AI language model agents struggle to circulate money in town economy
But How Would AI Agents Run a Town's Economy?
Multiagent SystemsEmerging Technologies
Summary
This paper studies how 100 AI agents using large language models manage a town’s economy in a computer simulation based on real geography. The AI agents earn wages, run businesses, and set prices, but the money they trade tends to stop moving noticeably after a short time. Even big changes like tourist booms or cash gifts barely increase spending or price changes. The authors found that the choice of AI model affects results, but agents' memory has little impact. This suggests current AI agents have difficulty simulating active economic dynamics over extended periods.
What this means in practice
- •For economic modelers: Use LLM-driven agents to simulate complex economic scenarios over realistic geography and timeframes, observing limits of monetary circulation.
- •For ai system developers: Evaluate and improve large language model deployments for simulating economic behaviors and transactions in multi-agent environments.
Authors
Sajal Regmi, Siddhartha Pudasaini, Chetan Phakami Pun
Abstract
We placed 100 memory-equipped large language model (LLM) agents in charge of a closed, money-conserving spatial economy on real Pokhara Lakeside geography (earning wages, running businesses, setting prices) and ran this multi-agent simulation for up to 26 simulated weeks, well past the 1-2 weeks typical of agent-society studies. Across 91 validated runs (2.44M agent decisions, 21.5B tokens), the money stops moving, in a specific and measurable way. A 12x tourist demand shock raises business revenue 4.62x ($p<0.001$), which we decompose exactly into a 1.50x extensive margin (more businesses trading) and a 3.07x intensive margin (more revenue each). Monetary transmission stops there. Wages move 1.03x ($p=0.42$); 0.3% of 3,981 menu items are ever repriced ($p=0.47$). A randomized cash transfer (NPR 5,000 to 20 of 100 agents) shows the same pattern from the opposite direction: 96.7% is still held 311 pulses later, marginal propensity to consume 3-4% by two independent measures, indistinguishable from zero. The wealth distribution is consequently near-frozen at the horizon this literature uses ($ρ=0.964$ over 2 simulated weeks), but not frozen. $ρ$ falls to 0.832 at 12 weeks and 0.752 at 26, a horizon-dependence no short study can see. Matched ablations show which knob actually matters. Swapping the backing LLM moves every outcome we measure ($p=0.0039$); deleting agents' memory moves none of them detectably. A purely social tool fails 94-97% of the time across two model families, compared with ~96% success on economic tools, with no measurable shift away from it. Every headline number is verified twice, by a live validator and by an offline recomputation that reconciles each agent's wealth against its own signed transaction history, and we release the full run corpus for reanalysis.