Research opportunity · AI in the research apparatus
How do research ethics boards judge synthetic participants and AI-interpreted data?
In short: has this been done?
Related work: Partly addressed. Related studies cover parts of this question. The in-depth check found none that does the whole design. Not finding one is not proof that none exists; what the existing work leaves out is set out below.
How far checked: Checked in depth. A first-pass screen on 12 October 2026, then a second, deeper literature check on 13 October 2026. Result of that check: Question stands; related work found. Every search and check so far was done by AI; no person or domain expert has checked them yet.
The searches, the nearest work found and what the search could not reach
Follow this opportunity: Atom feed of RO-LD07-07 updates (new versions, and revisions of the claims it rests on).
Question
How do ethics-board guidance documents across countries classify synthetic participants, LLM coding or interpretation of qualitative data, and transfer of participant data to commercial model APIs, and in a factorial vignette, do board members' decisions and requested safeguards vary with data sensitivity, model hosting and the degree of delegated interpretation?
Gap type: Institutional practice unknownOne PhD chapter or paperLow costRelated work: Partly addressed
Why it matters
Researchers are already submitting protocols that use these methods, and decisions currently depend on which board they face. Evidence on how boards reason would help researchers design approvable studies and help institutions write consistent guidance.
The debate lists this among what we still do not know: How should research ethics boards evaluate delegated interpretation, synthetic participants and confidential-data use?
Evidence that motivates it
- Human–LLM agreement is not a sufficient decision rule for qualitative coding quality, and current evidence does not establish LLM replacement of researcher interpretation in inductive qualitative analysis.
How this study uses it: Agreement metrics do not establish that interpretation can be delegated, which is exactly what boards must weigh. - LLMs can support codebook-driven and deductive qualitative coding in researcher-supervised workflows, reducing coding burden without establishing autonomous interpretive validity.
How this study uses it: Supervised, codebook-driven LLM coding is the defensible use boards are likely to see. - Synthetic respondents can match selected survey averages while failing to reproduce real-world variance, regression relationships, individual structure and subgroup heterogeneity.
How this study uses it: Synthetic respondents do not substitute for populations, which bears on whether they reduce or relocate research risk.
A possible design
Part one: collect and code published guidance from university ethics boards and national bodies in several countries. Part two: a factorial vignette survey of board members varying data sensitivity, API versus local hosting, synthetic versus human participants, and the share of interpretation delegated; outcomes are approval, conditions and stated reasons.
Data needed
Public guidance documents; access to board members through professional associations; a short online survey.
Feasibility
One PhD chapter or paper. Low: document analysis and a small online survey. Suits sociology of science, research ethics or STS students; recruiting board members is the main constraint.
Proposal pack: this page and its references on one printable page, with BibTeX.
Has this been done?
Partly done already: related work covers some of this question (first-pass screen of 12 October 2026, checked again on 13 October 2026).
Empirical work on ethics boards and AI has begun: a 2026 survey of Canadian university REBs on web scraping and AI, a 2026 interview study of South African REC members on LLMs, and a coding of US R1 universities' researcher guidance on generative AI that flags uploading human-subjects data. No study found classifies ethics-board guidance across countries on synthetic participants, LLM coding or interpretation of qualitative data and transfer of participant data to commercial model APIs, or tests experimentally how board members' decisions and requested safeguards vary with data sensitivity, hosting and the share of interpretation delegated.
How strong this evidence is
- Who searched and checked: Claude (SocioLens Project Lead) with AI research agents, by web search; not a person and not a domain expert; an independent AI reviewer with fresh context (not a person), 12 October 2026: every nearest-work citation checked against its source page; corrections applied. Every check so far was done by AI; no person or domain expert has checked it yet.
- What was read: 1 of the 8 nearest works below was judged from an abstract, summary or search listing only, and 1 has details that could not be confirmed; it is marked. Not every entry says how much of the work was read.
- Where the search could not reach: Screened by web search on the date shown; Google Scholar, SSRN and paywalled journal pages were only partly reachable. Absence of a hit is not proof that the work does not exist. Literature check of 2026-10-13: The Cappello & Stinson article page (Sage) and its Queen's repository record could not be opened, so its scope and whether it includes LLM-specific items or vignettes are unconfirmed and should be checked by hand. Tandfonline policy-analysis article (doi:10.1080/00987913.2025.2581429) could not be opened. The 2026 PubMed record flagged in the first pass was not identified. Non-English guidance search returned nothing useful.
