Algorithmic Assortative Matching on a Digital Social Medium

Conditionally Accepted at Information Systems Research 2021

Humans are increasingly interacting in and operating their daily lives through structured digital environments, mainly through apps that provide media for sharing photos, gaming, video watching, etc. Most of these digital environments are offered under “freemium” pricing to facilitate adoption and network effects. Users’ early social interaction and experience often have a substantial impact on their longer-term behavior. On this background, we study the impact of a new system that matches new users to existing communities in an assortative manner. We devise a machine learning-based matching system that identifies users with high expected value and provides them the option to join highly active teams (in terms of engagement and expenditure). We deploy this mechanism experimentally and find that assortative matching significantly increases user engagement, spending, and within-game output. This finding holds for more active communities and overall. Teams matched with low-activity new users are negatively impacted, leading to an overall more segregated social environment. We argue social experience and social behavior in groups are likely mechanisms that drive the impact of the matching system.

Conditionally Accepted at Information Systems Research

Download

Recommended citation: Lopez Vargas, K., Runge, J., Zhang, R. (2021). "Algorithmic Assortative Matching on a Digital Social Medium." Working Paper.