a dated backward timeline with the submission date fixed, and a list of activities that cannot begin before approval.* `RESEARCHER DECISION REQUIRED` where the timeline does not fit the dissertation deadline (K5 §2.7).

**Step 2. Describe the study in the committee's terms, not the supervisor's.**
The committee is not expert in your topic and is not assessing its academic merit. It is assessing risk to participants, adequacy of consent, and data handling. Write the study description accordingly: what will actually happen to a participant, in sequence, in plain language, with numbers and durations. "A qualitative exploration of lived experience through an interpretive phenomenological lens" tells a committee nothing; "up to 20 adults will be interviewed once, for 45 to 60 minutes, in a private room at their workplace or by video call, audio recorded with consent" tells them everything they need. Include the aim in one sentence, because a study whose purpose is unclear cannot have its risk-benefit balance assessed. *Correct result: a study description a non-specialist could use to picture a participant's entire experience.*

**Step 3. Assess risk proportionately, in four directions, and calibrate.**
Enumerate risks to **participants** (physical, psychological, social, legal, economic, reputational, and the risk of the research relationship itself), to the **researcher** (lone working, travel, exposure to distressing material, dual-role strain), to **third parties** (people discussed but not participating, the participant's employer or family), and to **data** (loss, breach, re-identification). For each: state the risk, its likelihood, its severity, and the specific mitigation. Then calibrate against the everyday standard, which is the benchmark most committees use: does this exceed the risk a person encounters in ordinary daily life or in routine professional activity? Two errors to avoid. **Inflation**: writing a 30-minute interview about software preferences as though it carried a serious psychological risk, which suggests you cannot judge risk and invites the committee to look harder at everything. **Dismissal**: writing "there are no risks in this study" for research on workplace bullying, which is the single fastest route to a full-committee referral. Where risk is genuinely minimal, say so and say why, briefly. Where it is real, name it precisely and mitigate it specifically. *Correct result: a four-direction risk table with likelihood, severity and a specific mitigation per row, and a calibration statement against the everyday standard.*

**Step 4. Write the recruitment materials and resolve the inducement question.**
Recruitment materials (poster, email, social media text, gatekeeper letter, verbal script) are part of the submission and must be consistent with the information sheet. Three things get checked. **Voluntariness**: the invitation must make clear that participation is optional, that declining carries no consequence, and, where the researcher has any authority over the participant, that the researcher will not know who declined, which usually requires an intermediary in recruitment. **Accuracy**: recruitment material cannot describe the study more attractively than the information sheet describes it. **Inducement**: an incentive that merely compensates for time and expense is acceptable; one large enough that a person would accept a risk they would otherwise refuse is undue inducement. The test is proportionality to the participant's circumstances, not the absolute amount, so the same payment can be reasonable for one population and coercive for another. State the amount, the form, the timing, and explicitly that it is not conditional on completing the study, since making payment conditional on completion undermines the right to withdraw. Where you recruit through a gatekeeper organisation, state how the gatekeeper's permission was obtained and how you will prevent the gatekeeper's endorsement from becoming pressure. *Correct result: a consistent recruitment set, with the incentive stated in all documents and its non-conditionality explicit.*

**Step 5. Write the participant information sheet at the reading level of the participants.**
Aim for the level of a general newspaper unless your participants are a specialist group, and lower where the group's literacy or language may vary. Short sentences, no methodological vocabulary, no acronyms, active voice, and questions as headings. The content is fixed by convention and by what a person needs to decide: what the study is about and who is doing it; why they have been invited; that participation is voluntary; what will happen to them, concretely, and how long it will take; what the possible risks and discomforts are, honestly; what the benefits are, without overstating them, since most studies offer no direct benefit and saying so is correct; how their data will be used, stored and protected; whether they will be identifiable in outputs and how quotes will be handled; how long data will be kept and what happens to it then; their right to withdraw, including the point after which withdrawal is no longer possible and why; who to contact with questions and who to complain to, which must be someone other than the researcher; and the ethics approval reference. Two things the sheet must never do: promise anonymity in a design that cannot deliver it, and describe a procedure that differs in any detail from the protocol. *Correct result: an information sheet a participant from your group could read once and act on, containing every required element, matching the protocol exactly.*

