8 studies across 5 public drafts · human peer review pending
For researchers

Pressure-test a study before you recruit participants.

Run a draft question and source packet through a moderated synthetic group. Use the transcript to refine the instrument and choose what needs human testing.

AnthroSim runs are exploratory. They do not predict an individual, verdict, vote, purchase, or population estimate.

Measurements behind the product
AnthroSim research program · 2026
Population fit on held-out survey responses
IE-ASIM-2026-01
On the tested ANES political outcomes, calibrated grounded personas cut aggregate error to one third of naive repeated prompting. Several cheaper readout methods performed about as well.
Read paper →
AnthroSim research program · 2026
Discussion mechanics and evidence use
IE-ASIM-2026-02 through IE-ASIM-2026-06
A private-thought mechanism improved one tested model and hurt another. The evidence study drew cases from a 369,215-debate archive; its source audit led to checks against the exact text a participant received.
Read paper →
What a run records

Inspect the synthetic participant and the discussion around them.

Audience group: commuters affected by a proposed rail extension
Fictional generated profile
Maya Torres
Age: 38
Occupation: Payroll administrator
Personality: Careful, moderately outgoing
Political stance: Moderate; wary of long tax commitments
Communication: Brief answers; asks for concrete costs
Values: Reliable transit; neighborhood businesses

Illustrative synthetic profile. It does not describe a real person.

Project material: Regional Light Rail Expansion Brief
Illustrative synthetic transcript
Maya Torres [14:03:21]
The brief says the sales-tax increase lasts 25 years. I need a clear answer on feeder-bus funding before I can support it.
Jordan Ellis [14:03:45]
East Market gets 22 months of lane closures. The plan needs a concrete relief proposal for shops there.
Moderator [14:04:02]
The packet projects boardings if feeder buses are funded, but it does not say that funding is committed. I am marking that as a question for the real town hall.

The quoted project facts come from AnthroSim's built-in example packet. The people and dialogue are fictional.

Audience construction

Choose one or more U.S.-grounded audience groups, with 1-50 generated participants per group. Unsupported filters are labeled before a run.

Orchestrated discussion

A moderator manages participation. Operators can inspect private decision traces and each participant's decaying in-run memory.

Project materials

Upload files, paste text, or import a page. Participants and the moderator can request bounded lookups from enabled sources.

Ballots and audit record

Every participant submits a final position with a rationale and confidence. Export the complete saved run as JSON.

Request Early Access

We're onboarding a first cohort: academic and applied researchers, market research firms, enterprise product teams, law firms, policy organizations, and teams planning sensitive or regulated studies.

Early access participants receive preferred rates and priority onboarding.

No commitment. Pricing finalized at launch.