For Political Organizations & Policy Teams

Pressure-test policy messaging before you go public.

Rehearse town halls, constituent meetings, and policy rollouts with demographically grounded participants. Spot where a message is vulnerable before you take it to real constituents.

Request Early Access

The blind spots in political research

Polling tells you positions. It doesn’t show how voters reason or what happens when a message meets competing views. And real constituent events are high-stakes and one-shot.

The Problem
Polls capture stated positions but miss reasoning, persuadability, and how voters respond when their views are challenged by neighbors with different backgrounds.
AnthroSim Solution
Simulate structured discussion among demographically accurate voter profiles. Observe how positions shift, what arguments land, and where opposition concentrates.
The Problem
Testing policy messaging in real constituent settings is high-risk. A poorly framed position in a town hall becomes a news story.
AnthroSim Solution
Run simulated town halls with synthetic constituent panels before any real-world deployment. Identify messaging failure modes in a zero-risk environment.
The Problem
Opposition research on voter response to policy alternatives requires costly dial testing or focus groups, and can't keep up with fast-moving legislative cycles.
AnthroSim Solution
Generate fresh voter panels in minutes. Test competing policy framings same-day and get structured vote outcomes and qualitative reasoning by demographic segment.

Example scenario: Ballot measure simulation

Step 1. Generate profiles: 60 participants · Guidance: “likely voters in a swing suburban district, mixed income, family-oriented, homeowners and renters”

Step 2. Add project materials: Ballot language, fiscal note, supporter message, and opposition critique

Step 3. Run simulation: Topic: “Measure J: 15% inclusionary zoning requirement for new developments over 10 units” · Duration: 90 min · Export: JSON

Output includes vote breakdown (Support / Oppose / Undecided) with representative statements organized by demographic group. Read the results as directional signals: where a message is vulnerable, which frames attract opposition, and what arguments move persuadable voters. Validate subgroup breakdowns with real voters before treating them as predictions.

Research validation

Controlled experiments on AI-assisted persuasion have found:

Peer-reviewed, 2025 · 2025
AI-personalized persuasion outperforms human debaters
Salvi et al.
Personalized AI outperformed human debaters in persuasiveness across 900 participants in controlled experiments. Direct evidence that AI-simulated persuasion dynamics are empirically meaningful, and that responsible-use governance is essential.

A note on responsible use

AnthroSim is a research and stress-testing tool. Results represent synthetic population behavior. Use them to prepare for real constituent engagement; they do not replace it. Simulation outputs are directional. Subgroup-level breakdowns generate hypotheses; they are not statistically definitive.

Research shows AI-powered persuasion can be potent. That makes it useful for pre-deployment message testing, and it makes careful handling a serious responsibility. Improbability Engineers, LLC does not support use of this platform for deceptive political operations, automated influence campaigns, or the targeting of real individuals.

Request Early Access

We're onboarding a first cohort: academic and applied researchers, market research firms, enterprise AI and product teams, law firms, and policy organizations.

Early access participants receive preferred rates and priority onboarding.

No commitment. Pricing finalized at launch.