**Step 6. Write a consent form that matches what will actually happen.**
The consent form is not a summary of the information sheet, it is a list of the specific things the person is agreeing to, each separately affirmable. Separate items matter, because bundling them removes real choices: consent to participate, to be audio or video recorded, to be quoted anonymously, to be quoted with an identifier such as a role, to the data being retained for a stated period, to being contacted for follow-up or member checking, and, where applicable, to the data being used in future research or archived. Where any of these is optional, the form must allow a participant to decline one and still take part, and the procedure must actually honour that, which means you need a plan for the participant who consents to interview but not to recording. State the withdrawal position precisely and truthfully: participants can withdraw at any time during the study, and their data can be removed up to a stated point (typically transcription, anonymisation or analysis), after which removal is impossible because the data can no longer be identified or has been aggregated. Saying "you may withdraw your data at any time" is usually false and is the commonest inaccuracy in student consent forms. Where consent is verbal (telephone, low-literacy contexts, cultural appropriateness), state how it will be recorded and witnessed. Where participants are under the age of legal consent in the jurisdiction, state the parental or guardian consent arrangement and the child's assent process separately. *Correct result: an itemised consent form whose every line corresponds to something that will actually happen, with a truthful withdrawal statement.*

**Step 7. Specify data management across the full lifecycle, artefact by artefact.**
Enumerate every data artefact the study will produce, then for each state: what it is, where it is stored, in what format, who has access, how it is transferred, how long it is retained, and how it is destroyed. Students routinely account for transcripts and forget the audio recordings, the consent forms, the contact list, the linking key between pseudonyms and identities, the recruitment correspondence and the copies on a personal device. Address specifically: how quickly recordings move off the recording device; whether any processing happens outside the institution's systems, including transcription and translation services; the linking key, which must be stored separately and destroyed at a stated point; and cloud storage location, which some institutions and jurisdictions restrict. State the lawful basis for processing where your jurisdiction requires it, in the institution's own wording. Retention periods are set by policy, not preference, so find the policy. *Correct result: a data table with one row per artefact and no blank cells, matching institutional policy on retention and storage.*

**Step 8. Address anonymisation honestly, including re-identification risk.**
Distinguish the three states, because using the wrong word is a substantive error. **Identifiable**: direct identifiers present. **Pseudonymised**: identifiers replaced with codes, with a key that still exists, so the data remains personal data and the participant can still be re-identified by you. **Anonymised**: no key exists and re-identification is not reasonably possible by anyone. Most student research is pseudonymised and describes itself as anonymised, which is inaccurate and matters because the two carry different obligations. Then assess re-identification risk from the reported output, which is where the real exposure lies in small and specialist samples. The mechanisms: a combination of role, tenure, department and demographic details unique within an organisation; a quote containing a distinctive turn of phrase or a specific incident colleagues would recognise; a participant characteristics table that, read alongside a quote, identifies the speaker; and the recruitment route, which tells readers who the pool was. Mitigations, stated in the application: reporting role at a coarser grain, generalising organisational and geographic detail, aggregating characteristics rather than tabulating them per participant, editing quotes to remove distinctive detail with the convention stated, and, in the highest risk cases, offering participants sight of their attributed quotes before publication. Where a participant cannot be protected from identification by colleagues, say so in the information sheet, because that lets the person decide. *Correct result: the correct term used throughout, a stated re-identification assessment for this specific sample, and named mitigations carried into the reporting plan.*

**Step 9. Apply the additional protections each vulnerable category requires.**
Vulnerability here is situational as much as categorical, and the question is always what specifically reduces this person's capacity to give free, informed consent or increases the harm the research could do. Children and young people: legal consent age in the jurisdiction, guardian consent plus age-appropriate assent materials, safeguarding clearance, a second adult present, and a disclosure protocol. Adults who may lack capacity: capacity assessment, the jurisdiction's legal framework for consultee or proxy arrangements, ongoing assent, and materials in accessible formats. People in institutional settings: gatekeeper permission plus specific protection against the gatekeeper's endorsement functioning as coercion, and a route to decline that staff cannot observe. Employees, students and clients of the researcher: recruitment via an intermediary, the researcher blind to who declined, and assessment consequences explicitly disconnected. People with insecure status of any kind: minimal data collection, no collection of the status itself unless essential, and an account of what happens if data is subject to legal demand. People discussing traumatic experience: distress protocol, control over pacing, and the right to stop without explanation. Where more than one applies, apply all. *Correct result: for each applicable category, a named additional protection, not a general statement of care.*

