With a Little Help From My Friends: Collective Manipulation in Risk-Controlling Recommender Systems

Dimensions overall score 5.0

GitHub Code Pulse

No public code linked for this paper yet.

• Risk-controlling recommender systems that use binary user feedback are inherently susceptible to coordinated adversarial user behavior.technical 90%
• A small coordinated group comprising only 1% of the user population can induce up to a 20% degradation in nDCG for non-adversarial users.result 90%
• Effective adversarial strategies for manipulating risk-controlling recommenders should target low-risk items.result 85%
• Coordinated users cannot selectively suppress specific content groups through reporting alone in the studied system.limitation 80%
• A mitigation strategy that shifts guarantees from the group level to the user level can reduce the impact of adversarial coordinated behavior.method 85%
• Simple, realistic attack strategies require little to no knowledge of the underlying recommendation algorithm to harm recommendation quality.technical 80%
• The KuaiRand dataset is the only publicly available dataset that includes realistic 'Not Interested' feedback for studying risk-controlling recommenders.technical 75%
• User-level personalized calibration of the risk-controlling threshold can be compared to standard population-level calibration to measure performance gaps.method 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card