Should Research Participants Be Able to Opt Out of AI Processing?

Can Participants Agree to Your Research but Refuse to Have AI Analyze Their Responses?

You are conducting a qualitative study involving university instructors. Participants have agreed to share their experiences through semi-structured interviews, and the research team plans to use generative AI to help organize selected transcript excerpts.

During recruitment, one prospective participant asks whether their interview can be analyzed without AI.

The participant is comfortable speaking with the researcher and allowing the university to use the interview for the approved study. However, they would prefer that their responses not be submitted to an external AI provider.

You could analyze that participant's interview manually, but doing so would create a different workflow from the one planned for the rest of the study.

Should participants have the right to refuse AI processing while remaining in the study? Or can the researcher reasonably require AI processing when it is part of the approved methodology?

An AI-Specific Opt-Out May Be Appropriate, but It Is Not Universally Required

Understanding Participant Choice and AI Processing in Research

Voluntary Research Participation Is Not the Same as an AI-Specific Opt-Out

Voluntary participation generally means that prospective participants can decide whether to join a study without improper coercion or undue influence.

Depending on the applicable framework and study arrangements, participants may also have rights concerning withdrawal and the processing of their personal information.

However, the ability to decline research participation does not necessarily establish a right to select every analytical method used by the research team.

For example, a participant may agree to a study that uses researcher-led thematic analysis without being offered a choice among different qualitative analysis software packages.

Generative AI may raise additional considerations when its use introduces an external provider, different processing conditions, or automated interactions that participants consider material.

Researchers should therefore distinguish the right to decide whether to participate from any additional choices available concerning AI-assisted procedures.

What Does an AI Processing Opt-Out Mean?

An AI processing opt-out is an arrangement through which a participant can decline a specified AI-assisted activity while potentially remaining involved in other parts of the research.

For example, a study might allow participants to complete interviews while requesting that their transcripts be analyzed only by authorized human researchers.

However, the meaning of opting out must be defined precisely.

Does the choice prevent AI-assisted transcription, external processing, preliminary coding, or every form of AI involvement with the participant's information?

A statement such as "You may opt out of AI" may be misleading if it does not identify which processing activities the choice covers.

Researchers should explain the actual scope of the option and any limitations that apply.

Distinguish Optional AI Processing From Essential AI Procedures

The research design is central to determining whether an alternative can be offered.

For example, a qualitative study examining instructors' experiences with institutional policies may be able to analyze some interviews without AI.

By contrast, an experimental study investigating students' interactions with an AI tutor may require participants in the relevant condition to interact with that system.

Removing AI from the latter procedure may fundamentally change what is being studied.

The researcher should not promise an AI-free alternative when doing so would be incompatible with the approved research design.

When Might an AI-Specific Opt-Out Be Appropriate?

Several circumstances may support considering an AI-specific choice, although they do not automatically establish a legal requirement.

The feasibility of an alternative should be assessed before participants are offered the choice.

Researchers should also consider whether the proposed alternative introduces new methodological or confidentiality concerns.

Does Informed Consent Automatically Require an AI Opt-Out?

No universal research ethics principle establishes that every AI-assisted procedure must have a separate opt-out.

Informed consent generally concerns whether participants receive appropriate information and voluntarily agree to the proposed research procedures.

For example, participants may consent to an approved study in which AI-assisted interaction is essential to the intervention.

However, additional choices may be required or appropriate when a particular processing activity is legally optional, separately consented to, or not necessary for participation.

Researchers should not assume that a signed research consent form eliminates every applicable data-protection right.

Equally, they should not assume that every use of AI creates a separate right to select a non-AI method.

Why the Legal Basis for Processing Matters

Research participation consent and the legal basis for processing personal information are related but not necessarily identical.

For example, under the GDPR, personal data processing may rely on different lawful bases depending on the circumstances.

Where processing relies on consent, the conditions for valid consent and withdrawal become relevant.

Where another lawful basis applies, different rights and requirements may govern the processing.

Article 21 of the GDPR also establishes a right to object to certain processing activities, subject to applicable conditions and exceptions.

Researchers should therefore avoid treating an AI opt-out as a universal legal entitlement or assuming that it is unavailable merely because research participation has already been agreed to.

The applicable legal framework and actual processing purpose must be examined.

What Does the Philippine Data Privacy Act Mean for Participant Choice?

In the Philippines, Republic Act No. 10173 establishes rights of data subjects and requirements concerning lawful personal information processing.

Section 16 includes rights relevant to information, access, objection, and other matters, subject to the Act's provisions and applicable exceptions.

However, these rights should not be interpreted as establishing an unrestricted ability to demand that every research procedure be performed without AI.

For example, the availability of an objection or withdrawal mechanism depends on the applicable processing arrangement and legal conditions.

Researchers should distinguish the right to decline participation from rights concerning particular personal information processing activities.

Institutional data protection officers may help determine whether an AI-specific choice is required or appropriate in a particular study.

Can Participants Opt Out of External AI Processing but Not Other Analysis?

Potentially. A research team may be able to separate external AI processing from the underlying research analysis.

For example, the study may require interview transcripts to be analyzed, but it may not require the analysis to occur through an external AI provider.

Participants could potentially agree to the research while declining that particular processing arrangement.

However, this requires an operationally feasible alternative and appropriate institutional authorization.

Researchers should avoid presenting an external provider as indispensable when the actual analytical task can be performed through an authorized alternative.

At the same time, the research team is not necessarily obligated to redesign every procedure in response to individual preferences beyond applicable legal and ethical requirements.

