Evolution of Cooperation and Cooperative Artificial Intelligence

A special issue of the Proceedings of the National Academy of Sciences USA highlights the connection between evolutionary game theory and AI – featuring three articles with contributions from the Max Planck Institute for Evolutionary

June 17, 2025

To the Point:

Diagram illustrating agents in prosperous and degraded states with matrices showing cooperation and defection outcomes, accompanied by transition probabilities (pr, pc) indicating recovery and collapse likelihoods. — Two agents face a classic dilemma: cooperation benefits all, but short-term selfishness is tempting. In this simulation, their choices also affect the environment – the more they defect, the greater the risk of collapse.

© W.Barfuss

Two agents face a classic dilemma: cooperation benefits all, but short-term selfishness is tempting. In this simulation, their choices also affect the environment – the more they defect, the greater the risk of collapse.

© W.Barfuss

Collective Cooperative Intelligence: In one article co-authored by Christian Hilbe (now at IT:U Linz), the focus lies on how agents seeking to maximise their own benefit can learn to cooperate with one another.
Enforced Cooperation: Another article, also involving Christian Hilbe, explores how incentives can be aligned with a self-interested opponent in complex and dynamic environments.
Strategy Selection for Cooperation: Julian Garcia (Melbourne) and Arne Traulsen discuss how outcomes in models of cooperation are influenced by field-specific traditions in strategy selection – highlighting opportunities for mutual learning between evolutionary game theory and cooperative AI.

Overall, the collection of three articles addresses several current research topics that are relevant both to the scientific community and to broader societal and technological developments. A central focus lies in linking concepts from evolutionary theory with methods from artificial intelligence. This interdisciplinary perspective offers new approaches to analysing and promoting cooperative behaviour.

At the same time, the articles explore how these insights can be applied to practical contexts – for instance, in international collaboration or in the design and regulation of autonomous systems. In addition, the contributions support the public discourse on artificial intelligence by highlighting the importance of cooperative structures in digital processes and examining their role in the interaction between humans and machines.

How does cooperation emerge – and under what conditions does it remain stable?

Three new scientific publications address this fundamental question and provide impulses for contemporary societal debates around artificial intelligence and global cooperation.

The interdisciplinary approach – at the intersection of biology, mathematics, and computer science – demonstrates how theoretical concepts can lead to practical insights.

"Cooperation is a key issue in many societal and technological contexts," says Christian Hilbe. "Our work shows how a well-grounded, cross-disciplinary perspective can open up new pathways for understanding and actively fostering collective action."