Small worlds: Modeling attitudes toward sources of uncertainty

We introduce the concept of a conditional small world event domain--an extension of Savage's [The Foundations of Statistics, Wiley, New York, 1954] notion of a 'small world'--as a self-contained collection of comparable events. Under weak behavioral conditions we demonstrate probabilistic sophistication in any small world event domain without relying on monotonicity or continuity. Probabilistic sophistication within, though not necessarily across, small worlds provides a foundation for modeling a decision maker that has source-dependent risk attitudes. This also helps formalize the idea of source preference and suggests an interpretation of ambiguity aversion, often associated with Ellsberg-type behavior, in terms of comparative risk aversion across small worlds.