The anatomy of a large-scale hypertextual Web search engine
WWW7 Proceedings of the seventh international conference on World Wide Web 7
Authoritative sources in a hyperlinked environment
Journal of the ACM (JACM)
The Eigentrust algorithm for reputation management in P2P networks
WWW '03 Proceedings of the 12th international conference on World Wide Web
Avoiding ballot stuffing in eBay-like reputation systems
Proceedings of the 2005 ACM SIGCOMM workshop on Economics of peer-to-peer systems
Prediction, Learning, and Games
Prediction, Learning, and Games
A content-driven reputation system for the wikipedia
Proceedings of the 16th international conference on World Wide Web
Algorithms and incentives for robust ranking
SODA '07 Proceedings of the eighteenth annual ACM-SIAM symposium on Discrete algorithms
Trusting spam reporters: A reporter-based reputation system for email filtering
ACM Transactions on Information Systems (TOIS)
Information and Computation
| Hi-index | 0.00 |
There are many online systems where millions of users post original content such as videos, reviews of items such as products, services and businesses, etc. While there are general rules for good behavior or even formal Terms of Service, there are still users who post content that is not suitable. Increasingly, online systems rely on other users who view the posted content to provide feedback. We study online systems where users report negative feedback, i.e., report abuse; these systems are quite distinct from much studied, traditional reputation systems that focus on eliciting popularity of content by various voting methods. The central problem that we study here is how to monitor the quality of negative feedback, that is, detect negative feedback which is incorrect, or perhaps even malicious. Systems address this problem by testing flags manually, which is an expensive operation. As a result, there is a tradeoff between the number of manual tests and the number of errors defined as the number of incorrect flags the monitoring system misses. Our contributions are as follows: We initiate a systematic study of negative feedbacks systems. Our framework is general enough to be applicable for a variety of systems. In this framework, the number of errors the system admits is bounded over the worst case of adversarial users while simultaneously the system performs only small amount of manual testing for multitude of standard users who might still err while reporting. Our main contribution is a randomized monitoring algorithm that we call Adaptive Probabilistic Testing (APT), that is simple to implement and has guarantees on expected number of errors. Even for adversarial users, the total expected error is bounded by "N over N flags for a given e Finally, we present empirical study of our algorithm that shows its performance on both synthetic data and real data accumulated from a variety of negative feedback systems at Google. Our study indicates that the algorithm performs better than the analysis above shows.