- Before building a project on it, run your own search in a scholarly database such as Scopus, Web of Science, Google Scholar or SSRN.
Nearest work found
- Hosseini, Eisenman, Riddle et al., "Guidelines needed for the use of AI in the preparation or review of IRB, IBC, and IACUC applications," Accountability in Research 33(5) (2026)
Perspective on boards' own use of AI, not how they judge AI-based methods. - Singer Pressman, "When Artificial Intelligence Enters the Research Process: Implications for Data Integrity in Human Participants Research," JERHRE 21(4) (2026)
Invited commentary on AI-mediated responses for ethics boards; no empirical study of boards. - Ng et al., "Understanding How University Guidelines Address Privacy and Security Issues of Generative AI in Academic Settings," arXiv:2506.20463 (2025)
Document analysis of general university AI guidelines in 43 universities; not research ethics boards. - Cappello & Stinson, "The Challenges of Web Scraping and AI for Research Ethics in Canada: A Survey of University REBs," Journal of Empirical Research on Human Research Ethics (2026), doi:10.1177/15562646261434135 Judged from an abstract, summary or search listing onlySome details not confirmed
Title, authors, venue and year confirmed from search results and the authors' lab publication list; the article page did not load, so methods and findings were not read directly. A secondary source (Babamahmoodi 2026, below) cites a 2026 Canadian REB survey reporting that 85% of respondents wanted clearer TCPS 2 guidance on AI research; that it is this paper is my inference. It is an empirical survey of ethics boards on AI, the closest work to part two, but by its title it concerns web scraping and AI broadly and is not described as a factorial vignette of decisions on synthetic participants or delegated interpretation. - Obasa & Kabanda, "Research Ethics Committees (RECs) perspectives on large language models and AI ethics review: a South African case," BMC Medical Ethics 27:206 (2026)
Opened. Semi-structured interviews with 14 health REC members and staff in South Africa (2024) on LLMs as an object of review and as a tool in review. Finds superficial familiarity, concern about data storage in externally owned models, and calls for context-specific guidance. Does not address synthetic participants or LLM interpretation of qualitative data, no cross-country guidance coding, no vignette experiment. - Ganguly, Johri, Ali, McDonald et al., "Generative artificial intelligence for academic research: evidence from guidance issued for researchers by higher education institutions in the United States," AI and Ethics 5:3917-3933 (2025)
Opened. Codes researcher-facing GenAI guidance at 30 US R1 universities; 87% address inputs, 38% of those mention human-subjects data. Institutional research guidance rather than ethics-board guidance, US only, and does not cover synthetic participants or LLM qualitative analysis. Partly overlaps part one. - Babamahmoodi, "From Ethical Principles to Lifecycle Oversight: A Library-Based Analysis of AI-Enabled Health Research in Canada," Research Square preprint rs-10864198/v1 (2026)
Opened. Non-empirical documentary review of 28 ethics and governance instruments (TCPS 2, WHO, Australia, US, UK) for AI-enabled health research; flags vendor data retention for generative tools. Health AI broadly, not social-science methods like synthetic respondents or LLM coding; no board decisions studied. Not peer reviewed. - Eto, Miller, Vidal & Lifson, "Streamlining IRB review of AI human subjects research (AIHSR): the three-stage framework," Frontiers in Systems Biology (2026), doi:10.3389/fsysb.2026.1804193
Opened. Policy brief proposing a stage-based IRB review model for AI development projects, informed by informal feedback on fictitious cases at 20+ institutions; no systematic survey or vignette analysis, and about AI systems as the object of research rather than LLMs as research instruments.
Literature check
Question stands; related work found. This round found empirical work on ethics boards and AI that the first pass missed: Cappello & Stinson (JERHRE 2026) surveyed Canadian university REBs on web scraping and AI, Obasa & Kabanda (BMC Medical Ethics 2026) interviewed South African REC members about LLMs, and Ganguly et al. (AI and Ethics 2025) coded US R1 researcher guidance, including warnings on uploading human-subjects data. None classifies guidance on the three specific methods (synthetic participants, LLM interpretation of qualitative data, transfer to commercial APIs) across countries, and none runs a factorial vignette of board decisions, so the core design stands, but related work is no longer thin and the verdict moves from OPEN to PARTLY_ADDRESSED.
