Exploring the Escalation of Source Bias in User, Data, and Recommender System Feedback Loop

TL;DR

This paper investigates how AI-generated content influences recommender systems, revealing source bias escalation and proposing a debiasing method to maintain content diversity and recommendation quality over time.

Contribution

It introduces a simulation environment to study AIGC's impact on recommender systems and proposes a debiasing technique to mitigate long-term bias and performance decline.

Findings

AIGC is ranked higher in recommendations, indicating source bias.

Bias toward AIGC increases with user clicks and training on simulated data.

Long-term dominance of AIGC can reduce recommendation performance.

Abstract

Recommender systems are essential for information access, allowing users to present their content for recommendation. With the rise of large language models (LLMs), AI-generated content (AIGC), primarily in the form of text, has become a central part of the content ecosystem. As AIGC becomes increasingly prevalent, it is important to understand how it affects the performance and dynamics of recommender systems. To this end, we construct an environment that incorporates AIGC to explore its short-term impact. The results from popular sequential recommendation models reveal that AIGC are ranked higher in the recommender system, reflecting the phenomenon of source bias. To further explore the long-term impact of AIGC, we introduce a feedback loop with realistic simulators. The results show that the model's preference for AIGC increases as the user clicks on AIGC rises and the model trains on simulated click data. This leads to two issues: In the short term, bias toward AIGC encourages LLM-based content creation, increasing AIGC content, and causing unfair traffic distribution. From a long-term perspective, our experiments also show that when AIGC dominates the content ecosystem after a feedback loop, it can lead to a decline in recommendation performance. To address these issues, we propose a debiasing method based on L1-loss optimization to maintain long-term content ecosystem balance. In a real-world environment with AIGC generated by mainstream LLMs, our method ensures a balance between AIGC and human-generated content in the ecosystem. The code and dataset are available at https://github.com/Yuqi-Zhou/Rec_SourceBias.