Tag: consensus algorithm

Abstract: This thesis aims at consensus-building on citations in Peer-to-Peer (P2P) systems. Citations, a source of various quantitative measures for intellectual products (e.g., scientific publications, patents, web pages), are more robust and productive if autonomous peers in a P2P system can determine and construct their true structure. However, this consensus-building has remained unreliable due to three problems that preceding studies have not addressed simultaneously: free-riding, strategic misreporting, and reviewer assignment. Therefore, we combined random walks on graphs with peer prediction methods and proposed two incentive mechanisms (ex-ante and ex-post consensus) that reward reviewers who participated in consensus-building. Experimental studies support the usefulness of the two incentive mechanisms for all three problems, by showing that peers can (i) be reviewers more often as they get higher PageRank scores and (ii) maximize the expected rewards per review by always reporting true beliefs. Our proposal—rewards from the consensus-building on citation relationships—also contributes to open-access intellectual products as an alternative scheme to grants, royalties, and advertisements. On the other hand, potential applications require future studies to prevent spamming and Sybil attacks and make the reward a sufficient incentive.

Abstract (日本語)　Main Texts (English)　SSRN: pre-print server