For Researchers & Academics

Synthetic populations for social science, at scale.

Run behavioral studies without recruitment delays, IRB constraints, or demographic gaps. AnthroSim produces the diverse, structured participant data your research requires.

Request Early Access

The problems with human participant research

Recruiting diverse, representative participants for social science studies is slow, expensive, and often produces homogeneous samples — typically WEIRD (Western, Educated, Industrialized, Rich, Democratic) populations. IRB approval timelines constrain iteration. And running counterfactual conditions with the same participants is practically impossible.

The Problem

Recruiting 50 demographically diverse participants for a 90-minute discussion study takes weeks and costs thousands. Geographic and demographic homogeneity is nearly impossible to avoid.

AnthroSim Solution

Generate 50 personas in under a minute with controllable demographic distribution — age, occupation, regional background, political stance, values. Run the study immediately.

The Problem

IRB approval processes constrain research questions and add months to timelines. Running sensitive topic studies (political polarization, extremist rhetoric) with real participants raises ethical and safety concerns.

AnthroSim Solution

Synthetic participants have no welfare constraints. Study sensitive dynamics — radicalization, conflict escalation, disinformation spread — without ethical review bottlenecks.

The Problem

Running the same participant group through multiple conditions introduces memory and learning effects. Counterfactual analysis requires separate recruitment rounds.

AnthroSim Solution

Generate fresh cohorts for each experimental condition with identical demographic profiles. Strong counterfactual control with unlimited replication — results are most reliable as directional signals for hypothesis generation.

Example scenario: Climate opinion study

A team studying how new information changes climate policy opinions needs to run 5 experimental conditions across a demographically stratified sample of 50 participants each.

With AnthroSim:

Step 1 — Generate profiles: 50 participants · Guidance: “US adults with varied political and regional backgrounds”

Step 2 — Run condition 1 (baseline): Topic: “federal carbon tax” · Duration: 60 min

Step 3 — Run condition 2 (EPA data injection): Same topic, same cohort profile, different information environment

Total time: under 3 hours. Human equivalent: 6–12 weeks.

Validated methodology

The simulation approach is grounded in peer-reviewed research:

Nature Machine Intelligence · 2025

A psychometric framework for evaluating and shaping personality traits in large language models

University of Cambridge & Google DeepMind

Instruction-tuned LLMs accurately emulate the full range of human personality traits, with results reproducible across prompt conditions.

Read paper →

Request Early Access

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

Early access participants receive preferred rates and priority onboarding.