Algorithmic Design for Mitigating Homophily Effects in Science

Past research has emphasized the importance of diversity in teams for innovation, including in science. However, homophily effects and existing awareness often influence scientists’ choices in whom to collaborate with. Research has shown that prescribed interaction can lead to new collaborations. Therefore, this research aims to design conference interactions in an effort to mitigate these effects on team formation. To achieve this, we designed a sorting algorithm to initiate connections among the participants who never collaborated before a conference. The algorithm assigned the participants to several group meetings and maximized each group's diversity based on its members’ gender, academic disciplines, and countries of residence. It also minimized repeated meetings among all possible pairs during multiple iterations of such conversations. We show that the groups generated using the algorithm are more diverse than those generated randomly. Hence, the sorting algorithm can catalyze novel interdisciplinary collaborations and, at the same time, combat homophilic effects among conference participants. We will describe extensions of this work to designing a novel online platform that aims to support idea generation, scientific collaboration, and innovation.