Entropy, heterogeneity, and their impact on technology progress

Abstract

This study seeks to determine whether the entropy of patent assignees and the heterogeneity of patented technology within a technology domain positively contribute to the domain's influence on others. This question is motivated by the diversity-performance debates that have been explored across diverse disciplines. Three entropy indices are considered: Shannon, Herfindahl, and Lorenz indices. In addition, the semantic heterogeneity index is developed by employing a pre-trained deep neural network for word embedding. This study investigates about 2 million patents from 1976 to 2021 in the eight Cooperative Patent Classification (CPC) sections that constitute the entire patent landscape. The major findings are two folds. First, the semantic heterogeneity of knowledge created within a technology domain has a positive impact on its influence on others. Second, a negative impact can be exerted on a domain's influence, as the entropy of inventing entities increases. This suggests that a technology domain tends to be more influential when inventions are concentrated among a few prolific entities rather than being distributed across small entities.