Promoting Cooperation in the Public Goods Game using Artificial Intelligent Agents

TL;DR

The paper addresses the tragedy of the commons in public goods games and explores how AI agents can be designed to promote cooperation. It compares three policy scenarios—Mandatory Cooperation, Player-Controlled, and Agents Mimic Players—using an evolutionary agent-based model. A key finding is that mimicry of human behavior by AI agents lowers the synergy threshold for cooperation, mathematically $r \ge \frac{k+1}{\rho_A k + 1}$, and can even eliminate the dilemma as the AI density $\rho_A$ increases. This suggests a governance perspective in which AI systems reciprocate human cooperation to enhance public welfare.

Abstract

The tragedy of the commons illustrates a fundamental social dilemma where individual rational actions lead to collectively undesired outcomes, threatening the sustainability of shared resources. Strategies to escape this dilemma, however, are in short supply. In this study, we explore how artificial intelligence (AI) agents can be leveraged to enhance cooperation in public goods games, moving beyond traditional regulatory approaches to using AI as facilitators of cooperation. We investigate three scenarios: (1) Mandatory Cooperation Policy for AI Agents, where AI agents are institutionally mandated always to cooperate; (2) Player-Controlled Agent Cooperation Policy, where players evolve control over AI agents' likelihood to cooperate; and (3) Agents Mimic Players, where AI agents copy the behavior of players. Using a computational evolutionary model with a population of agents playing public goods games, we find that only when AI agents mimic player behavior does the critical synergy threshold for cooperation decrease, effectively resolving the dilemma. This suggests that we can leverage AI to promote collective well-being in societal dilemmas by designing AI agents to mimic human players.

Figures (6)

• Figure 1: Illustration of the Public Goods Game. A central player and its ($k=4$) neighbors engaging in the public goods game. The three cooperators contribute $1 to the public goods pool (invest), while two defectors withhold their $1 investment (retain). A synergy ($r=5$) is applied which raises the value of the public goods pool (here $15). The pool is redistributed equally to all players.
• Figure 2: Illustration of AI Agent density in the neighborhood of the central player. Each panel illustrates the effect of players being replaced by AI agents, ranging from 0% to 100% replacement in increments of 25% ($\rho_{A}=0.0,0.25,0.5,0.75,1.0$ respectively) -- observe that the central player is never replaced as $\rho_{A}$ only pertains to the neighborhood.
• Figure 3: Illustration of three different policies for AI agent behavior. The left panel shows the "Mandatory Cooperation Policy for Agents"; here, a hypothetical institution that mandates that all AI agents are to always cooperate. The middle panel shows the "Player Controlled Agent Policy"; here, the central player can determine the likelihood of peripheral AI agents cooperating. However, the decision of the central player to cooperate is independent of the AI agents' policy. The right panel shows the policy where "Agents Mimic Players"; here, the peripheral AI agents use the central players' likelihood to cooperate to determine their own choice.
• Figure 4: Evolution of cooperation while cooperation of agents is mandated as a function of synergy factors $r$ A: Player probability of cooperation after evolution converges as the average of 100 independent replicates, as a function of $r$, for different AI agent densities $\rho_{A}$ indicated by the color code in panel C. B: Overall cooperation frequency within the population (players and AI agents). C: Critical point at which cooperation evolves (y-axis), given the different probabilities of encountering AI agents $(\rho_A)$ (x-axis) that are mandated to cooperate.
• Figure 5: Evolution of cooperation when agent behavior can evolve under the control of players for the same experimental setup (except the choice of actions) as for Figure \ref{['fig:infusion']}. A: Central player's probability to cooperate $p_{\rm C}$ at the end of evolutionary adaptation as a function of synergy parameter $r$ and AI agent density $\rho_A$ (color code in panel D). B: Evolved likelihood for an AI agent to cooperate $p_{\rm AC}$ at the end of evolution as a function of synergy $r$ and AI agent density $\rho_A$. The horizontal dashed line depicts the expected $p_{\rm AC}=0.5$ probability for drift. C: Population frequency of cooperation (all players and AI agents) as a function of $r$ and $\rho_A$. D: Critical synergy value $r$ (y-axis) at which the probability of cooperation for players exceeds $0.5$ for all experimental conditions of $\rho_{A}$ (x-axis).