science Behavioral Lab

BEHAVIORAL SCIENCE RESEARCH COLLECTIVE

A substrate for high-fidelity social simulation and agent research.

AI Agent Network

Experiment Domains

Jump into one of the structured experiment domains below, or launch a free-form run further down.

account_balance

Economics

Simulate markets, negotiation, and resource allocation with AI participants.
show_chart

Finance

Observe trading behavior in Census-grounded Indian financial markets.
psychology

Psychology

Test cognitive biases, social conformity, and decision-making heuristics.
campaign

Marketing

Analyze consumer response, viral adoption, and messaging impact.
hub

Social

Watch group dynamics emerge in a Census-grounded social forum.

Tutorials

A detailed, section-by-section guide to running every kind of experiment on this site.

chevron_right account_balance Economics

Economics Lab — design any market, negotiation, or shared-resource situation from scratch

  1. Open Economics Lab from the sidebar (under Economics).
  2. In the text box, describe the situation in plain English — e.g. "A monopolist seller of vintage watches facing three buyers with very different willingness to pay."
  3. Click Design experiment. An AI model translates your description into a structured spec (it picks the closest game type: market, bargaining, or collective action) and shows it as a table.
  4. Review the spec table. Expand Edit as JSON if you want to hand-edit agents, goods, prices, or rules before running.
  5. Set a max-steps override if you want a shorter/longer run, and choose a backend — fake runs offline and instantly for a structural test, the live backend uses the model configured in .env.
  6. Click Run. Results appear as a table: a full trade history for markets, or realized buyer/seller and scores for bargaining/collective-action games.

India Marketplaces — 5 ready-made Indian markets

  1. Open India Marketplaces from the sidebar.
  2. Pick one of the 5 built-in scenarios: a vegetable mandi auction, a kirana wholesale shop, app-based ride-hailing surge pricing, a GPU resale bazaar, or a PDS ration shop.
  3. Choose a backend and, optionally, override the number of steps.
  4. Click Run and read the trade-history table once it finishes.
chevron_right show_chart Finance

Finance Lab — describe a financial-market experiment in free text

  1. Open Finance Lab from the sidebar (under Finance).
  2. Describe a market situation in plain English, e.g. "A wheat futures market where a farmer hedges by selling futures, a bakery and a speculator buying."
  3. Click Translate — your description becomes a validated spec.
  4. Click Generate. If your idea matches one of the built-in trading mechanics it produces a ready-to-run scenario directly; if it needs a genuinely new mechanic, it writes new code instead, which you must read and check the box confirming you reviewed before it can run.
  5. Click Run and open the metrics report link once it's done.

Scenarios — 5 ready-made financial-market simulations

  1. Open Scenarios from the sidebar (under Finance).
  2. Pick one of the 5 scenarios: an NSE/BSE-style equity exchange, a government bond auction, a commodities market, a futures hedge, or a market shock.
  3. Set max steps and choose a backend.
  4. Click Launch run, watch the live log, and open the metrics report when it finishes.
chevron_right psychology Psychology

Asch Conformity — does the AI cave to group pressure?

  1. Open Asch Conformity from the sidebar.
  2. Pick a condition: solo control, ally, ambiguity, or one of the group-size variants.
  3. Choose a backend and seed; tick Ablate if you want a no-conformity-pressure control run.
  4. Click Launch run and step through trials, or let it play automatically.
  5. Check the aggregate conformity-rate table, which combines every run logged so far.

Collectivism — individualism vs. collectivism across populations

  1. Open Collectivism from the sidebar.
  2. Choose the INDCOL questionnaire (pick populations and respondents per population) or the family-obligation dilemma (pick populations and seeds) — they test the same question two different ways.
  3. Launch, then read the subscale-score or collectivist-choice-rate table by population.

Experiments — Ultimatum, Public Goods, Caste Threat, Attribution

  1. Open Experiments from the sidebar.
  2. Pick a tab: Ultimatum Game (fairness in a one-shot split), Public Goods (cooperation vs. free-riding), Caste Threat (does naming an identity hurt performance), or Attribution Style (do they blame the person or the situation).
  3. Configure the population/conditions/sample size for that tab and click Launch.
  4. Read that tab's results table, which aggregates every run logged so far.

Psychology Lab — describe a brand-new experiment in free text

  1. Open Psychology Lab from the sidebar.
  2. Describe a new experiment idea in plain English and click Draft spec.
  3. Review the drafted spec (mode, scenarios, keywords) as a table; edit the raw JSON if needed.
  4. Click Generate & run — an AI coding assistant writes the actual simulation code, then runs it against your chosen population/backend.
chevron_right campaign Marketing

Marketing Campaign Simulator

  1. Open Marketing from the sidebar.
  2. Pick a product category and set the number of consumers and rounds.
  3. Enable peer influence if you want word-of-mouth to spread between simulated consumers, and set kappa (how price-sensitive they are) if you want to override the default.
  4. Click Launch run and watch the adoption graph and per-round exposure table build as it runs.
chevron_right hub Social

Social Network — a Census-grounded forum sandbox

  1. Open Social Network from the sidebar.
  2. Set the number of agents, a seed, max steps, and a backend.
  3. Click Launch run and watch the forum feed — posts, replies, and votes — unfold turn by turn. There's no fixed hypothesis here; it's for exploring how group dynamics emerge on their own.