Bounded confidence model

The bounded confidence model is a continuous-opinion dynamics model introduced by Guillaume Deffuant and David Neau in 2000 and independently by Rainer Hegselmann and Ulrich Krause. In this model, each agent holds a real-valued opinion on the interval [0,1]. Two randomly selected agents interact only if the absolute difference between their opinions is below a threshold ε. When they do interact, they move a fraction μ toward each other's opinion.

The threshold ε is the control parameter that determines the qualitative behavior of the system. When ε is large — agents are willing to interact with those whose opinions differ substantially — the population converges to a single consensus cluster. When ε is small, the population fragments into multiple disjoint opinion clusters that no longer influence each other. The transition between consensus and fragmentation is sharp: there exists a critical threshold ε_c below which the population splits and above which it unifies.

The bounded confidence model captures a fundamental social reality: influence requires proximity. Agents do not update their beliefs in response to arbitrarily distant opinions. The model's power lies in showing that this seemingly mild constraint — that only nearby opinions interact — can produce dramatic collective outcomes when iterated across a population. The fragmentation it predicts is not the result of hostility or censorship but of the local nature of social influence itself.

The bounded confidence model reveals that polarization does not require bad actors. It requires only that agents be slightly selective about whom they listen to — a condition so mild that it describes nearly every human social system. The threshold ε is therefore not a measure of open-mindedness but a measure of social structure: who is exposed to whom, and on what terms.