Concave Expected Utility and Event Separability

We introduce a family of decision making models, referred to as eventseparable, based on a non-additive probability and on a general integration scheme. To characterize such models we take a different approach to independence and present the subjective codecomposable independence axiom that determines when the decision maker exhibits ambiguity neutrality. The new approach allows us to: (a) introduce the Concave Expected Utility model of decision making, adhering to ambiguity aversion where uncertainty is captured through a non-additive probability; and (b) provide sufficient conditions, weaker than those employed by previous formulations hinging on the independence axiom, to subjective and Choquet expected utility mod-