**Step 10. Build a distress protocol with a named escalation route.**
A distress protocol that does not name a service and a person is not a protocol. Write it as a sequence with thresholds: what indicates mild discomfort and what you do (pause, offer a break, remind of the right to skip or stop); what indicates significant distress and what you do (stop the recording, stop the interview, remain with the participant, offer the support sheet); and what constitutes a disclosure requiring action, which is where the protocol becomes serious. Disclosures of risk of harm, or of a safeguarding concern, cannot be held in confidence, and the information sheet must say so in advance in specific terms, because a confidentiality promise the researcher will break on discovering a safeguarding issue is a promise that should never have been made. Name the escalation route: the supervisor, then the named institutional contact, with what happens and how quickly. Provide participant-facing support information as a separate sheet naming a service appropriate to the population and topic, checked as currently operating. Include the researcher: who they debrief with after a difficult interview, since researcher distress in sensitive-topic work is real and committees look for it. *Correct result: a threshold-based protocol naming the actions, the people and the services, with the confidentiality limits carried into the information sheet.*

**Step 11. Establish whether secondary data use is covered by the original consent.**
For any existing dataset about people, the question is not whether the data is available to you but whether the people agreed to this. Establish: who collected it, under what consent, what that consent said about future use and sharing, whether the data is identifiable or pseudonymised in your hands, what the data sharing agreement or licence permits, and whether an ethics approval already covers this use. Three outcomes. The original consent explicitly covers further research use, in which case cite it. The original consent is silent or narrower, in which case you need either an approval that addresses the mismatch, or consent from the data controller with a documented justification, and the committee will ask about proportionality. The data is genuinely anonymous and openly published, in which case the review requirement may not apply, but check rather than assume, particularly for social media and web-scraped content, where the fact that data is public does not establish that its subjects consented to research use and where institutional positions differ sharply. *Correct result: a documented statement of the original consent's scope and an explicit finding on whether it covers this use.* `RESEARCHER DECISION REQUIRED` where the consent does not clearly cover the use (K5 §2.4).

**Step 12. Disclose AI and third-party processing of participant data.**
State plainly which tools will process participant data and at which step: transcription, translation, coding or analysis assistance, summarisation, or writing support that involves pasting participant data. For each, state where the data goes, whether it leaves the institution's controlled environment, whether it may be retained or used for model training by the provider, and what the participants have been told. Three rules that committees increasingly apply. Participants must be told in the information sheet if their data will be processed by an external automated service, in plain terms. Identifiable data should not be sent to a service that retains it or trains on it, and where a service is used, the institutional position on approved tools governs. And a human must verify any AI-produced transcription, coding or classification that feeds a finding, with that verification described (per K4 §7 and K5 §6). Where the institution has an approved-tools list, use it and say so. Where AI will not touch participant data, say that explicitly, since silence now reads as omission. *Correct result: an explicit statement of AI and third-party processing, matched by a plain-language sentence in the information sheet, or an explicit statement that none will occur.*

**Step 13. Run the cross-document consistency check, which is what actually causes rejections.**
Read every document against every other and check that these match exactly: study title, participant numbers, procedure duration, what is recorded and how, the incentive and its conditions, the withdrawal deadline, the retention period, the anonymity claim, the contact names and the complaints route. Then check the application against the methodology chapter, which must not diverge. The classic failures: the poster says 30 minutes and the information sheet says an hour; the consent form offers withdrawal at any time and the protocol says data is anonymised at transcription; the application says 20 participants and the sampling section says 25; the information sheet promises anonymity and the design reports participants by job title in a single organisation. Each of these individually triggers a revision cycle, and a cycle is typically a month. Then check completeness against the form's own checklist, since committees reject on missing appendices more often than on substance. *Correct result: a consistency matrix across all documents with every field matching, and a completed submission checklist.*

