Should Participants Be Told When They Are Interacting With AI Rather Than a Human Researcher?

Do Participants Have a Right to Know Their Research Interviewer Is AI?

Suppose a researcher develops a chatbot that conducts convincing semi-structured interviews. It introduces itself, asks relevant questions, and responds conversationally. Participants might reasonably assume that a human researcher is typing the messages.

Should the researcher explicitly identify the interviewer as AI? Could disclosure influence participants' answers, particularly when the study investigates sensitive experiences? And if knowing the interviewer's identity changes participant behavior, is withholding that information ever methodologically justified?

These questions concern more than technological transparency. They involve informed consent, participant autonomy, research validity, and the ethical boundaries of deception in human participant research.

Participants Should Generally Be Informed That Their Interviewer Is AI

Why AI Interviewer Disclosure Matters in Human Participant Research

Why Is the Interviewer's Identity Relevant to Informed Consent?

Informed consent requires more than obtaining a participant's signature or agreement to a statement. It involves providing information necessary for a meaningful, voluntary decision about research participation.

Under the U.S. Common Rule, informed consent generally requires an explanation of the research purpose, procedures, reasonably foreseeable risks, confidentiality protections, and participants' rights. The regulations also require information to be presented in language understandable to prospective participants.

The U.S. Office for Human Research Protections emphasizes that consent is an ongoing communication process rather than merely a completed form. Its informed consent guidance explains the importance of meaningful disclosure and opportunities to ask questions.

When an AI system conducts an interview, its identity may reasonably influence whether someone wishes to participate. A person who agrees to speak with a human researcher has not necessarily agreed to disclose the same information to an automated system.

AI involvement may also introduce additional considerations, including external processing of responses, automated conversational decisions, and limitations in the system's ability to respond appropriately to unexpected disclosures.

For these reasons, researchers should treat AI interviewer identity as information relevant to the consent decision rather than assuming it is an inconsequential technical detail.

Does Research Ethics Explicitly Require AI Disclosure?

The answer depends on the applicable regulatory and institutional framework.

Traditional human participant research regulations were not necessarily written with generative AI interviewers in mind. Consequently, some frameworks establish broad informed consent requirements without separately specifying that every use of an AI chatbot must be disclosed.

However, the absence of an explicit AI-specific provision should not be interpreted as permission to conceal a material feature of the research procedure.

The Council for International Organizations of Medical Sciences (CIOMS), in collaboration with the World Health Organization, emphasizes informed and voluntary participation in its International Ethical Guidelines for Health-related Research Involving Humans. Guideline 9 addresses informed consent, while Guideline 22 considers privacy and other ethical issues arising from digital tools and online research.

These guidelines are especially relevant to health-related research and provide useful ethical principles for evaluating digital data collection. They should not be presented as legislation governing every educational or social science study.

Researchers should consult the requirements of their institution, ethics committee, and jurisdiction. Where applicable, additional rules concerning AI systems, privacy, or automated interactions may also apply.

What Exactly Should Researchers Disclose?

Meaningful disclosure should explain the aspects of AI involvement that could reasonably affect participation, rather than relying on a vague statement that the study uses artificial intelligence.

Not every study requires an equally extensive explanation of the underlying technology. A chatbot that simply presents fixed survey questions differs from an autonomous system that generates sensitive follow-up questions.

The disclosure should be proportionate to the system's actual role and risks, while remaining understandable to the intended participants.

Is Saying "This Interview Uses AI" Enough?

Not necessarily. Such a statement identifies the technology but leaves important questions unanswered.

For example, participants may not know whether AI is merely transcribing their answers, suggesting questions to a human interviewer, or independently conducting the conversation.

They may also assume that the chatbot's messages are monitored by a researcher in real time, even when the system operates without live supervision.

A more informative disclosure identifies what the AI does and what it does not do.

The second example communicates the system's interviewing role and the limits of human monitoring. It should only be used when those statements accurately describe the actual study.

When Should Participants Be Told About AI Involvement?

Ordinarily, disclosure should occur before participants decide whether to take part in the AI-mediated procedure.

Researchers should not wait until the end of an interview to reveal AI involvement when that information would reasonably have influenced the participant's original consent decision.

Disclosure may appear in the participant information sheet, consent process, and introductory message displayed by the chatbot. Repeating the essential information at the beginning of the interaction can help prevent confusion, particularly when recruitment and interviewing occur at different times.

For electronic consent, the joint guidance issued by the U.S. Department of Health and Human Services and the Food and Drug Administration emphasizes understandable information, opportunities to ask questions, and appropriate electronic procedures.

The guidance does not establish a universal AI-specific disclosure script, but it supports designing digital consent procedures around participant comprehension rather than merely presenting information on a screen.

Can Disclosure Change How Participants Respond?

Yes. Knowing that an interviewer is AI may influence participants' expectations, willingness to disclose sensitive information, and manner of responding.

Some participants may feel less personally judged by a chatbot. Others may distrust the technology, worry about data retention, or doubt whether it can understand their experiences.

Research by Pickard, Roster, and Chen (2016) examined preferences for disclosing sensitive information to humans and avatars, while Pickard and Roster (2020) investigated disclosure across different automated and human-facing interviewing conditions. These studies indicate that interviewer presentation and perceived social presence can influence disclosure behavior.

