Group inference method of attribution theory based on Dempster–Shafer theory of evidence

Abstract

Kelley’s attribution theory has been widely popular in recent years. Lots of efforts have been spent on improving it with the assumption that there is only one expert to make attributions and the expert is assumed to be omniscient and omnipotent. However, such an assumption hardly exists in reality for the reason that the knowledge of each expert to make judgments is always limited. In order to solve this problem, this paper proposes a group inference method under the framework of Kelley’s attribution theory based on Dempster–Shafer theory of evidence. An information extraction mechanism is introduced to ensure that the real judgments of each expert can be well described. Then Shafer’s discounting is used to generate the basic probability assignment (BPA) functions by integrating the weights of experts on different criteria into the judgments of experts. The Dempster’s rule is employed to make fusion for the BPA functions, and a consensus reaching model which can increase the satisfaction degrees of group decision as much as possible is established to determine the probabilities of external and internal causes for the evaluated behavior. Finally, an algorithm is summarized, and illustrative example and discussion are provided to demonstrate its applicability.