**Step 14. Respond to requested revisions specifically and without re-litigating.**
Committee feedback is usually a numbered list. Respond with a numbered response document that mirrors it exactly: the comment, what you changed, and where to find the change (document, section, page). Change the documents themselves and mark the changes visibly. Where you disagree with a request, comply if it does not damage the study, and where it genuinely would, explain the methodological reason briefly and propose an alternative that meets the committee's underlying concern, since the concern is nearly always legitimate even when the requested remedy is not. Never argue that the committee has misunderstood without also making the passage clearer, because if they misunderstood it, so will participants. Resubmit within the same cycle where possible. *Correct result: a point-by-point response document, revised materials with changes marked, and no unaddressed comment.*

## 8. Analytical framework

**The consistency triangle.** Three things must describe the same study, and the committee's core test is whether they do:

    What the protocol says will happen
      ↔ What the information sheet tells the participant will happen
        ↔ What the consent form asks the participant to agree to

Applying it: every element of the study appears in all three, in compatible terms, at the appropriate register. A promise in one that the others cannot support is the defect that causes most rejections, and it is found by reading the three documents side by side rather than in sequence. The most common breaks are anonymity (promised in the sheet, impossible in the design), withdrawal (offered without limit in the form, bounded in the protocol) and recording (assumed in the protocol, optional in the form with no plan for the person who declines).

**The proportionality ladder.** Applied to every protection in the application:

    Risk identified → Likelihood and severity assessed →
    Mitigation proportionate to both → Residual risk stated →
    Participant told, in terms they can act on

Applying it: protections are justified by the risks they address, not by their impressiveness. A mitigation with no risk above it is inflation and wastes the committee's attention; a risk with no mitigation below it is the gap the committee will find; and a residual risk not communicated to the participant is a consent failure, because the participant agreed to a study they were not accurately described.

## 9. Output format

The submission is a set of documents, and the set is the output.

**1. The institutional form**, completed in full, in its own structure and terminology.

**2. Study description.** Aim in one sentence, participants, procedure in sequence, duration, setting, data collected.

**3. Risk assessment.**

| Direction | Risk | Likelihood | Severity | Mitigation | Residual risk |
|---|---|---|---|---|---|
| Participant | | | | | |
| Researcher | | | | | |
| Third party | | | | | |
| Data | | | | | |

**4. Recruitment materials.** Poster, email or script, and gatekeeper letter, with the incentive stated identically in each.

**5. Participant information sheet.** Plain language, question headings, every required element.

**6. Consent form.** Itemised, separately affirmable, with a truthful withdrawal statement.

**7. Data management plan.**

| Artefact | Format | Storage location | Access | Transfer method | Retention period | Destruction method |
|---|---|---|---|---|---|---|

**8. Anonymisation and re-identification assessment.** The correct term for the data's state, the specific re-identification mechanisms for this sample, and the reporting mitigations.

**9. Vulnerable participant protections**, per applicable category, with the specific protection named.

**10. Distress and disclosure protocol.** Thresholds, actions, named escalation route, participant support sheet with a checked service.

**11. Secondary data consent statement**, where applicable.

**12. AI and third-party processing statement.**

**13. Instrument.** Full questionnaire, interview guide or observation schedule.

**14. Consistency matrix and submission checklist.**

**When something is not yet decided, the format must not force invention (K4 §1).** An unresolved element is stated as unresolved with the decision route named, or the submission waits. An invented retention period, an unchecked support service, a support organisation that no longer operates, or a fabricated data storage arrangement are all worse than a blank, because the application is a set of commitments the study will be held to, and a committee's approval is granted on the basis that what is written is true.

## 10. Quality checks

Run before the supervisor sees it. These sit on top of K4 §8.

