Amid widespread conflict and division in the news, a UCLA study reveals striking parallels in how mice and AI systems learn to cooperate by pursuing common goals.
The results, published in Science, show that both biological brains and AI neural networks independently arrived at similar strategies and neural patterns when working together—pointing to universal principles of cooperation that go beyond the divide between biology and technology.
The Essential Role of Cooperation in Society and AI Development
Cooperation is a cornerstone of human society, critical for everything from workplace collaboration to global diplomacy. Gaining insight into how cooperation arises and persists could help address social conflict, inform treatments for social behavior disorders, and improve the design of cooperative AI systems.
Cooperation is crucial for boosting individual success and ensuring group survival, while its failure often results in harmful social conflict and instability.
With the growing complexity of artificial intelligence, researchers have discovered that both biological and artificial agents can develop comparable behavioral strategies and neural patterns. This opens up new avenues for exploring how cooperative behavior arises in AI systems and whether these interactions are guided by neural dynamics similar to those found in living organisms.
This study marks the first direct comparison of cooperative learning between biological brains and artificial intelligence, shedding new light on a key aspect of social behavior and helping to inform the development of more collaborative AI systems.
Innovative Behavioral Task Links Mouse and AI Cooperative Learning
UCLA researchers designed a novel behavioral task in which pairs of mice had to synchronize their actions within progressively shorter time frames—eventually as brief as 0.75 seconds—to earn rewards.
To track brain activity during the task, scientists used advanced calcium imaging to monitor individual neurons in the anterior cingulate cortex (ACC).
The researchers then developed AI agents using multi-agent reinforcement learning and trained them on a similar cooperative task in a virtual setting. This parallel design enabled a direct comparison of how biological and artificial systems acquire cooperative behavior.
The mice successfully learned to coordinate their actions to earn shared rewards. They adopted three main strategies: moving toward their partner’s side of the chamber, waiting for their partner before nose-poking, and engaging in social interactions before making decisions. These behaviors significantly increased with training, with interaction behaviors more than doubling as the mice improved their cooperative skills.
Neural Mechanisms of Cooperation in Mice and AI Agents
The study found that neurons in the anterior cingulate cortex (ACC) encoded these cooperative behaviors and decision-making processes. Mice that cooperated more effectively showed stronger neural representations of their partner’s actions. Notably, when researchers inhibited ACC activity, cooperative behavior dropped significantly, confirming that this brain region plays a crucial role in coordinating joint actions.
The AI agents developed remarkably similar strategies to those of the mice, such as waiting for their partner and precisely coordinating actions. In both biological and artificial systems, networks formed functional groups that enhanced their responsiveness to cooperative cues, with information about the partner becoming increasingly important as coordination improved.
When researchers selectively disrupted neurons linked to cooperation in the AI models, their performance declined significantly—demonstrating that specialized neural circuits are essential for effective cooperation in both biological and artificial systems.
Moving forward, the research team aims to explore whether similar neural mechanisms exist in other brain regions involved in social behavior. They also plan to investigate how understanding these core principles of cooperation could deepen our broader understanding of how social behaviors emerge and operate.
Bridging Biology and AI: Shared Principles of Cooperation
The similarities between biological and artificial systems indicate that insights from animal cooperation studies could guide the development of more advanced, collaborative AI. At the same time, AI models offer a valuable tool for testing theories about brain function that would be challenging to explore directly in living animals.
“We discovered remarkable similarities in how mice and AI agents learn to cooperate,” said Weizhe Hong, the study’s senior author and professor in UCLA’s Departments of Neurobiology and Biological Chemistry.
“Both systems independently formed comparable behavioral strategies and neural patterns, indicating that fundamental computational principles of cooperation exist beyond the divide between biological and artificial intelligence.”
This research is part of Hong’s larger investigation into prosocial behavior in both biological and artificial systems. His recent study published in Nature demonstrated that mice and AI develop strikingly similar “shared neural spaces” during social interactions.
Together with his 2024 Nature study on the role of the anterior cingulate cortex in helping others in pain and his 2025 Science research on rescue-like behavior in animals toward unconscious mice, these findings provide a detailed understanding of the neural mechanisms behind various forms of prosocial behavior.
“Grasping the nature of cooperation is essential for tackling many of society’s greatest challenges,” Hong said. “By examining how both biological brains and AI systems learn to collaborate, we can gain deeper insights into the neural foundations of human social behavior and develop more effective, cooperative artificial intelligence.”
Read the original article on: Tech Xplore
Read more: AI Can Write Your College Essay, But it Won’t Capture Your Unique Voice