Schelling segregation dynamics in densely-connected social network graphs (2504.16307v2)

Abstract: Schelling segregation is a well-established model used to investigate the dynamics of segregation in agent-based models. Since we consider segregation to be key for the development of political polarisation, we are interested in what insights it could give for this problem. We tested basic questions of segregation on an agent-based social network model where agents' connections were not restricted by their spatial position, and made the network graph much denser than previous tests of Schelling segregation in social networks. We found that a dense social network does not become as strongly segregated as a sparse network, and that agents' numbers of same-group neighbours do not greatly exceed their desired numbers (i.e. they do not end up more segregated than they desire to be). Furthermore, we found that the network was very difficult to polarise when one group was somewhat smaller than the other, and that the network became unstable when one group was extremely small; both phenomena may help explain the complexity of real-world polarisation dynamics, such as unique risks faced by very small group sin a society. Finally we tested Fossett's (2006) "paradox of weak minority preferences", a well-established result in grid- and map-based models which shows that an increase in the minority group's desire for same-group neighbours can create more segregation than a similar increase for the majority group. In a densely connected social network, we find that the evidence for this effect is mixed.

Schelling Segregation Dynamics in Densely-Connected Social Network Graphs

The paper "Schelling segregation dynamics in densely-connected social network graphs" provides a comprehensive analysis of the Schelling segregation model within the framework of agent-based models on dense social network graphs. The objective is to explore the intricacies of segregation and how it contributes to political polarization in contexts where geographical constraints are not predominant, a departure from traditional spatial models.

The authors initiated their investigation by evaluating how the density of connections affects segregation, contrasting it with scenarios in sparse networks. The results indicate dense networks do not become strongly segregated, and individual agent preferences for same-group neighbors do not exceed their initial desires. This aspect is intriguing, as it differs from results observed in less densely-connected networks.

Exploring the agent-based models with parameters such as agent tolerance and group size, the authors observe divergent results relative to previous models. The 'tipping point' for polarization in dense networks arises only under heightened levels of group intolerance, which illustrates how dense networks structurally resist polarization compared to traditional models. Notably, asymmetrical tolerance within groups reveals that even strongly intolerant groups fail to manifest segregation beyond specific thresholds.

Additionally, the size differential between groups in the network—traditionally known to exacerbate segregation—is explored thoroughly. Smaller groups exhibit significant instability, yet fail to polarize the network overall. Such interactions are further detailed in their explorations of the "paradox of weak minority preferences." Mixed evidence emerges regarding whether minority preferences have stronger segregation-inducing effects compared to majority preferences in dense networks.

The implications of this research are significant for understanding social behaviors in modern online environments. Dense networks akin to online social platforms resist simple models of polarization, suggesting new avenues for mitigating political polarization by focusing on network structure rather than purely on individual or collective preferences. Moreover, this paper prompts further investigations into secondary effects, such as ideological divides foregrounding polarization, and proposes embedding dimensionality as a valuable metric for observing network polarization dynamics.

Future explorations could consider extensions such as integrating ideological preferences or further analyzing fuzzy group membership to capture deeper dynamics. The groundwork laid here opens up substantial opportunities for addressing segregation and polarization in increasingly dense social networks. The focus on embedding dimensionality as a measure provides a novel lens for examining polarization and has potential applications across broader social network analyses beyond simple structural considerations.