Deception and Debriefing in Research: When a Cover Story Is Justified and What You Owe Participants

Answer first: Masking and deception are not the same act, and conflating them causes teams to either over-disclose and ruin their data or under-disclose and breach a code they never read. Masking means withholding the specific hypothesis while giving an honest account of the study's general purpose. Every major research code licenses it, and good design usually requires it. Deception means asserting something that is not true, and it triggers a justification test, a debriefing duty, and a cost you pay in every study you run afterwards. The practical rule: you may decline to say what you expect to find; you may not say something false about what is happening.

The disclosure ladder

Most real research decisions sit somewhere on a five-rung ladder, and it is worth being able to name which rung you are on.

Rungs 1 to 3 are the working range for almost all commercial research. Rungs 4 and 5 are where the codes impose tests, and the honest answer is that most product teams who think they need rung 4 have a rung 2 design problem they have not solved.

What the codes actually require

The ICC/ESOMAR International Code is the global self-regulatory benchmark for market, opinion and social research, and its wording is precise in ways worth quoting.

Key Fundamental 4 states that respondents cooperation is voluntary and must be based on adequate, and not misleading, information about the general purpose and nature of the project when their agreement to participate is being obtained, and that all such statements shall be honoured. The word general is the licence for masking. You owe an honest account of what the study is about. You do not owe your hypothesis.

Article 3(a) states that respondents cooperation is entirely voluntary at all stages, and that they shall not be misled when being asked for their co-operation. Note where the prohibition attaches: to the recruitment moment. The code is protecting the integrity of the decision to take part, which is exactly the decision a false premise corrupts.

Article 4(a) requires that researchers shall promptly identify themselves and unambiguously state the purpose of the research, and Article 4(b) that respondents shall be able to check the identity and bona fides of the researcher without difficulty. Article 1(d) requires that market research shall be clearly distinguished and separated from non-research activities including any commercial activity directed at individual respondents. Article 6 requires respondents to be informed before observation techniques or recording equipment are used, and gives them the right to have the record destroyed.

The APA Ethical Principles give you the justification test for rung 4. Standard 8.07(a): psychologists do not conduct a study involving deception unless they have determined that the use of deceptive techniques is justified by the study's significant prospective scientific, educational, or applied value and that effective nondeceptive alternative procedures are not feasible. Standard 8.07(b): they do not deceive prospective participants about research that is reasonably expected to cause physical pain or severe emotional distress. Standard 8.07(c): they explain any deception that is an integral feature of the design as early as is feasible, preferably at the conclusion of their participation, but no later than at the conclusion of the data collection, and permit participants to withdraw their data.

Standard 8.08 covers debriefing generally: provide a prompt opportunity for participants to obtain appropriate information about the nature, results and conclusions of the research, and take reasonable steps to correct any misconceptions of which you are aware. Where scientific or humane values justify delaying or withholding that information, take reasonable measures to reduce the risk of harm. Where you become aware that procedures have harmed a participant, take reasonable steps to minimise the harm.

Read together, 8.07(a) is a two-part test and the second part is the one teams skip. It is not enough that the study is valuable. You have to be able to say what nondeceptive alternatives you considered and why they were not feasible. That is a documentable answer, and it belongs in your study plan.

Sugging is not merely unethical

There is one form of research deception that is straightforwardly unlawful in the UK, and it is the one commercial teams drift into without noticing.

Sugging is selling under the guise of research: a contact presented as a survey or interview whose real function is to qualify a lead, build a marketing list, or open a sales conversation. ESOMAR Article 1(d) prohibits it as a matter of code. UK law goes further. Schedule 20 of the Digital Markets, Competition and Consumers Act 2024, in force since 6 April 2025, lists practices that are unfair in all circumstances, requiring no proof that anyone was actually misled. Paragraph 25 covers falsely claiming or creating the false impression that the trader is not acting for purposes relating to the trader's business, or falsely representing oneself as a consumer. Paragraph 12 covers using editorial content to promote a product where the trader has paid for the promotion without making that clear.

The practical test is what happens to the data. If a research contact feeds a CRM record that a sales team will action, it was not research, whatever the subject line said. ESOMAR Key Fundamental 6 is unambiguous: researchers shall never allow personal data collected in a market research project to be used for any purpose other than market research.

This matters even for teams with no intention of sugging, because the drift is structural. A product team runs discovery interviews, the interviews surface three enthusiastic prospects, and somebody forwards them to sales. That is the breach, and it happened after the research was over.

The contamination cost nobody prices

The strongest argument against casual deception is not ethical, it is methodological, and there is data on it.