However, such findings do not establish that telling participants about AI always increases or decreases disclosure. The effect may depend on the topic, interface, population, and wording of the disclosure.

The possibility of an interviewer effect is a methodological consideration, but it does not automatically justify withholding information that participants need to make an informed choice.

The relationship between disclosure and differences in participant responses to AI and human interviewers should be considered during study design and interpretation.

Can Researchers Conceal AI Interviewer Identity for Experimental Purposes?

In some research, investigators may wish to examine whether participants behave differently when they believe they are communicating with a human rather than an AI system.

Such a design may involve deception or incomplete disclosure. However, methodological interest alone does not establish that concealment is ethically permissible.

Under the U.S. Common Rule, an institutional review board may approve certain alterations or waivers of informed consent when the applicable regulatory criteria are satisfied. These include conditions relating to risk, participant rights and welfare, practicability, and the provision of additional information afterward when appropriate.

CIOMS Guideline 10 similarly addresses modifications and waivers of informed consent under specified ethical review conditions.

These provisions do not constitute automatic authorization to conceal AI involvement. Researchers must obtain the required approval before implementing a deceptive or incomplete disclosure procedure.

Is Debriefing Enough to Justify Withholding AI Disclosure?

Not by itself. Debriefing can help participants understand aspects of a study that were initially withheld, but it does not automatically make an otherwise inappropriate deception acceptable.

Where an ethics committee permits altered disclosure, the approved procedure may require participants to receive additional information afterward. Depending on the circumstances, the research team may also need to explain how participants can raise concerns or exercise any applicable choices regarding their data.

The appropriateness of debriefing depends on the study's design, risks, and regulatory framework.

Researchers should not assume that revealing AI involvement after data collection fully restores the informed choice participants would have had before the interview.

Should Researchers Explain How Participant Data Are Processed by AI?

Yes, to the extent necessary for meaningful consent and compliance with applicable data protection requirements.

Participants may reasonably wish to know whether their responses remain within an institutionally controlled system or are transmitted to an external AI provider.

Relevant information may include the categories of data collected, processing purposes, access arrangements, retention practices, and whether responses may be used for purposes beyond the approved research.

Researchers should verify these arrangements through the actual service configuration, institutional agreements, and applicable provider documentation. They should not assume that every chatbot service retains data in the same way or that disabling one feature eliminates every form of processing.

CIOMS Guideline 22 emphasizes privacy-protective measures when digital tools are used in health-related research, including risks associated with linking or inferring personal information from datasets.

For studies involving identifiable information, privacy obligations may extend beyond the consent statement. Appropriate legal bases, security controls, and institutional authorization may also be necessary.

Does Disclosure Mean Participants Must Understand the AI Model's Technical Architecture?

No. Meaningful informed consent does not ordinarily require participants to understand neural network architectures, model parameters, or the mathematical processes used to generate text.

The relevant information concerns what the system does, how its involvement affects participation, and what risks or limitations participants should reasonably consider.

For example, participants may need to know that an AI chatbot generates follow-up questions, but they do not necessarily need an explanation of transformer attention mechanisms.

Excessive technical detail may actually reduce comprehension by obscuring information relevant to the consent decision.

The appropriate level of explanation should reflect the participant population, the nature of the research, and the system's actual capabilities.

Who Is Responsible When Participants Interact With AI?

The research team remains responsible for the conduct of the study and compliance with applicable ethical and institutional requirements.

Identifying the chatbot as AI does not transfer responsibility for participant protection to the technology provider or to the participant.

Researchers must still establish suitable procedures for data handling, participant questions, withdrawal, and foreseeable problems during the interview.

When the study involves potentially sensitive disclosures, participants should understand whether human assistance is available and how it can be accessed.

This becomes particularly important when the chatbot may encounter participant distress or disclosures suggesting imminent harm.

Disclosing AI Interviewing in a Study of Teachers' Experiences

A possible introductory statement for this hypothetical study is:

This wording is illustrative and must be adapted to the actual system, approved consent procedure, and applicable requirements. It should not replace a complete participant information sheet or consent process.

Most importantly, every operational claim must be accurate. If a researcher monitors the conversation live, the disclosure should say so. If the AI provider processes identifiable responses, that arrangement should not be obscured by describing the interview as confidential without qualification.

Misconceptions About Disclosing AI Interviewers

How to Make AI Interviewer Disclosure Meaningful

Researchers should design disclosure around the information participants need to make a reasonable decision about participation.

For a chatbot administering fixed, low-risk questions, a concise explanation may be sufficient alongside the ordinary consent information. For a generative interviewer handling sensitive experiences or external data processing, more detailed disclosure may be necessary.

Researchers should also distinguish between consent to participate in the research and any additional permissions required for particular data processing activities. These requirements may differ according to jurisdiction and the system used.

Finally, disclosure should be supported by actual operational safeguards. Participants should not be told that their data are anonymous, that no external service receives their responses, or that a human researcher is available immediately unless those statements are accurate.

The broader process of validating AI-mediated research procedures should include checking whether participant-facing explanations accurately reflect system behavior.

Before Obtaining Consent for an AI-Mediated Interview

Frequently Asked Questions About AI Interviewer Disclosure

Participants Should Know Who or What Is Interviewing Them

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.