Workshop: Subset Selection in Machine Learning: From Theory to Applications
Multi-objective diversification via Submodular Counterfactual Scoring for Track Sequencing on Spotify
Recommender systems powering online multi-stakeholder platforms often face the challenge of jointly optimizing for multiple objectives, in an attempt to efficiently match suppliers and consumers. Examples of such objectives include user behavioral metrics (e.g. clicks, streams, dwell time, etc), supplier exposure objectives (e.g. diversity) and platform centric objectives (e.g. promotions). Jointly optimizing multiple metrics in online recommender systems remains a challenging task. In this work, we propose a multi-objective diversification technique via submodular optimization and counterfactual scoring to sequence music tracks in sequential recommendation scenarios. Based on a large scale live AB test deployment of the proposed sequencing method, we empirically demonstrate the value of submodular diversification in balancing across multiple objectives.