1. Does the study description let a non-specialist picture a participant's entire experience?
2. Is every risk in the four directions named, with a proportionate mitigation and a stated residual?
3. Is the risk assessment calibrated, neither dismissing real risk nor inflating trivial risk?
4. Do the protocol, the information sheet and the consent form describe the same study, in every detail that appears in more than one?
5. Is the information sheet readable by the participants, not by the marker?
6. Does the consent form itemise separately affirmable elements, and can a participant decline one and still take part?
7. Is the withdrawal statement true, with a stated cut-off point and a reason?
8. Is the word "anonymous" used only where no key exists, and is pseudonymised data called pseudonymised?
9. Has re-identification been assessed for this specific sample and reporting plan, not in general?
10. Does every applicable vulnerable category have a named additional protection rather than a general assurance?
11. Does the distress protocol name a service, a person and a threshold, and has the service been checked as operating?
12. Are the limits of confidentiality stated in the information sheet before consent, not discovered at disclosure?
13. Does the data table have a row per artefact, including recordings, consent forms and the linking key, with no blank cells?
14. Are the retention period and storage location taken from institutional policy rather than chosen?
15. Is AI and third-party processing disclosed in both the application and the information sheet, or explicitly ruled out?
16. Is the incentive stated identically everywhere, and explicitly not conditional on completion?
17. Is every appendix the form requires actually attached?
18. Does the timeline still work, given the committee's meeting dates?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Incompleteness** | A blank field, a missing appendix, an unattached instrument | Complete the form's own checklist last (Step 13) |
| **Document inconsistency** | The poster and the sheet give different durations | Cross-document consistency matrix (Step 13) |
| **Risk inflation** | A low-risk study written as though it were clinical | Calibrate against the everyday standard (Step 3) |
| **Risk dismissal** | "There are no risks" on a sensitive topic | Enumerate in four directions before concluding (Step 3) |
| **Reading level mismatch** | Methodological vocabulary in the information sheet | Write for the participants, test it on someone outside the field (Step 5) |
| **The false anonymity promise** | Anonymity promised, participants reported by role in one organisation | Assess re-identification for this sample (Step 8) |
| **Untrue withdrawal statement** | "Withdraw your data at any time" with no cut-off | State the real limit and the reason (Step 6) |
| **Bundled consent** | One tick box covering participation, recording and quotation | Itemise, and plan for partial consent (Step 6) |
| **Forgotten artefacts** | Recordings, consent forms or the linking key absent from the data plan | One row per artefact (Step 7) |
| **The unnamed escalation route** | "Participants will be signposted to support" | Name the service, the person, the threshold (Step 10) |
| **Coercive recruitment** | The researcher recruits their own students or staff directly | Intermediary recruitment, researcher blind to refusals (Steps 4, 9) |
| **Conditional incentive** | Payment on completion, undermining withdrawal | Non-conditional, stated as such (Step 4) |
| **Assumed secondary consent** | Public data used because it was available | Establish the original consent's scope (Step 11) |
| **Undisclosed AI processing** | Transcription sent to an external service, participants not told | Disclose in the application and the sheet (Step 12) |
| **Timeline collapse** | Submission planned for the week before fieldwork | Build the timeline backwards first (Step 1) |
| **AI-invented support service** | A plausibly named helpline that does not exist | Never generate a service, a contact or a policy (§12) |
| **Template drift** | Content written to a generic template, transferred badly | Write into the institution's own form from the start (Step 1) |

## 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, who normally co-signs the submission and carries formal responsibility for it.

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 supply the checklists, the risk enumeration structure, the required-elements lists, the consistency matrix and a critique of the student's drafts, and it will point out where a document contradicts another. Note additionally that some institutions require ethics materials to be the applicant's own composition, and where that is so, the assistance stops at structure and critique.

2. **Never invent an institutional fact.** Not a retention period, not a storage location, not a review level, not a committee deadline, not a lawful basis, not an ethics reference number, not a policy position. Every one of these is a property of a specific institution, and a plausible answer is a commitment the student will be held to and cannot honour. Name what must be found and where to look for it.

3. **Never generate a support service, helpline, referral contact or safeguarding lead.** A distress protocol pointing to a service that does not exist, or that closed, or that does not serve this population, is worse than no protocol, because it will be relied on at the moment it matters. Services are found and verified by the student, and checked for current operation.

4. **Never write a consent statement, an anonymity promise or a confidentiality assurance that the described design cannot deliver.** If the design reports participants by role within one organisation, the materials cannot promise anonymity, and the correct action is to say so and change either the promise or the design.

5. **Never assert that a study does not require ethics review.** Review requirements are institutional and jurisdictional. The correct output is the question to ask and who to ask it of.

6. **Never assess capacity, vulnerability or risk for a specific individual or group on the student's behalf as a final judgement.** Supply the categories, the questions each raises and the protections each conventionally requires; the applied judgement belongs to the student and the committee (K5 §2.4).

7. **Never help conceal a change from an approved protocol.** Where the study has deviated from what was approved, the correct route is an amendment or a reported deviation, and assisting with either is legitimate. Writing the application or the methodology so that the divergence is invisible is not.

