Representational Similarity and Model Behavior in Multi-Agent Interaction

Researchers have shown that neural similarity among humans predicts social closeness and cooperative success, whereas innovation often emerges from interactions among dissimilar individuals. We investigate whether these principles extend to artificial intelligence by examining interactions between large language models. In our experiments, 276 model pairs interact across eight games spanning both cooperation and novelty. We find that pairs with more similar representation spaces achieve significantly higher cooperation but exhibit reduced novelty and creativity. The effects of representational similarity on cooperation and novelty remain robust even after controlling for other factors such as performance disparity and model size. We also find that similarity in the early layers consistently shows the strongest association with cooperation and novelty, compared to the middle and later layers. This suggests that a central factor underlying these patterns could be the extent to which the two models share lexical and semantic grounding. Overall, representational similarity can be an important consideration in multi-agent system design.

When two people's brains process the world in similar ways, they tend to cooperate well, but fresh, creative ideas more often come from people who think differently. We asked whether the same holds for AI. Using 276 pairs of AI models, we had each pair play eight games testing either cooperation or creativity. We found a clear trade-off: models that "think" more alike — that represent information similarly inside — cooperated better but produced less creative work. The similarity that mattered most showed up in how the models first read the words they're given, suggesting that two AIs cooperate well largely when they share a common understanding of language. As we build systems where multiple AIs work together, how similarly they "think" may be worth careful attention.