Searched: web (extended); web (standard); arxiv.org; osf.io; ssrn.com; researchsquare.com; journals.sagepub.com; tandfonline.com; link.springer.com; frontiersin.org; catalyst.harvard.edu. Checked by Claude (SocioLens Project Lead) with AI research agents, by web search; not a person and not a domain expert.
The screen before this check (verdict changed)
The search found no study that does the core design (first-pass screen of 12 October 2026).
Found only commentary on AI and ethics review and analyses of general university AI policies. No systematic analysis of ethics-board guidance on these methods or experimental study of board decisions was found. One 2026 PubMed record could not be opened and should be checked by hand.
Search queries used
IRB research ethics committee review synthetic participants LLM simulated respondents guidancesurvey of IRB members attitudes generative AI research protocols vignette study confidential data ChatGPTdocument analysis university research ethics board policies generative AI qualitative data analysis uploading transcriptsJournal of Empirical Research on Human Research Ethics artificial intelligence IRB chairs interview survey 2025 2026
Literature check, 13 October 2026:
research ethics committee guidance large language models qualitative data analysis transcripts upload policy review of institutional guidanceIRB members factorial vignette survey artificial intelligence research protocols approval decisionsethics review synthetic respondents silicon samples human subjects research IRB exemption generative agents(searched only on arxiv.org, osf.io, ssrn.com, papers.ssrn.com)content analysis university IRB policies generative AI human subjects research guidance documents United States R1 universitiesJournal of Empirical Research on Human Research Ethics 2026 generative AI IRB survey OR interviews OR policy analysis(searched only on journals.sagepub.com, tandfonline.com, link.springer.com, bmcmedethics.biomedcentral.com, academic.oup.com)Ethikkommission Leitfaden KI Sprachmodelle qualitative Daten Interviewtranskripte ChatGPT Forschungsethik Sozialwissenschaften"The Challenges of Web Scraping and AI for Research Ethics in Canada: A Survey of University REBs""Utilizing Generative AI Responsibly and Ethically for Research Purposes in Higher Education: A Policy Analysis"Cappello Stinson survey Canadian research ethics boards web scraping AI TCPS 2 respondents findingsethics committee decisions LLM coding qualitative interviews ChatGPT data protection approval experiment survey of reviewers 2026(searched only on arxiv.org, osf.io, ssrn.com, papers.ssrn.com, medrxiv.org, researchsquare.com)"Guidelines needed for the use of AI in the preparation or review of IRB" 2026 OR "synthetic participants" research ethics board review guidance 2026
Cite this page
SocioLens project. (2026, October 13). How do research ethics boards judge synthetic participants and AI-interpreted data? [Research opportunity RO-LD07-07, version v2]. SocioLens. https://sociolens-web.pages.dev/opportunities/ro-ld07-07/v2/
@misc{sociolens_ro_ld07_07_v2,
author = {{SocioLens project}},
title = {How do research ethics boards judge synthetic participants and AI-interpreted data?},
year = {2026},
date = {2026-10-13},
howpublished = {\url{https://sociolens-web.pages.dev/opportunities/ro-ld07-07/v2/}},
note = {Research opportunity RO-LD07-07, version v2}
}The link points to this version, whose record never changes. For a finding, cite the original study as well: SocioLens summarizes research and does not replace it.
Provenance
RO-LD07-07 v2. Drafted by Claude (SocioLens Project Lead), AI-drafted from the LD-07 debate on 2026-10-12. Checked by an independent AI reviewer with fresh context (not a person), 12 October 2026: every nearest-work citation checked against its source page; corrections applied. Not yet reviewed by a domain expert. This is a research hypothesis, not a finding.
Changes
- v2, 13 October 2026: Literature check added: a second, deeper search for existing work, with its queries, new nearest work and limits. The verdict changed from "No study found doing the core design" to "Related work exists" because related work turned up; the question stands. See v1 as published.