Dynamical complexity in the perception-based network formation model

TL;DR

This paper models social network evolution based on individuals' perceptions and strategic link formation, revealing how perception accuracy and initial conditions lead to complex, discontinuous changes in network structure.

Contribution

It introduces a perception-based strategic link formation model that accounts for individual perceptions and costs, highlighting the impact of initial conditions on network dynamics.

Findings

Stable link density exhibits discontinuous jumps depending on linking cost.

Dynamical complexity is influenced by initial perception biases.

Highly connected individuals facilitate information spread in heterogeneous networks.

Abstract

Many link formation mechanisms for the evolution of social networks have been successful to reproduce various empirical findings in social networks. However, they have largely ignored the fact that individuals make decisions on whether to create links to other individuals based on cost and benefit of linking, and the fact that individuals may use perception of the network in their decision making. In this paper, we study the evolution of social networks in terms of perception-based strategic link formation. Here each individual has her own perception of the actual network, and uses it to decide whether to create a link to another individual. An individual with the least perception accuracy can benefit from updating her perception using that of the most accurate individual via a new link. This benefit is compared to the cost of linking in decision making. Once a new link is created, it affects the accuracies of other individuals' perceptions, leading to a further evolution of the actual network. As for initial actual networks, we consider homogeneous and heterogeneous cases. The homogeneous initial actual network is modeled by Erd\H{o}s-R\'enyi (ER) random networks, while we take a star network for the heterogeneous case. In any cases, individual perceptions of the actual network are modeled by ER random networks with controllable linking probability. Then the stable link density of the actual network is found to show discontinuous transitions or jumps according to the cost of linking. As the number of jumps is the consequence of the dynamical complexity, we discuss the effect of initial conditions on the number of jumps to find that the dynamical complexity strongly depends on how much individuals initially overestimate or underestimate the link density of the actual network. For the heterogeneous case, the role of the highly connected individual as an information spreader is discussed.