Recommender Systems for a Fair and Sustainable Future

Abstract

Recommender systems have become essential in navigating today’s information-rich digital landscape, shaping what billions of people watch, read, and buy. Yet, while optimizing for accuracy and engagement, the field has overlooked two urgent challenges: fairness and sustainability. This talk reflects on the role and purpose of recommender systems from these perspectives. I will examine the environmental impact of research and deployment, showing how the energy demands of state-of-the-art models translate into substantial carbon costs. At the same time, I will discuss how recommender systems influence fairness in access to information, opportunities, and cultural goods, and why existing metrics only capture part of this complex issue. Through examples ranging from e-commerce to healthcare, I argue that accuracy alone is not enough: recommender systems must be designed and evaluated with fairness and sustainability at their core. Building recommender systems for a fair and sustainable future is not optional — it is essential.