Promoting Cooperation in the Public Goods Game using Artificial Intelligent Agents

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.

💡 Research Summary

**
The paper investigates how artificial intelligence (AI) agents can be used to promote cooperation in the public goods game (PGG), a classic model of the tragedy of the commons. Three policy scenarios are examined: (1) a Mandatory Cooperation Policy that forces every AI agent to always cooperate, (2) a Player‑Controlled Policy in which human players evolve an auxiliary probability that determines how likely surrounding AI agents are to cooperate, and (3) an Agents‑Mimic‑Players policy where AI agents copy the cooperation probability of the central human player.

Using an agent‑based evolutionary simulation, the authors first derive analytically that the critical synergy factor r_c (at which cooperation becomes individually advantageous) depends only on group size k and not on the proportion of cooperators. Consequently, any policy that merely changes the density of cooperating agents without altering the payoff structure should not shift r_c.

In the Mandatory Cooperation scenario, AI agents are programmed to cooperate unconditionally. Simulations confirm that while the overall frequency of cooperative actions rises proportionally with the AI density ρ_A, the human players’ evolved cooperation probability p_C remains a function of r identical to the baseline case. The dilemma persists because human payoffs are unchanged; the presence of “always‑cooperating” AI does not make cooperation a better response for humans when r < k + 1.

The Player‑Controlled scenario introduces a second evolvable trait p_AC that determines the likelihood that AI agents in a player’s neighbourhood will cooperate. Evolution optimizes both p_C and p_AC. When it is beneficial, both traits converge to 1, making AI agents cooperate whenever they appear. This raises the total cooperation level in the population, yet the critical synergy threshold remains unchanged. Human players still face the same incentive gap, so the tragedy of the commons is not resolved.

The third and decisive policy, Agents‑Mimic‑Players, lets AI agents adopt the exact cooperation probability of the central human player. In effect, AI agents become behavioral mirrors: if the human cooperates, nearby AI agents also cooperate; if the human defects, they defect. This creates strategic homogeneity between humans and AI, dramatically increasing the payoff to cooperation because the number of cooperators in any group rises in tandem with the human’s own decision. Simulations show that the critical synergy factor r_c shifts to substantially lower values (e.g., from k + 1 ≈ 5 down to ≈ 2–3), meaning that cooperation becomes the rational choice even when the multiplication factor is modest. Moreover, the effect scales with AI density: higher ρ_A amplifies the mimicry impact, making cooperation robust as AI agents become more prevalent.

The authors argue that policies which merely regulate AI behavior (forcing cooperation) or leave it to market/owner decisions (player‑controlled) are insufficient because they do not alter the underlying incentive structure for human players. By contrast, embedding AI agents that dynamically mimic human strategies directly reshapes the payoff landscape, lowering the threshold at which cooperation dominates. This insight has broad policy implications: governments could mandate “behavior‑mirroring” AI designs in domains such as autonomous vehicles, energy grids, or social media algorithms to harness AI as a catalyst for collective welfare.

In summary, the study demonstrates that AI agents can be powerful tools for resolving social dilemmas, but only when they are designed to adaptively mirror human actions, thereby creating a feedback loop that makes cooperative behavior individually advantageous. The findings suggest a new direction for AI governance—moving from static regulation toward dynamic, behavior‑aligned AI systems that promote societal cooperation.