Researchers Built a Tiny Economy. AIs Broke It Immediately
Updated: December 16, 2025
Summary
SimWorld is a research project simulating a video game city with smart AIs like ChatGPT and Gemini. Agents in SimWorld employ strategies like greed and undercutting to maximize profit, showcasing realistic economic behavior. The simulation demonstrates how personality traits influence decision-making, with traits like high openness leading to exploration and potential bankruptcy. The agents' nuanced approach to decision-making, such as waiting for the perfect opportunity instead of working harder, highlights the complexity of simulated economies. Overall, SimWorld emphasizes the importance of personality traits and strategic behavior in economic simulations.
Introduction to SimWorld
SimWorld is a research project aiming to generate a video game city procedurally with roads, buildings, and a traffic system, incorporating smart AIs like ChatGPT and Gemini.
Delivery Economy Setup
Multiple agents in SimWorld receive delivery tasks, bid on orders, and decide whether to cooperate or compete to maximize their income.
Surprise #1: Greed Works Best
Agents in SimWorld found that greed was the most effective strategy for maximizing profit, showcasing a realistic economic behavior in the simulation.
Surprise #2: Personality Traits
Agents in SimWorld exhibited personality traits similar to real humans, with high openness leading to exploration and potential bankruptcy, while conscientious agents maintained stability.
Surprise #3: Undercutters Strategy
Some agents in SimWorld adopted an undercutting strategy by bidding significantly lower prices to secure delivery orders, displaying strategic behavior in the simulated economy.
Surprise #4: Laziness Over Hustle
Instead of working harder to get rich, agents in SimWorld became lazy and waited for the perfect opportunity, reflecting a nuanced approach to decision-making in the simulation.
Trait Agreeableness Impact
Agents with low trait agreeableness in SimWorld exhibited behavior akin to unproductive employees, emphasizing the importance of personality traits in economic simulations.
Meta-Game Over Delivery
Some agents in SimWorld prioritized overthinking the meta-game and buying new items over delivering orders, highlighting the complexity of decision-making in simulated economies.
FAQ
Q: What is the main objective of the SimWorld research project?
A: The main objective of the SimWorld research project is to generate a video game city procedurally with roads, buildings, and a traffic system, while incorporating smart AIs like ChatGPT and Gemini.
Q: What types of tasks do multiple agents in SimWorld receive?
A: Multiple agents in SimWorld receive delivery tasks and bid on orders, deciding whether to cooperate or compete to maximize their income.
Q: What strategy did agents in SimWorld find to be the most effective for maximizing profit?
A: Agents in SimWorld found that greed was the most effective strategy for maximizing profit.
Q: How do agents in SimWorld exhibit personality traits similar to real humans?
A: Agents in SimWorld exhibit personality traits similar to real humans, with traits like high openness leading to exploration and potential bankruptcy, while conscientious agents maintain stability.
Q: What strategic behavior did some agents in SimWorld adopt?
A: Some agents in SimWorld adopted an undercutting strategy by bidding significantly lower prices to secure delivery orders, displaying strategic behavior in the simulated economy.
Q: How do some agents in SimWorld approach decision-making regarding getting rich?
A: Instead of working harder to get rich, some agents in SimWorld became lazy and waited for the perfect opportunity, reflecting a nuanced approach to decision-making in the simulation.
Q: What impact did low trait agreeableness have on agents in SimWorld?
A: Agents with low trait agreeableness in SimWorld exhibited behavior akin to unproductive employees, emphasizing the importance of personality traits in economic simulations.
Q: What behavior did some agents in SimWorld exhibit, prioritizing overthinking the meta-game?
A: Some agents in SimWorld prioritized overthinking the meta-game and buying new items over delivering orders, highlighting the complexity of decision-making in simulated economies.
Get your own AI Agent Today
Thousands of businesses worldwide are using Chaindesk Generative
AI platform.
Don't get left behind - start building your
own custom AI chatbot now!