Believe and Let Believe: Axiomatic Foundations for Belief Dependent Utility Functionals
A large body of experimental data demonstrates that people’s beliefs influence their well-being beyond the indirect eﬀect through the actions taken. I present a model that incorporates beliefs into an agent’s utility function. The paper provides axiomatic foundations for a special class of non-additive utility indices deﬁned over inﬁnite streams of beliefs and actions. I assume that: 1) there exists a (null) belief that does not have any eﬀect on future preferences; 2) the agent has ﬁnite memory – only ﬁnite histories have an eﬀect on current preferences; and 3) if the agent knows she will not be getting any additional future information, she prefers today’s beliefs to be consistent with her past choices. Preferences satisfying these assumptions admit a generalized discounted utility representation in which all terms depend on both actions and beliefs. Experimental testability of the proposed framework is also discussed.
Yariv, Leeat, "Believe and Let Believe: Axiomatic Foundations for Belief Dependent Utility Functionals" (2001). Cowles Foundation Discussion Papers. 1606.