Can Users Fix Algorithms? A Game-Theoretic Analysis of Collective Content Amplification in...
summary
Paper models recommendation as a game using matrix completion and game theory and proves that coordinated uprating by a subset of majority users can force a rank-constrained recommender to retain minority items, yielding Pareto improvements in social welfare; analysis uses a K-truncated SVD abstraction called the alpha-loss-tolerant learner and majority-minority block matrices. Empirical work includes Goodreads experiments and a Twitter audit with 1,120 bots and >20K impressions showing For You feeds amplify right-leaning content; authors provide an algorithm to find effective collectives and show it materially improves both user welfare and recommender utility.
AI-written summary · about
chapters (8)
- 0:00Collective content amplification
- 3:55User agency in recommendations
- 9:01Recommendation systems as games
- 21:11Activist user deviations
- 29:04Altruism in recommendations
- 37:51Trust decline in algorithms
- 43:30Auditing algorithms with AI bots
- 54:11Q&A
related talks
26mA Game-Theoretic Perspective on the Recommender (Eco-)System
32mLet's talk about Recommendation Systems!
20mBuilding Recommender system
54mKeynote - Fairness and Diversity in Online Social Systems
37mA primer on recommendation systems.
52mReinforcement Learning in Recommender Systems: Some Challenges
20mLucky Talk: Low-regret Recommendations and Item-item Collaborative Filtering
45mAGT and Machine Learning