8. **Never draft a response to a committee that argues the committee is wrong without also fixing the underlying clarity problem.** If a reviewer misread a passage, participants will too.

9. **Never fabricate the study's procedure to fit the form.** Where the form asks something the study has not decided, the answer is that it is undecided, together with what must be decided and by whom. An application is a set of commitments, and a fabricated commitment is a governance failure rather than a drafting shortcut.

## 13. Best-practice principles

- **Most rejections are for completeness and consistency, not ethics.** Read the three core documents side by side, not in sequence, and the majority of defects appear immediately.
- **Find a recently approved application from your department.** It is worth more than any amount of general guidance, because it shows what this committee accepts.
- **Write the information sheet first, and let it discipline the protocol.** Anything you cannot explain simply to a participant is usually something you have not fully decided.
- **Proportionality runs in both directions.** Under-stating risk gets you referred; over-stating it gets you scrutinised. Calibrate against ordinary life and say which way you calibrated.
- **A protection is a named action, a named person and a threshold.** "Care will be taken" is not a protection.
- **Never promise anonymity you cannot deliver.** In a small organisation, colleagues identify each other from a quote and a job title, and the honest position is to say so and let people decide.
- **Confidentiality limits are disclosed before consent, not at the point of disclosure.** A promise you intend to break under safeguarding conditions is not a promise.
- **Approval is a dependency with a fixed rhythm.** Committees meet on dates you cannot influence, so the plan bends around them.
- **Amendments are cheap; unreported deviations are not.** A protocol change reported and approved costs a form; the same change discovered later costs the study's standing.
- **The application and the methodology chapter must remain the same document in two registers.** When one changes, change the other the same day.
- **Include the researcher in the risk assessment.** Lone working, distressing material and dual-role strain are real, and committees notice when they are absent.
- **Keep the approved bundle.** The reference number, the approved documents and the date go into the methodology chapter and the appendix, and they are the evidence that the study was conducted as approved.

## 14. Worked example

Generic fictional scenario, academic, in a non-governmental organisation context.

**INPUT**

A masters candidate in development studies plans interviews with 18 frontline staff of a humanitarian organisation about their experience of a restructuring, including its effects on morale and workload. The candidate previously worked for the organisation and has been offered access by a country director who is enthusiastic about the study. The interviews will be conducted by video call, recorded and transcribed with an external transcription service. Fieldwork is planned to start in six weeks.

**PROCESS**

*Step 1.* The committee meets monthly, with a submission deadline ten days before each meeting. The next deadline is in eight days, and the following one is five weeks away. Working backwards, submitting to the next deadline gives a decision in eighteen days, leaving room for one revision cycle before the planned fieldwork start. Missing it makes the plan impossible. This is stated to the supervisor immediately.

*Step 3.* Risk enumeration. Participants: reputational and employment risk, since staff criticising a restructuring in an organisation where the country director sponsored the study is the central risk and is not minor. Researcher: exposure to accounts of distress, and dual-role strain as a former colleague. Third parties: named managers discussed but not participating. Data: recordings and transcripts, plus the external transcription step.

*The judgement call.* Two problems surface together at Step 4 and Step 9. The country director's enthusiasm is a recruitment risk, not an asset: if the director circulates the invitation and knows who volunteered, participation is not voluntary in any meaningful sense, and staff will reasonably assume their responses may reach management. And with 18 participants in one country office, role plus tenure plus team will identify many of them. The candidate's first instinct is to promise strict confidentiality and proceed. That promise cannot be kept by the design.

The resolution has three parts. Recruitment moves away from the director: the organisation provides a staff distribution list, the invitation comes from the candidate directly, responses go to the candidate only, and the application states explicitly that the organisation will not be told who participated or declined, with the country director's written agreement to that condition attached. Reporting protections are specified in advance: no job titles, roles reported only at three coarse levels, no team or location detail, quotes edited to remove identifying incident detail with the convention stated, and no participant characteristics table. And the information sheet says honestly that although the researcher will not identify anyone, colleagues might recognise an account, so participants should consider that before agreeing, and may ask for any quote to be withheld.

*Step 6.* The consent form is itemised: participation, recording, anonymous quotation, retention for the stated period, and a separate item for being contacted to review quotes. Recording is optional, and the protocol includes a note-taking fallback. Withdrawal is stated truthfully: data can be removed until anonymisation, which occurs at transcript checking, approximately two weeks after interview, after which the link to the individual no longer exists.

