False-Name-Proof Recommendations in Social Networks

Conference item

We study the problem of finding a recommendation for an uninformed user in a social network by weighting and aggregating the opinions offered by the informed users in the network. In social networks, an informed user may try to manipulate the recommendation by performing a false-name manipulation, wherein the user submits multiple opinions through fake accounts. To that end, we impose a no harm axiom: false-name manipulations by a user should not reduce the weight of other users in the network. We show that this axiom has deep connections to false-name-proofness. While it is impossible to design a mechanism that is best for every network subject to this axiom, we propose an intuitive mechanism LEGIT+, and show that it is uniquely optimized for small networks. Using real-world datasets, we show that our mechanism performs very well compared to two baseline mechanisms in a number of metrics, even on large networks.

Authors: Brill, M; Freeman, R; Conitzer, V; Shah, N

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Association for Computing Machinery
• Host title: AAMAS '16: Proceedings of the 2016 International Conference on Autonomous Agents and Multiagent Systems
• Publication date: 2016-05-09
• ISSN: 1558-2914
• Keywords: Theory; Social choice; Economics; Recommendation Systems; False-name-proofness; Algorithms; Social networks