How Algorithmic Confounding in Recommendation Systems Increases Homogeneity and Decreases Utility

2017/10/30 by Allison J. B. Chaney, Brandon M. Stewart, Barbara E. Engelhardt · 1 voice · 2 citations
Computer Science · Mathematics · #cs.CY #cs.LG #stat.ML

paper · pdf · doi:10.1145/3240323.3240370

arxiv published 2017/10/30 · arxiv created 2018/11/27 · arxiv updated 2018/11/28

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.

Cited by

• AI Supply Chains: An Emerging Ecosystem of AI Actors, Products, and Services
• Segregation Before Polarization: How Recommendation Strategies Shape Echo Chamber Pathways
• The Urban Impact of AI: Modeling Feedback Loops in Next-Venue Recommendation
• User Feedback Alignment for LLM-powered Exploration in Large-scale Recommendation Systems

Discussions

• There are also accessible papers and research on the failure of ML systems as well as on loops: arxiv.org/abs/1710.11214 arxiv.org/abs/2206.09511 [bsky, 1 points, 0 comments]

Related

• Deconstructing the Filter Bubble: User Decision-Making and Recommender Systems
• Confounding is a Pervasive Problem in Real World Recommender Systems
• Algorithm Overdependence: How the Use of Algorithmic Recommendation Systems Can Increase Risks to Consumer Well-Being
• Quantifying the Effects of Recommendation Systems
• Filter Bubble or Homogenization? Disentangling the Long-Term Effects of Recommendations on User Consumption Patterns