A paper of mine titled “Controversial Users demand Local Trust Metrics: an Experimental Study on Epinions.com Community” (pdf) got accepted for the Twentieth National Conference on Artificial Intelligence (AAAI-05)! Cool! The email I received this morning says “Your paper was one of 148 accepted to AAAI-05, out of 803 submissions. AAAI is a highly selective conference, and you are to be congratulated on your paper’s acceptance.” This means acceptance rate is 18%. Let me know if you like/dislike the paper or want to discuss its topic a bit. I think controversiality is an important theme and I think there are too many papers that assume that every user/agent has a global goodness value that is the same for everyone (there are some users that are bad for everyone and the goal of the technique is to spot them out). This assumption is unrealistic: just think of Bush or Berlusconi … some people like them (yeah, I know it’s kinda incredible) and some other don’t. My paper hopefully provide some evidence about this intuitive phenomena. You might also want to check other papers of mine.
Title: Controversial Users demand Local Trust Metrics: an Experimental Study on Epinions.com Community
Abstract: In today’s connected world it is possible and very common to interact with unknown people, whose reliability is unknown. Trust Metrics are a recently proposed technique for answering questions such as “Should I trust this user?”. However, most of the current research
assumes that every user has a global quality score and that the goal of the technique is just to predict this correct value. We show, on data from a real and large user community, epinions.com, that such an assumption is not realistic because there is a signicant
portion of what we call controversial users, users who are trusted and distrusted by many. A global agreement about the trustworthiness value of these users cannot exist. We argue, using computational experiments, that the existence of controversial users (a normal phenomena in societies) demands Local Trust Metrics, techniques able to predict the trustworthiness of an user in a personalized way, depending on the very personal view of the judging user.