The psychology literature ran a decades-long natural experiment. Analyses summarised by Ortmann and Hertwig in their review of the costs of deception show the proportion of deception studies in the leading social psychology journal and its predecessor rising steadily: an average of 5 percent of articles between 1921 and 1948, 9 percent in 1948, 51 percent by 1968, peaking around 69 percent in 1975, still 50 percent in 1983, declining to 31 percent in 1994, then back up to 47 percent in 1992 on one analysis and 42 percent in 1996 on another.

Participant suspicion tracked it. Stang's exhaustive review of the conformity literature identified 21 studies reporting the proportion of participants classified as suspicious, and found that until the mid-1960s roughly one in ten participants reported being suspicious, with that number rising to between 40 and 60 percent for the period 1967 to 1973. In Stang's own 1976 study, 20 percent of participants were excluded from analysis based on self-reported suspicion.

The measurement itself was probably too generous. Stricker's review of 88 deception studies across four leading journals found that only 16 even attempted to quantify participant suspicion, with a median of 4 percent and a range from zero to 23 percent, and the authors later concluded that the overall rate of actual suspicion in those 16 studies was seriously underestimated because the criteria used were inadequate.

Two things follow, and the second is the one that should change your behaviour.

First, suspicion is not a nuisance variable, it is a validity threat that grows with exposure. A suspicious participant is not giving you a noisy version of the real answer; they are answering a different question, namely what they think you are really up to.

Second, and specific to commercial research: in academia the contamination is a commons problem, so each researcher pays a fraction of the cost they create. In a company-owned participant panel it is fully internalised. You are the only organisation fielding into your customer base. Every masked study, every unexplained premise, every survey that turned out to be a sales call raises the suspicion baseline of the exact people you will need to interview next quarter. You are not borrowing from a shared pool. You are spending your own.

That is a quantifiable argument for a research operations budget, and it is one leadership understands: the panel is an asset with a depreciation schedule you control.

Three questions that replace the deception debate

When somebody proposes a cover story, these three questions resolve it faster than a principles discussion.

Can the design get the same answer without the false statement? Usually the honest version of the question, asked about specific past behaviour rather than about intentions, gets you there. Deception is frequently a substitute for a well-written question. If you cannot articulate which nondeceptive alternatives you tried, you have not satisfied APA 8.07(a) and you probably have not tried.

Would this participant, on learning the truth, feel they had been used? This is the practical version of the respect test, and it separates rung 2 from rung 4 reliably. Almost nobody objects to learning that the researcher had a hypothesis they did not share. People object strongly to learning that the friendly person in the session was an actor, or that the feedback they received was fabricated.

Who else fields into this pool, and how often? If the answer is only us, and monthly, the contamination cost lands entirely on you and compounds.

The debriefing specification

A debrief is not a thank-you message. A complete debriefing contains five elements:

• What was withheld or misstated, and what is actually true. Specific, not euphemistic. If the second product concept was fictitious, say so.
• Why the design required it. People accept masking readily when the reason is explained, and the explanation is usually short: knowing what we expected would have changed how you answered.
• What the study was really measuring. The honest version of the purpose you gave in general terms at recruitment.
• The right to withdraw the data, with a mechanism. APA 8.07(c) requires that participants be permitted to withdraw their data. A right with no working process is not a right. Give a link or an address and a deadline, and honour it.
• A named contact, and where the topic warrants it, a support resource. For anything touching finances, health, bereavement or personal safety, see our guidance on trauma-informed research.

On timing, 8.07(c) sets the outer bound at the conclusion of data collection, with the strong preference being at the conclusion of the individual's participation. Delay is permitted only where later participants would otherwise learn the design from earlier ones, and even then the delay should be defined in advance with a date, not left open.

Two further distinctions are worth keeping. Dehoaxing is correcting the false belief. Desensitising is addressing how the participant feels about what they did or revealed while holding it. A debrief that dehoaxes but does not desensitise can leave someone worse off than no debrief at all, which is exactly the harm 8.08(c) tells you to minimise.

Where AI moderation helps, and where it does not

An AI-moderated interview changes three things about this problem.

The script is the record. With a human moderator, what was actually said in the room is reconstructed afterwards from memory and notes, and moderators improvise. With an AI interviewer working from a defined guide and a full transcript of every session, what each participant was told is auditable. If a study is later challenged, the question of whether the researcher went beyond masking into deception is answerable from the record rather than argued from recollection.

The debrief is deliverable and verifiable. Because the platform holds the transcript and the participant record, a debriefing message can be delivered to every participant at the end of the session and logged, rather than depending on a researcher remembering to send it after the twentieth interview. The right to withdraw data becomes operable rather than theoretical.

