Influence of network topology on cooperative problem-solving systems

TL;DR

This study examines how different social network topologies affect the ability of groups to solve complex optimization problems, revealing that network structure influences performance depending on landscape ruggedness and group size.

Contribution

It provides a detailed analysis of how network topology impacts cooperative problem-solving efficiency in groups tackling NK fitness landscapes, highlighting the role of connectivity and information flow.

Findings

High connectivity and centralization improve performance on smooth landscapes.

Slowing information transmission benefits large groups on rugged landscapes.

Long-range links and modularity have limited effects except in specific cases.

Abstract

The idea of a collective intelligence behind the complex natural structures built by organisms suggests that the organization of social networks is selected so as to optimize problem-solving competence at the group-level. Here we study the influence of the social network topology on the performance of a group of agents whose task is to locate the global maxima of NK fitness landscapes. Agents cooperate by broadcasting messages informing on their fitness and use this information to imitate the fittest agent in their influence networks. In the case those messages convey accurate information on the proximity of the solution (i.e., for smooth fitness landscapes) we find that high connectivity as well as centralization boost the group performance. For rugged landscapes, however, these characteristics are beneficial for small groups only. For large groups, it is advantageous to slow down the information transmission through the network to avoid local maximum traps. Long-range links and modularity have marginal effects on the performance of the group, except for a very narrow region of the model parameters.