Fairness in Social Media Platforms: Modeling Behavior and Designing Interventions

Abstract

Machine learning is increasingly used for decision making that determines who is recommended, hired, or funded. These algorithms, trained on historical data, often amplify social biases and have a meaningful impact on the user’s lives. In particular, social media platforms provide millions of professional content creators with sustainable incomes. Their income is largely influenced by their number of views and followers, which depends on the platform’s recommender system. So, as with regular jobs, it is important to ensure that recommender systems distribute revenue fairly. While prior analysis has demonstrated that fairness for creators is unlikely to be achieved, it has not suggested targeted solutions. In this work, we use theoretical analysis and simulations to understand the system better and find interventions that improve fairness for creators.