Annelien Smets

Recommender systems are not neutral tools; they are AI systems that actively shape human behaviour while being shaped by it in return. When platforms prioritise short-term revenue over content quality, these human-algorithm feedback loops can push ecosystems into self-reinforcing decline. This paper presents a game-theoretic model of co-adaptation between a platform deploying an algorithmic recommender and a population of creators responding to its incentives. The model yields two equilibria: a cooperative outcome with sustained high-quality content, and a defective outcome in which both parties prioritise short-term gains. The transition between them can be irreversible: a temporary relaxation of quality standards may shift the system beyond a tipping point from which restoring the original policy no longer recovers the prior equilibrium. The location of this tipping point depends on the initial health of the ecosystem, implying that early intervention is structurally more effective than delayed correction. The model is intentionally minimal, providing a baseline for future extensions while highlighting how human-AI system design can generate persistent societal effects.

Research on how people experience music emphasizes the importance of exploration and diversity in listening. However, music recommender systems struggle with facilitating exploration. Even when music recommender systems are able to recommend something valuable to users that is outside their typical preferences, it still remains difficult to spark their interest. This paper presents a user study examining the efficacy of immersive and informative introductions in stimulating interest in songs that are beyond one’s usual preferences, an experience called Taste-Broadening Serendipity. We uncover two important mechanisms behind the effect of introductions: transportation and cognitive elaboration. Our findings indicate that transportation (i.e., being absorbed into a narrative world) is the strongest predictor of Taste-Broadening Serendipity, while cognitive elaboration (i.e., learning something new about the artist or social context in which the music emerged) has a weaker effect but is easier to stimulate. We propose that song introductions can play an important role in facilitating exploration and increasing diversity of listening on music streaming platforms.

News recommender systems (NRSs) are increasingly implemented in the news industry to enhance digital distribution, yet their development and evaluation are shaped by a complex interplay of organizational stakeholders with diverse objectives. This study investigates the real-world complexities of NRS implementation, considering the multitude of objectives on the NRS provider side by exploring how different stakeholder groups —from journalistic, commercial, and tech logic— conceptualize NRS goals and assess their performance. Through 22 semi-structured expert interviews across four news organizations in Europe, we uncover a ‘jungle of metrics’ where an overabundance of data insights does not necessarily lead to better decision-making. While traditional accuracy-based metrics like click-through rate (CTR) remain widely used, they fail to capture the broader strategic objectives that guide NRS adoption in industry. This paper identifies a critical gap between current academic evaluations of NRS and the business key performance indicators (KPIs) used in practice. Rather than evaluating NRS in isolation, stakeholders assess their contribution to overarching organizational goals, including reach, engagement, conversion, and retention. Our findings highlight the need for a more nuanced, multi-stakeholder approach to NRS evaluation that aligns with both business imperatives and journalistic values.