Trust representation under confusion and ignorance

Kazuhiko Ohkubo, Tetsuhisa Oda, Yuki Koizumi, Tetsushi Ohki, Masakatsu Nishigaki, Toru Hasegawa, and Yoshinobu Kawabe
IWIN2018, September 2018. (Best Presentation Award)


Handling trust among agents/users is an important issue in the Internet, and many researchers have tackled on describing and analyzing trust values. Marsh and Dibben used a real value in the range of [−1, 1) to express a trust value, where value 1 corresponds to a state for an agent fully trusted and value −1 corresponds to a state of a complete distrust. However, there are cases where such a one-dimensional representation of trust is not sufficient; for example, we cannot give a proper trust value on an agent if the agent is both trusted and distrusted — this kind of confusion may happen since trust is a property closely related to human’s impressions. Also, ignorance on an agent may affect giving a trust value. To deal with trust values for cases where some confusion or ignorance is allowed, this study employs a multidimensional representation of Fuzzy-set Concurrent Rating method, FCR for short. The FCR method is originally developed for the study of mathematical psychology, and it can handle an answer like ”basically good, but somewhat bad simultaneously”. With the FCR method, we can describe trust values in a two-dimensional way, and this enables a proper description of trust states. In this study we introduce how to represent a trust value, and we discuss how the trust value corresponds to a one-dimensional trust value defined by Marsh and Dibben.