Cumulative structure and path length in networks of knowledge

Abstract

An important knowledge dimension of science and technology is the extent to which their development is cumulative, that is, the extent to which later findings build on earlier ones. Cumulative knowledge structures can be studied using a network approach, in which nodes represent findings and links represent knowledge flows. Of particular interest to those studies is the notion of network paths and path length. Starting from the Price model of network growth, we derive an exact solution for the path length distribution of all unique paths from a given initial node to each node in the network. We study the relative importance of the average in-degree and cumulative advantage effect and implement a generalization where the in-degree depends on the number of nodes. The cumulative advantage effect is found to fundamentally slow down path length growth. As the collection of all unique paths may contain many redundancies, we additionally consider the subset of the longest paths to each node in the network. As this case is more complicated, we only approximate the longest path length distribution in a simple context. Where the number of all unique paths of a given length grows unbounded, the number of longest paths of a given length converges to a finite limit, which depends exponentially on the given path length. Fundamental network properties and dynamics therefore characteristically shape cumulative structures in those networks, and should therefore be taken into account when studying those structures.