Some reasons for deception disappear. A large share of commercial cover stories exist to manage social desirability: the participant is trying not to disappoint the person in the room. An AI interviewer has no visible stake in the answer, and participants routinely report finding it easier to be blunt with. That does not eliminate the bias, but it removes one of the more common excuses for reaching for a false premise. Koji supports six structured question types (open_ended, scale, single_choice, multiple_choice, ranking, and yes_no), and the combination that does most of the work here is a yes_no or single_choice question about specific past behaviour, followed by open_ended probing that the AI generates itself. Where you would otherwise have used a fabricated comparison to force a preference, a ranking question gets the same discrimination honestly, and a scale question captures intensity without implying a right answer. Asking what did you actually do last time rather than what would you do produces less socially managed answers than any cover story.

What AI moderation does not change is the obligation. Article 4(a) requires the researcher to promptly identify themselves and unambiguously state the purpose. That applies to an AI-conducted interview exactly as it does to a human one, and participants should also be told plainly that they are speaking with an AI interviewer and that the session is recorded, as Article 6 requires. Concealing the nature of the interviewer is itself a rung 4 deception, and an unnecessary one: disclosure of AI moderation does not depress participation in the way teams fear.

Frequently asked questions

Is it deception to not tell participants what my hypothesis is?

No. Withholding the specific hypothesis is masking, and it is standard practice because telling people what you expect to find changes what they do. The ICC/ESOMAR Code requires cooperation to be based on adequate and not misleading information about the general purpose and nature of the project. The word general is doing real work: you owe an honest account of what the study is about and who is running it, not a preregistration. Deception begins when you assert something that is not true.

Can I conceal which company commissioned the research?

Often yes, and frequently you should, because naming the sponsor produces sponsor-pleasing answers. What you cannot do is deny it, invent a different sponsor, or present yourself as an independent party when asked directly. ESOMAR requires researchers to promptly identify themselves and unambiguously state the purpose of the research, and requires respondents to be able to check the identity and bona fides of the researcher without difficulty. Blinding the brand is compatible with that. Lying about it is not, and refusing to answer a direct question is a signal you should treat as a design failure.

What is sugging and why does it matter legally?

Sugging is selling under the guise of research: a contact framed as a survey or interview whose actual purpose is to generate a lead or make a sale. The ICC/ESOMAR Code requires market research to be clearly distinguished and separated from non-research activities including any commercial activity directed at individual respondents. In the UK it is more than an ethics breach. Schedule 20 of the Digital Markets, Competition and Consumers Act 2024 lists falsely claiming or creating the false impression that the trader is not acting for business purposes, or falsely representing oneself as a consumer, as unfair in all circumstances, requiring no proof that anyone was misled.

What must a debriefing actually contain?

Five elements: what was withheld or misstated and what the truth is; why the design required it; what the study was really measuring; the participant's right to withdraw their data together with a working mechanism to exercise it; and a named contact plus, where the topic warrants, a support resource. APA Standard 8.07(c) requires the explanation as early as is feasible, preferably at the conclusion of participation and no later than the conclusion of data collection, and requires that participants be permitted to withdraw their data.

Does deception really damage future research?

The evidence says yes and it is measurable. Analyses of the psychology literature found the proportion of deception studies in the leading social psychology journal tripling from 16 percent to 47 percent by 1971, with participant-reported suspicion tracking it: roughly one in ten participants reported suspicion until the mid-1960s, rising to between 40 and 60 percent for 1967 to 1973. In one 1976 study, 20 percent of participants were excluded on self-reported suspicion. In a company-owned panel you are the only organisation fielding into that pool, so you absorb the entire contamination cost yourself rather than sharing it across a discipline.

Do I need deception to avoid social desirability bias?

Almost never. Social desirability is better handled by design than by lying: ask about specific past behaviour rather than intentions, remove your own stake from the question wording, use indirect framings, guarantee and then honour anonymity, and let a neutral AI interviewer run the session so there is no human whose approval the participant is managing. Deception should be what you reach for after nondeceptive alternatives have failed, which is precisely what the APA justification test requires you to be able to demonstrate.

Related Resources

• Research Ethics and Informed Consent covers the consent framework that masking and deception sit inside.
• Research Consent Form Templates provides GDPR-compliant wording for stating a general purpose honestly.
• Structured Questions in AI Interviews explains the six question types and how behaviour-anchored questions reduce social desirability.
• Trauma-Informed User Research covers desensitising and support resources for sensitive topics.
• How to Record Customer Interviews sets out the recording disclosure obligations referenced above.
• Research Peer Review is where a proposed cover story should be challenged before the study fields.
• How to Build a Research Participant Panel explains panel health, which is the asset deception depreciates.

---

Want research your participants trust enough to answer honestly? Start free with 10 credits and run your first AI-moderated study today.