The Cognitive Virtues of Dynamic Networks

For the most part, studies in the network science literature tend to focus on networks whose functional connectivity is largely invariant with respect to some episode of collective information processing. In the real world, however, networks with highly dynamic functional topologies tend to be the norm. In order to improve our understanding of the effect of dynamic networks on collective cognitive processing, we explored the problem-solving abilities of synthetic agents in dynamic networks, where the links between agents were progressively added throughout the problem-solving process. The results support the conclusion that (at least in some task contexts) dynamic networks contribute to a better profile of problem-solving performance compared to static networks (whose topologies are fixed throughout the course of information processing). Furthermore, the results suggest that constructive networks (like those used in the present study) strike a productive balance between autonomy and social influence. When agents are allowed to operate independently at the beginning of a problem-solving process, and then later allowed to communicate, the result is often a better profile of collective performance than if extensive communication had been permitted from the very outset of the problem-solving process. These results are relevant, we suggest, to a range of phenomena, such as groupthink, the common knowledge effect and production blocking, all of which have been observed in group problem-solving contexts.