How Algorithmic Confounding in Recommendation Systems Increases Homogeneity and Decreases Utility

Abstract

Recommendation systems are ubiquitous and impact many domains; they have the potential to influence product consumption, individuals' perceptions of the world, and life-altering decisions. These systems are often evaluated or trained with data from users already exposed to algorithmic recommendations; this creates a pernicious feedback loop. Using simulations, we demonstrate how using data confounded in this way homogenizes user behavior without increasing utility.

References in corpus (5)

• European Union regulations on algorithmic decision-making and a "right to explanation"
• Estimating the Causal Impact of Recommendation Systems from Observational Data
• A probabilistic model to resolve diversity-accuracy challenge of recommendation systems
• Modeling mutual feedback between users and recommender systems
• Reducing Offline Evaluation Bias in Recommendation Systems

Cited by in corpus (54)

• Causal Intervention for Leveraging Popularity Bias in Recommendation
• Evaluating Stochastic Rankings with Expected Exposure
• Deconfounded Recommendation for Alleviating Bias Amplification
• Fighting Fire with Fire: Using Antidote Data to Improve Polarization and Fairness of Recommender Systems
• Correcting for Selection Bias in Learning-to-rank Systems
• Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation
• A Graph-based Approach for Mitigating Multi-sided Exposure Bias in Recommender Systems
• User-controllable Recommendation Against Filter Bubbles
• A Classification of Feedback Loops and Their Relation to Biases in Automated Decision-Making Systems
• Exploring Longitudinal Effects of Session-based Recommendations
• Debiasing Recommendation with Personal Popularity
• Measuring Recommender System Effects with Simulated Users
• A Situated-Infrastructuring of WhatsApp for Business in India
• Dynamic Causal Collaborative Filtering
• Runtime Monitoring of Dynamic Fairness Properties
• Distributionally-Informed Recommender System Evaluation
• Much Ado About Gender: Current Practices and Future Recommendations for Appropriate Gender-Aware Information Access
• Disentangling User Interest and Conformity for Recommendation with Causal Embedding
• Scalable Realistic Recommendation Datasets through Fractal Expansions
• How Do Fair Decisions Fare in Long-term Qualification?
• Towards Psychologically-Grounded Dynamic Preference Models
• It's Not You, It's Me: The Impact of Choice Models and Ranking Strategies on Gender Imbalance in Music Recommendation
• Popularity Bias Is Not Always Evil: Disentangling Benign and Harmful Bias for Recommendation
• Practitioners Versus Users: A Value-Sensitive Evaluation of Current Industrial Recommender System Design
• Do Offline Metrics Predict Online Performance in Recommender Systems?
• Oh, Behave! Country Representation Dynamics Created by Feedback Loops in Music Recommender Systems
• Exploration in Online Advertising Systems with Deep Uncertainty-Aware Learning
• Recommenders with a mission: assessing diversity in newsrecommendations
• Understanding Filter Bubbles and Polarization in Social Networks
• Unintended Selection: Persistent Qualification Rate Disparities and Interventions
• Reconciling the Quality vs Popularity Dichotomy in Online Cultural Markets
• Understanding Longitudinal Dynamics of Recommender Systems with Agent-Based Modeling and Simulation
• Deconstructing the Filter Bubble: User Decision-Making and Recommender Systems
• Deep Bayesian Bandits: Exploring in Online Personalized Recommendations
• Beyond Item Dissimilarities: Diversifying by Intent in Recommender Systems
• Correcting the User Feedback-Loop Bias for Recommendation Systems
• Understanding or Manipulation: Rethinking Online Performance Gains of Modern Recommender Systems
• Multiversal Simulacra: Understanding Hypotheticals and Possible Worlds Through Simulation
• To Recommend or Not? A Model-Based Comparison of Item-Matching Processes
• DCAF: A Dynamic Computation Allocation Framework for Online Serving System
• Performative Debias with Fair-exposure Optimization Driven by Strategic Agents in Recommender Systems
• Scaling Up Collaborative Filtering Data Sets through Randomized Fractal Expansions
• Correcting Exposure Bias for Link Recommendation
• Simulation as Experiment: An Empirical Critique of Simulation Research on Recommender Systems
• Recommendation and Temptation
• Reusable Self-Attention Recommender Systems in Fashion Industry Applications
• Not All Videos Become Outdated: Short-Video Recommendation by Learning to Deconfound Release Interval Bias
• User and Recommender Behavior Over Time: Contextualizing Activity, Effectiveness, Diversity, and Fairness in Book Recommendation
• Individual Fairness for Social Media Influencers
• When the Label Ignores the Request: Auditing Policy-Selected Targets in Synthetic Conversational Music Recommendation
• Trend-responsive User Segmentation Enabling Traceable Publishing Insights. A Case Study of a Real-world Large-scale News Recommendation System
• Quantifying the Effects of Recommendation Systems
• Searching Personal Collections
• Position Paper on Simulating Privacy Dynamics in Recommender Systems