Algorithmic Bias in Recommender Systems: Implications for Consumer Choice and Information Diversity

Recommendation systems have become the core mechanism through which digital platforms shape information exposure and consumer choices. As a central infrastructure of platformized digital markets, these systems increasingly influence users’ attention allocation and decision-making processes across e-commerce, media, and social platforms. However, they exhibit systematic biases. This article analyzes biases, including popularity bias, exposure bias, ranking bias, preference reinforcement, and filter bubbles from three interrelated perspectives: consumer behavior, market incentives, and algorithmic system logic. It explores the formation mechanisms and structural impacts of these biases. The study finds that these biases are not isolated technical errors but structural outcomes jointly shaped by data distribution imbalance, ranking objective functions, and long-term reinforcement of user feedback loops. At the individual level, the recommendation mechanism reshapes the consumer choice framework, narrowing independent exploration; at the market and social levels, biases intensify visibility concentration and competition inequality and weaken information diversity. This article argues that to alleviate algorithmic biases, it is necessary to promote multi-party collaborative governance under transparency, fairness, and diversity goals. Based on these findings, the article argues that mitigating algorithmic bias requires multi-party collaborative governance that integrates transparency, fairness, and diversity-oriented objectives into platform design and regulation.