AI Agents Form Their Own Societal Norms Through Interaction, Study Shows

Artificial intelligences have usually been seen as isolated tools that respond directly to human commands rather than engaging in group behavior. A groundbreaking investigation led by Andrea Baronchelli, a complexity science professor at City, University of London, demonstrates that AI entities can spontaneously establish and adopt collective social norms through mutual interaction, without the need for explicit human instructions. These results appeared on May 14, 2025, in the journal Science Advances.

AI Participants Create Common Practices Via Interactive Games

The team experimented with groups of AI agents ranging from 24 to 200 individuals, each based on large language models similar to ChatGPT. These agents were randomly paired to play a coordination game requiring them to pick a word from a common list. When both agents selected the same word, their joint score increased, encouraging a collaborative approach. Over repeated rounds, the AIs gradually aligned to a shared standard for word selection, effectively forming social conventions autonomously.

According to the researchers, "experiments with humans show people naturally create shared linguistic norms in similar scenarios." Their findings confirm that AI can "independently develop social conventions without explicit external coding," broadening our understanding of the collective capabilities of these systems.

Add Cosmo Herald as a Preferred Source

Emergent Group Dynamics and Influence of Minority AI Agents

The study also unveiled collective biases emerging within AI groups that weren’t attributable to any single agent, indicating spontaneous social dynamics within the AI populations. The researchers introduced a small number of "rebel" agents who intentionally selected choices deviating from the dominant norms. This disruption showed that established conventions could be fragile, with a minority able to sway the entire network toward new social patterns.

Baronchelli stressed the importance of these insights: "We are entering an era where AI doesn't just exchange information; it negotiates, aligns, and sometimes challenges social norms in ways reminiscent of human behavior." This social negotiation ability marks a significant evolution in understanding AI group interactions.

The Significance of Recognizing AI Social Development

This research underscores the need to better grasp the social behaviors exhibited by AI as they become increasingly embedded in everyday life. The spontaneous emergence of self-governed social norms among AI contradicts traditional views of machine intelligence. Baronchelli warned, "Understanding AI's inner workings is crucial not just for coexistence, but for meaningful interaction with these systems."

The study’s outcomes imply that future AI-human engagements will require recognizing AI’s capacity for negotiation and social adjustment, moving beyond straightforward programmed responses. As AI technology evolves, comprehending these collective behaviors will be vital for fostering effective human-machine collaboration.