Research Ethics and Consent Design

What makes consent genuinely informed?

Six elements, each present, specific and comprehensible: the purpose in concrete terms, what is collected including recording and metadata, how it is used including whether verbatim words may be published, who sees it in identifiable form, how long it is kept and what triggers deletion, and the right to withdraw with its real limits. The standard is comprehension, not disclosure: the practical test is whether a participant from the actual population could restate those six elements in their own words afterwards.

Do I need consent to run participant data through an AI system?

If the consent does not mention it, it does not cover it. Tell participants that a system will transcribe, code, summarise, moderate or analyse their words, what kind of processing that is, and what a human does with the result. De-identification does not remove the obligation, because participants care about what reads their words and are entitled to know. Consent given before AI processing was contemplated cannot be stretched to cover it.

How do I know if my sample is too small to be anonymous?

Write out the attribution string you intend to publish and ask how many people in the world it describes. "Head of procurement, mid-size logistics firm, Nordic market" is a name to anyone in that industry, and the industry is who reads the report. If the answer is one person or a handful, broaden the attribution, aggregate to theme and prevalence, paraphrase and label it as the analyst's summary, seek specific consent to be identifiable, or do not report it.

What extra protections do vulnerable participants need?

It depends which vulnerability. Children need consent from the person with parental responsibility plus the child's own assent as a real veto, and materials at their reading level. Patients need recruitment that does not run through the treating clinician in a way implying care is affected. Employees reporting on their employer need recruitment management cannot observe, a true statement of what the employer will see, and reporting thresholds that prevent identification by team or tenure. People in distress need an incentive that is not the reason for participating and topic limits that hold even when they volunteer more.

Who counts as a vulnerable participant in commercial research?

More people than usually get treated as one. Vulnerability is situational rather than a property of a person, and it comes from capacity to consent, dependence on someone in the participation chain, and exposure to harm from taking part. A senior professional discussing their employer, in a study the employer commissioned and will read, is a vulnerable participant on the second and third grounds and is almost never treated as one.

What should a distress protocol contain?

A stop rule for when a participant becomes distressed, beginning with stopping rather than continuing gently. A disclosure rule stating in the consent what happens if harm is disclosed, since a promise of confidentiality that cannot be kept is worse than an honest limit. Signposting to support prepared in advance and appropriate to the market. A named human, reachable during fieldwork hours, with authority to stop a session. And support for the researcher, since exposure to distressing material accumulates. A protocol nobody can execute is worse than none.

Can AI moderate interviews on sensitive topics?

Not without designed human escalation. An automated moderator cannot reliably recognise distress, judge capacity, exercise a stop rule with judgement, or take responsibility. The requirements are a low-threshold escalation trigger, a route to a real person, a visible always-available exit, and a clear statement that the participant is speaking to a system and how to reach a human. A sensitive-topic study that cannot provide human escalation should not be run this way.

How much should I pay research participants?

Enough to recognise their time and costs, benchmarked against the participant's circumstances rather than the study's budget. It becomes an inducement when it is large enough that someone accepts a risk or discomfort they would otherwise decline, and the same amount is a token to one population and a week's food to another. Make it unconditional on the content of answers, payable at least pro rata if the participant stops early, and state it up front.

Can participants really withdraw after the research is finished?

Partly, and the honest thing is to say which parts in the consent. Data can be removed from a dataset up to the point analysis is run. Once a report is published, a contribution cannot be extracted from an aggregate and a published quote cannot be recalled. So state a deadline as a date or an event, and state what withdrawal still delivers after it: deletion of identifiable records and recordings, no further use, and no inclusion in anything not yet produced.

How do I decide whether research should be done at all?

Four tests, and failing one is enough. Necessity: can the question be answered without asking these people this. Proportionality: is the burden and risk proportionate to what will be learned, weighed for the participant rather than the commissioner. Harm to the group: could the findings be used against the people who provided them. And the refusal test: would a fully informed participant agree, or does the design depend on them not fully understanding. Some research should not be conducted, and a discipline that cannot say so has no ethics in it.