What Makes an Opt-Out Meaningful?

An opt-out is meaningful only when participants understand the choice and the research team can implement it as described.

For example, a participant may select an option stating that their interview will not be processed by an external AI provider.

If the researcher subsequently uploads the transcript to that provider, the option has not been honored.

Similarly, an opt-out offered after the information has already been processed may not reverse the earlier transfer.

Researchers should therefore establish when the choice can be exercised, what processing it prevents, and whether any limitations apply.

They should also ensure that the selected option is communicated to everyone responsible for handling the participant's information.

How Should Researchers Track AI Processing Preferences?

When an AI-specific choice is offered, the research team needs a reliable way to implement it.

For example, a participant management record may indicate whether each participant's information is authorized for the specified AI-assisted activity.

The research team should avoid placing unnecessary identifying information into AI prompts merely to manage these preferences.

A coded participant identifier may help distinguish records, provided the associated linkage information remains appropriately protected.

Researchers should also consider how preferences will be communicated to authorized analysts and whether processing restrictions remain effective throughout the study.

Where the preference is part of the approved consent arrangements, the implementation procedure should be documented.

Can Offering an Opt-Out Affect Research Validity?

Potentially. Different analytical procedures may introduce differences that researchers need to consider.

For example, if some interviews are analyzed with AI-assisted coding and others through researcher-led coding, differences in analytical procedures may influence how categories are generated or applied.

Researchers should establish a consistent analytical framework and appropriate verification procedures rather than assume that the two approaches automatically produce equivalent results.

In an experimental study, allowing participants to select AI or non-AI conditions may also introduce self-selection effects that complicate causal interpretation.

These methodological considerations do not automatically justify refusing participant choice.

Instead, researchers should determine whether the choice can be accommodated while preserving a defensible research design.

Should Participants Be Told That an Opt-Out May Affect Their Participation?

Yes, where the choice has genuine consequences for participation or the use of their information.

For example, if AI interaction is essential to the study, declining that procedure may mean the individual cannot participate in that particular research condition.

Researchers should explain this accurately without suggesting that the participant is obligated to accept AI processing.

Where an alternative is available, the materials should explain how the participant can select it.

Any consequences should be consistent with the approved protocol and applicable ethical requirements.

Researchers should not introduce punitive or misleading conditions to discourage participants from exercising legitimate choices.

Can Participants Opt Out After AI Has Already Processed Their Information?

The answer depends on the applicable rights, processing arrangement, and stage of the research.

For example, a participant may request that their information no longer be submitted to an AI provider.

The research team may be able to prevent future processing, but it may not be able to reverse processing that has already occurred.

Deletion or withdrawal requests may also be subject to applicable legal, institutional, and provider-specific conditions.

Researchers should distinguish stopping future AI processing from deleting previously processed information.

They should not promise that an opt-out retroactively removes all records from an external provider unless the actual arrangement supports that outcome.

This is closely related to AI provider data retention and deletion conditions.

How Should Researchers Explain an Optional AI Processing Choice?

Where the approved study genuinely offers an alternative, participant information should explain the choice in accessible language.

This example assumes that the alternative is available, methodologically acceptable, and approved for the study.

Researchers should not use the wording if AI processing is essential to the research design or if the promised alternative cannot actually be provided.

Can AI Help Design an Opt-Out Procedure?

Generative AI may assist researchers in identifying practical questions about implementing participant preferences.

However, the model should not determine whether an opt-out is legally required or institutionally authorized without verified information.

The researcher should verify the proposed procedure and obtain the appropriate institutional determination before offering it to participants.

What If an AI-Free Alternative Is Not Feasible?

Sometimes the research team cannot reasonably provide a separate non-AI procedure without changing the study's essential design.

For example, a study investigating students' experiences with an AI tutor cannot examine the same intervention if the AI tutor is removed.

In such circumstances, researchers should explain the AI involvement clearly and allow prospective participants to decide whether to participate under the applicable consent conditions.

However, the absence of an alternative does not automatically eliminate legal rights concerning personal information processing.

If the proposed processing cannot satisfy applicable rights or authorization requirements, the research design may need to be modified or the activity may not be permissible.

Researchers should not use methodological convenience as a substitute for examining those obligations.

Does an Opt-Out Automatically Make AI Processing Ethical?

No. Providing a choice does not independently establish that the AI service is authorized, secure, or suitable for the research purpose.

For example, offering participants an opt-out does not justify uploading other participants' identifiable transcripts to an unauthorized provider.

The processing arrangement must still satisfy the relevant ethics, data-protection, and institutional requirements.

Researchers should therefore distinguish the ethical value of participant choice from the separate obligation to establish appropriate processing safeguards.

Allowing an Instructor to Participate Without External AI Analysis

The participant remains in the study because the alternative is both feasible and authorized.

The example illustrates how an AI-specific opt-out can support participant choice without assuming that every research design must provide the same option.

Common Misconceptions About AI Processing Opt-Outs

Offer an AI Processing Choice Only When It Is Genuine and Appropriately Authorized

Before promising an AI-specific opt-out, determine whether the research can accommodate a non-AI procedure and whether the choice is required or appropriate under the applicable framework.

Then establish how the preference will be recorded, implemented, and explained to participants.

Do not promise participant control that the research team cannot actually provide.

Respect for participant choice requires accurate explanations and procedures that work in practice, not simply an additional checkbox on a consent form.

Before Offering an AI Processing Opt-Out

Frequently Asked Questions About Research Participant AI Opt-Outs

Participant Choice Matters, but an Opt-Out Must Be More Than a Promise

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.