Reading Between the Lines: A Study of Thematic Bias in Book Recommender Systems

Niche interests are underserved: Users with specialized or "long-tail" thematic preferences receive less personalized recommendations compared to users with diverse interests, meaning book recommender systems may inadvertently push certain readers toward mainstream content rather than helping them discover books aligned with their actual preferences.

Bias starts with data imbalances: The research identifies that thematic bias originates from imbalances in the underlying book collections themselves, which are then amplified by how users engage with recommendations—suggesting that fixing recommender systems requires attention to both the data and the algorithms that process it.

Popular themes get amplified: User engagement patterns reinforce existing biases, meaning that books fitting popular themes receive disproportionate visibility while underrepresented themes get neglected, potentially limiting the diversity of stories and perspectives available to readers.

Different users experience different fairness: The study reveals that fairness in recommendations is not uniform across user groups—some users benefit from personalization while others face systematic disadvantages based on their thematic interests, highlighting the need for recommender systems designed with diverse user needs in mind.

This problem extends beyond books: By demonstrating thematic bias in book recommendations, this work provides a framework that could be applied to other recommendation domains (music, film, news), suggesting that similar fairness issues may exist across multiple platforms that shape what content people discover.