Glossary
Agent-based research
Also called agent-based market research · multi-agent simulation
A design in which respondents are not merely questioned but set in motion over time, so that one reaction can produce the next.
A survey captures a moment: what people say when you ask. Agent-based research models something else — how a position moves once information arrives, once the people around someone react, once a decision starts to circulate.
The claimed gain is propagation: threshold effects, tipping points, opinion cascades, gradual adoption. It appeals most for decisions whose effect cannot be judged on day one — a price change, a site, a reform.
What it is not
It is not a forecast. The more turns of interaction a model simulates, the further it drifts from anything that was calibrated at the start, and the less checkable the resulting trajectory becomes. A three-year simulation does not carry the evidential weight of a reaction to a stimulus, and the deliverable should say so.
What a buyer can check
What exactly was calibrated: the starting state only, or the interaction rules as well? What happens when a structural assumption is varied — a robust result holds, a fragile one scatters. And whether the vendor keeps what was measured visibly apart from what was assumed.
See also
- AI focus group — Several synthetic respondents brought together with a moderator, reacting to a stimulus and to each other. The depth route, on a small group.
- Synthetic population — The generated set taken as a whole, built to match the known statistical margins of a territory. The term comes from microsimulation, not from research.
- Digital twin — An attempt to replicate a specific, real individual rather than represent a segment. The strongest claim in the field — and therefore the most demanding to evidence.
Further reading
- Why ESOMAR Asks Twenty QuestionsFlashInsight blog
- Writing — columns, articles, researchFlashInsight
- Our methodFlashInsight