Lada Adamic
University of Michigan
To friend and to trust: eliciting truthful and useful ratings online

You need Flash player 8+ and JavaScript enabled to view this video.

Online rating and reputation systems have shown themselves to be essential for filtering content, building trust, and fostering communities. However, these ratings should not be taken at face value. When individuals submit ratings online, especially ratings of other people, they are being asked to quantify inherently subjective feelings. To complicate matters, they may formulate their ratings differently if these are shown to others, and if those others can reciprocate. In this talk I will present two studies that combine data analysis of several online data sets. For one such system, CouchSurfing.org, I will discuss findings from a large- scale survey and in-depth interviews to examine from multiple angles the challenges that users have in providing useful and truthful ratings. We find, for example, that the potential to reciprocate produces higher and more correlated ratings than when individuals are unable to see how others rated them. Ratings further can depend on the gender, age and nationalities of the raters and rates.All of these findings indicate that ratings should not be taken at face value without considering social nuances.