What does it take to eliminate Coalitional Manipulation in Online Discussion Forums?
Consider an online discussion forum where each participant is to be assigned a real valued credit score based on the value of her contribution (estimated based on popularity indicators such as likes, upvotes, etc.) in the forum. Notice that strategic participants can manipulate a credit score assignment scheme by forming coalitions, i.e., by strategically awarding popularity indicators among a subset of agents to maximize their credit scores. To prevent such coalitional manipulation, we propose a coalition resistant credit score function which discourages such strategic endorsements. Our key idea is to use community detection algorithms to identify close-knit communities in the graph of interactions and characterize a coalition identifying community detection metric. In particular, we show that the metric “modularity” is coalition identifying. We provide theoretical guarantees on modularity based credit score functions. Extensive simulations on illustrative datasets validate our theoretical findings.
Reference:
Ganesh Ghalme, Sujit Gujar, Amleshwar Kumar, Shweta Jain, Y. Narahari: Design of Coalition Resistant Credit Score Functions for Online Discussion forums. AAMAS 2018: 95-103.