Bias Mitigation for AI-Feedback Loops in Recommender Systems: A Systematic Literature Review and Taxonomy

2025-08-28

AI Summary

Recommender systems operating in AI feedback loops undergo continuous retraining, which risks amplifying biases and degrading long-term fairness; yet existing debiasing methods are predominantly validated on static datasets, leaving their dynamic robustness poorly understood. Method: We conduct the first systematic literature review focused on bias mitigation within AI feedback loops, synthesizing 24 studies (2019–2025) empirically validated via multi-round simulations or online A/B tests. Contribution/Results: We propose the first six-dimensional taxonomy—covering mitigation algorithms, bias types, dynamic settings, evaluation foci, application domains, and ML tasks—revealing three critical gaps: absence of shared simulation platforms, insufficient joint evaluation of fairness and utility, and lack of standardized metrics. Our framework supports industry practitioners in method selection and guides academic research toward long-term fairness modeling, fairness-utility co-optimization, and benchmark standardization.

Technology Category

• Philosophy and Ethics of AI: Bias, Fairness & Equity
• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Recommender systems continually retrain on user reactions to their own predictions, creating AI feedback loops that amplify biases and diminish fairness over time. Despite this well-known risk, most bias mitigation techniques are tested only on static splits, so their long-term fairness across multiple retraining rounds remains unclear. We therefore present a systematic literature review of bias mitigation methods that explicitly consider AI feedback loops and are validated in multi-round simulations or live A/B tests. Screening 347 papers yields 24 primary studies published between 2019-2025. Each study is coded on six dimensions: mitigation technique, biases addressed, dynamic testing set-up, evaluation focus, application domain, and ML task, organising them into a reusable taxonomy. The taxonomy offers industry practitioners a quick checklist for selecting robust methods and gives researchers a clear roadmap to the field's most urgent gaps. Examples include the shortage of shared simulators, varying evaluation metrics, and the fact that most studies report either fairness or performance; only six use both.

Problem

• Addressing bias amplification in recommender systems' AI feedback loops
• Evaluating long-term fairness of mitigation methods in dynamic settings
• Identifying research gaps through systematic literature review and taxonomy

Innovation

• Systematic review of bias mitigation in feedback loops
• Taxonomy organizes methods across six key dimensions
• Identifies gaps like shared simulators and metric consistency