Research in algorithmic responsibility at Spotify combines machine learning research with social science to promote high quality data decisions and equitable algorithmic outcomes. We carry out in-depth research, perform product-focused case studies, and develop practical tools that teams can apply. Research projects include safety modelling, measuring fairness in machine learning models, and establishing frameworks and tools for auditing recommender systems to mitigate potential negative impacts like reinforcement loops, exposure to harmful content, and algorithmic bias.