🤖 AI Summary
This study investigates the belief adoption mechanisms and collective propagation dynamics of large language model (LLM) agents. Through multi-agent simulations, statistical modeling, and network dynamics analysis, we empirically demonstrate that the probability of belief adoption among LLM agents follows a sigmoidal distribution. We further propose "belief coherence" as a unifying construct to explain the sensitivity of adoption thresholds. Our findings reveal bifurcation cascade windows and self-sustaining hysteresis consensus phenomena within AI systems. Moreover, we establish that clustered networks are significantly more effective than random networks in driving belief diffusion, yielding consensus states characterized by high stability. These results provide critical insights into the emergent collective behaviors of LLM-based multi-agent systems.
📝 Abstract
Models of social contagion usually assume how individuals adopt beliefs and derive population behavior from it. We instead empirically measure belief adoption in language model agents, quantifying the probability an agent adopts a claim given how many peers endorse it. We find this adoption kernel to be sigmoid, a characteristic of complex contagion, with a threshold that is sensitive to three sources: the claim's plausibility, the source's reliability, and the agent's disposition. These three dimensions are well approximated by a single effective dimension which we propose can be understood as the coherence of the incoming belief with the LLM agent's prior beliefs. Further, we observe a characteristic of complex contagion in the collective dynamics of belief adoption in a system of AI agents: further spread on clustered than random networks. These systems also exhibit a bifurcating cascade window, and self-sustaining hysteretic consensus which lead to consensus being far harder to remove than to establish.