*Steps 7 and 12.* The data table lists recordings, transcripts, consent forms, the contact list and the linking key. The external transcription service is examined and fails the check: it retains uploaded audio. The plan changes to transcription within the institution's approved environment, and the information sheet gains a plain sentence stating that recordings are transcribed using a tool approved by the university and are not shared with anyone outside the research team. The candidate also states that no analysis tool will process identifiable data, and that any AI-assisted coding will run on anonymised transcripts and be verified by the candidate against the source.

*Step 10.* The distress protocol names thresholds and actions, names the supervisor as the escalation route, and provides a participant support sheet listing the organisation's own staff welfare service and one independent service, both checked as currently operating. It also states that the candidate will debrief with the supervisor after any difficult interview.

**OUTPUT**

A submission comprising the completed institutional form, a study description, a four-direction risk table, recruitment materials routed away from the sponsor, an information sheet at general reading level that is honest about recognisability, an itemised consent form with a truthful withdrawal cut-off, a full data table, a re-identification assessment with reporting rules carried into the analysis plan, a distress protocol with named services, an AI processing statement, the interview guide, and a consistency matrix. The committee approves with one minor revision, a request to add the complaints contact to the recruitment email, which is returned within two days.

`RESEARCHER DECISION REQUIRED` was raised at Step 4 on the recruitment route, since proceeding through the sponsor would have made consent non-voluntary (K5 §2.4), and `RESEARCHER SIGN-OFF REQUIRED` on the full submission before it left the department.

## 15. Advanced usage

**Designing for approvability from the start.** Where a design decision has an ethics cost, knowing it early is worth more than solving it late. Sponsor-routed recruitment, identifiable reporting, incentives that scale with vulnerability and external processing of identifiable data are all easier to design around than to defend. Reading the committee's criteria before finalising the methodology is an hour that routinely saves a month.

**The amendment habit.** Studies change: a site withdraws, a population is added, an instrument is revised, the mode shifts. Learn the amendment route early and use it, because a study conducted outside its approval is a governance problem that surfaces at examination and can affect publication. Keep a running list of changes and submit them in batches where the committee allows it.

**Multi-site and gatekeeper layers.** Where a study spans organisations, each may require its own permission, and their timelines are independent and often longer than the university's. Map the permissions as a dependency chain, and start the slowest first.

**Ethics materials as a reporting asset.** The information sheet, consent form and data plan feed directly into the methodology chapter's ethics section and into the appendix, and the re-identification assessment becomes the reporting rules for the analysis chapter. Written properly once, they are used three times. See **15.10 Data Analysis Chapter Development** for how the reporting rules constrain quote selection.

**Where the standard approach does not fit.** Research with communities that have their own governance and consent traditions may require collective as well as individual consent, and the university's forms rarely anticipate this; document the additional process and explain it to the committee. Research in emergency or rapidly changing settings may need a protocol with pre-approved flexibility rather than a fixed procedure. Online and social media research raises questions the forms often handle badly, particularly around what public availability implies about consent; where the institution has no position, say so and reason it explicitly rather than assuming permission.

## 16. Skill chain

**Recommended previous skills:**
- **15.08 Research Design and Methodology Chapter.** Hands over the settled procedure, population, instrument and data plan the application restates.
- **13.05 Research Ethics and Consent Design.** Hands over the substantive ethical reasoning about consent, risk and participant protection that this skill documents.
- **01.06 Sampling Strategy.** Hands over the recruitment route, which is where voluntariness problems originate.

**Recommended next skills:**
- **15.10 Data Analysis Chapter Development.** Takes the re-identification assessment as the binding rules on quote selection and participant description.
- **13.06 AI Research Governance.** Takes the AI processing statement and the verification obligations into practice during analysis.
- **15.11 Discussion and Contribution Development.** Takes the reporting constraints into the write-up of findings about identifiable groups.

**Runs well alongside:**
- **02.03 Interview Question Development**, since the instrument is submitted with the application and its content affects the risk assessment.
- **03.01 Recruitment and Sample Sourcing**, for the practical recruitment route the application must describe.
- **15.08**, continuously, since the application and the methodology chapter must remain consistent throughout.

---
A Yazi Supplied Skill and resource.