Forward-looking beyond expectations: Anticipating outcomes in environments with multi-attribute alternatives

The information retrieval behavior of decision makers (DMs) dealing with alternatives defined by multiple characteristics is generally determined by the expectations operator. Even when considering decision trees in sequential evaluation environments, the central assumptions imposed state that given a probability function defined on the set of potential realizations from a given alternative, the expected value suffices as a decision-making operator. However, suppose DMs evaluate the combinations of multiple characteristics relative to their expected values when selecting an alternative. Why don’t they incorporate these combinatorial possibilities beforehand into their information retrieval process? We analyze the consequences of incorporating this combinatorial behavior within an online information retrieval environment and illustrate the differences in utility that arise when evaluating sets of alternatives of different cardinality. We simulate 1,000,000 runs for different selection settings determined by the information retrieved by DMs before choosing and computing the utility received from implementing a standard expected utility and a combinatorial forward-looking strategy. We illustrate how this latter strategy pays off if DMs behave according to the postulates defining rational behavior through the whole decision process. This requirement implies evaluating the entire set of characteristics composing the alternatives before making a decision and considering their corresponding certainty equivalents as a benchmark to evaluate the alternatives. Finally, we use the information retrieval profiles generated to analyze the capacity of an artificial neural network to categorize DMs across the different evaluation and selection settings correctly.