Events

An agent’s intuitive beliefs over a state space are modelled as the output of a network of associations between events. Intuitive beliefs are generically nonadditive and feature non-Bayesian updating. A characterization of additivity and Bayesian updating is provided. Formalizing the idea that the network is shaped by past experience, we hypothesize that intuitive beliefs minimize their difference from objective probabilities of “simple” events. It is shown that heterogeneous intuitive beliefs may arise from any given objective probability distribution. A characterization is provided for when intuitive beliefs over/under react to information relative to the Bayesian update of the objective distribution.