Flooding-induced fragmentation in an adaptive bounded confidence model

The widespread use of social media has transformed how information spreads through society. In this paper, we examine how social media’s ability to amplify voices affects different types of information consumers. Building on Hegselmann and Krause’s bounded confidence model, we introduce adaptive parameters that mirror documented patterns of social media behavior: users becoming less receptive to divergent viewpoints and placing increasing weight on social consensus as their perceived peer group grows. Our simulations reveal that this micro-level adaptiveness leads to a macro-level vulnerability. While beneficial in normal conditions, it makes the system susceptible to fragmentation when misinformation agents (“social bots”) flood the environment. We identify a critical threshold in the proportion of these agents beyond which the community fails to converge on the truth. However, we also show that a sub-population of “epistemically responsible” agents with fixed, non-adaptive parameters are remarkably resilient and can act as anchors, preserving the system’s ability to find the truth. • An adaptive bounded confidence model incorporates social media behavioral patterns. • Agents reduce openness and reliance on evidence as their peer group size increases. • “Flooding” with misinformation agents exploits adaptiveness to fragment consensus. • A critical threshold of misinformation agents prevents convergence on the truth. • Non-adaptive agents function as epistemic anchors that enhance system resilience.