Run the standard models of behavioural economics, compare the results, and measure the same parameters in real people with a shareable survey or in a synthetic audience drawn from census records.
Generate up to 100 AI respondents for any experiment. Each one is a real anonymised census record narrated as a person, answered by a language model, and fitted like any human respondent. The silicon sampling method of Argyle et al. (2023).
Sample
Personas are drawn from real ONS and US Census microdata, so age, income, occupation, and region keep their true joint distribution.
Answer
A language model answers the identical choice tasks while conditioned on each persona. Answers are constrained to the survey; nothing else gets through.
Compare
Fitted parameters land beside your human cohort and the published literature, with a fidelity panel that says how closely the two crowds match.
A woman aged 16 to 24, associate professional, living in the South East of England.
Real ONS census record · answered by a language model · zero humans surveyed
ONS census microdata · silicon sampling · fitted by maximum likelihood
Input Parameters
Simulation Output
v(gain) +105.90 · v(loss) −106.33 · prospect value +20.29
Input Parameters
Simulation Output
hyperbolic (solid) vs exponential (dashed) · divergence −£20.24 at day 5
Input Parameters
Simulation Output
logistic curve · Z = 1.231 → adoption 77.4%
Create an experiment from a model and share one link. Respondents answer a short series of choice questions, and the platform fits the model's parameters from their answers. Respondents do not need an account.
Create
Pick a model. The platform generates a set of choice questions calibrated to identify its parameters.
Share
Send the survey link to a class, a team, or an audience. The survey takes about two minutes.
Analyse
Parameter distributions update as responses arrive. You can export every response as CSV.
Which would you prefer?
Loss Aversion (λ)
μ = 2.220Value Curvature (α)
μ = 0.832Probability Weighting (γ)
μ = 0.637Revealed-preference estimation · 13-task battery · Prospect Theory
Plus
Thirty simulations a day, every model, and synthetic audiences.
Pro
Unlimited simulations and experiments, 500 synthetic respondents a month, and exports.
Lifetime
One payment for permanent access to everything.