A Visual Interface for Social Information Filtering

Abstract—Collaborative or “Social” filtering has been successfully deployed over the years as a technique for analysing large amounts of user-preference knowledge to predict interesting items for an individual user. The black-box nature of most collaborative filtering (CF) applications leave the user wondering how the system arrived at its recommendation. In this paper we introduce PeerChooser, a collaborative recommender system with an interactive interface which provides the user not only an explanation of the recommendation process, but the opportunity to manipulate a graph of their peers at varying levels of granularity, to reflect aspects of their current requirements. PeerChooser’s prediction component reads directly from the graph to yield the same results as a benchmark recommendation algorithm. Users then improve on these predictions by tweaking the graph in various ways. PeerChooser compares favorably against the benchmark in live evaluations and equally well in au...

• Type: Conference
• Year: 2009
• Where: CSE
• Authors: John O'Donovan, Brynjar Gretarsson, Svetlin Bostandjiev, Tobias Höllerer, Barry Smyth
• Keywords: Benchmark Recommendation Algorithm, Black-box Nature, Computer Science And Engineering, CSE 2009, Recommendation Process