Algorithms increasingly shape what we see, hear, and engage with in daily life. Recommender systems, in particular, aim to connect users with relevant content through personalization mechanisms—yet, they often reproduce and amplify existing societal biases. In this talk, I examine the persistence of systematic biases, focusing on the complex interactions between algorithmic models, user behavior, and feedback loops.
Drawing on empirical research in music recommender systems, I highlight how underlying distributional patterns can undermine fairness interventions. Beyond identifying challenges, I will discuss approaches that integrate algorithmic adjustments with broader socio-technical considerations, raising important questions about responsibility, design choices, and how to evaluate systems beyond traditional performance metrics. This work reflects my commitment to developing systems that serve both consumers and providers more equitably.