Skip to content

Antonio Mele

An agent-based political economy simulation built on a DSGE-HA (Dynamic Stochastic General Equilibrium with Heterogeneous Agents) framework.

Core trick

Each citizen agent gets a Claude decision per turn for economic, entrepreneurial, and voting choices, benchmarked against a numerical DSGE solver.

Tags

Description

An agent-based political economy simulation built on a DSGE-HA (Dynamic Stochastic General Equilibrium with Heterogeneous Agents) framework. 50+ citizen-agents with heterogeneous preferences, endowments, and values self-organize governance from scratch. An LLM (Claude) makes all economic, entrepreneurial, and political decisions for each agent, while a numerical solver provides a benchmark. The simulation tracks the emergence of property rights, voting rules, taxation, coalition dynamics, and measures outcomes such as Pareto efficiency, Gini coefficient, and social welfare.
meleantonio/AgentSociety★ 31 forksPythonCreated Feb 11, 2026 · last push Feb 22, 2026

README excerpt

# The Emergent Constitution

An agent-based political economy simulation built on a DSGE-HA (Dynamic Stochastic General Equilibrium with Heterogeneous Agents) framework. 50+ citizen-agents with heterogeneous preferences, endowments, and values self-organize governance from scratch. An LLM (Claude) makes all economic, entrepreneurial, and political decisions for each agent, while a numerical solver provides a benchmark. The simulation tracks the emergence of property rights, voting rules, taxation, coalition dynamics, and measures outcomes such as Pareto efficiency, Gini coefficient, and social welfare.

What happens when you give AI agents an economy and no rules? **The Emergent Constitution** is a heterogeneous agents economic simulation where 50+ citizen-agents build governance from scratch. Each agent has unique wealth, productivity, preferences, and ideological values. Every period, agents make consumption-savings decisions, choose between employment and entrepreneurship, propose constitutional rules, and vote on others' proposals. A Walrasian market clearing algorithm finds equilibrium prices across labor, capital, and bond markets. Agents can create any institution they want -- tax schedules, transfer programs, public goods, regulations, voting procedures -- through democratic voting. The twist: an LLM makes all decisions for each agent, receiving Bellman-equation-derived preferences as context so political choices are grounded in economic self-interest. The simulation 

Team