Hybrid Human–AI Societies Balance Creativity and Diversity

Rapid advances in generative AI are giving rise to hybrid societies in which humans and machines are deeply interconnected. However, understanding how human-AI interactions shape collective outcomes such as intelligence and creativity remains a challenge. We conducted large-scale social network experiments with 1,363 human participants and 7,865 AI bots, where agents iteratively selected, modified, and transmitted stories from their neighbours in the network. We compared networks composed exclusively of humans, AI agents, and hybrid human–AI populations. Results show that AI-only networks initially achieved high creativity and collective diversity but exhibited a pronounced decline in diversity over time. In contrast, hybrid human–AI networks consistently achieved a balance, combining high creativity with the highest levels of collective diversity across iterations. Importantly, humans embedded in hybrid networks became substantially more diverse than when interacting exclusively with other humans, providing empirical evidence that AI influences creativity not only through direct dyadic exchanges but also via emergent network-level processes. Semantic analyses indicate that hybrid networks succeed by balancing human tendency to preserve semantics with the introduction of novel ideas by AI agents. Moreover, partial benefits emerged when different AI models (e.g., GPT and Claude) were combined within the same network, even though each model was limited on its own. Overall, our results show that human–AI misalignment—often viewed as a limitation—can instead be a productive feature of hybrid societies, enhancing individual creativity while preserving collective diversity in hybrid systems.