The Pew Research Center‘s recent publication exploring human vs. synthetic survey respondents has renewed interesting public conversations.
The underlying idea is not entirely new. Researchers have long used personas, agent-based models, simulations, and synthetic data. What has changed is the ability of large language models (LLMs) to combine natural-language interaction with demographic, behavioral, contextual, survey, interview, and first-party data. Researchers can now create simulated participants that respond dynamically rather than following only predetermined rules.
Commercialization is proceeding faster than methodological consensus. Vendors already offer synthetic interviews, survey responses, concept and message testing, UX testing, A/B-test screening, conversational rehearsal, and software-quality testing.
For higher education, the immediate opportunities are likely to emerge outside high-stakes research: prospective-student journey testing, communication testing, software and workflow testing, AI-system evaluation, and advancement rehearsal. More ambitious uses in institutional or academic research may follow, but only as validation practices mature.
Research also exposes inconsistency in success and validity.
Current systems can conduct synthetic interviews, generate survey responses, evaluate concepts, simulate customer segments, navigate software, or rehearse conversations. Research reviews are most supportive of their use in upstream activities such as qualitative exploration, research pretesting, pilot work, and hypothesis development rather than unconditional replacement of human samples.
Synthetic systems can generate hundreds, thousands, or potentially millions of responses at comparatively low marginal cost. That scale is operationally attractive, but it can also create an appearance of empirical precision.
The research does not currently justify a simple transition from human participants to artificial ones. The more plausible maturity path is a hybrid model with human and synthetic participants. Human evidence can ground a participant model. Synthetic participants can then run large numbers of exploratory simulations. Those simulations can identify promising options, poorly designed questions, unexpected edge cases, or areas of disagreement. Real participants can validate the consequential findings. Observed results can then be used to recalibrate the synthetic system.
For a deeper look at the evolution of synthetic use cases, levels of persona, a new conceptual architecture for synthetic participants, and recommendations, read in my latest publication, 2026 Trend Analysis: The Evolution of Synthetic Personas.
Interested in the potential of synthetic research participants in higher education? Connect with me to discuss opportunities, challenges, and emerging applications.
Originally posted by Nipuna Ambanpola on LinkedIn. Be sure to follow him on LinkedIn to catch all his great industry insights.
Share Article:
Get exclusive access to higher education analysts, rich research, premium publications, and advisory services.
