Content area
Full text
1. Introduction
The current paper defines the optimal sequential information acquisition process of a rational utility maximizing decision maker (DM) when allowed to acquire n pieces of information from a set of bi-dimensional products whose characteristics have a continuum of variants.
The optimal allocation of the information available to the DM is based on two well-defined real-valued expected utility functions. One of the functions defines the expected utility obtained from continuing gathering information on one of the previously partially observed products. The other function describes the expected utility that follows from starting observing the characteristics of a new product. The crossing points, if any, between the graphs of both functions correspond to optimal thresholds for the information gathering process that define the dynamic behavior of the corresponding algorithmic search structure.
Moreover, we incorporate a heuristic mechanism to the information acquisition process of the DM. When deciding whether to start gathering information on a new product or continuing with any of the products whose characteristics have been observed, the DM will use as a reference point to define his optimal behavior the threshold value arising from the two observations setting. This value determines the final forward looking decision of the DM when calculating how to use his last observation and is, at the same time, carried backwards to determine his current information acquisition strategy. That is, it affects his current choice behavior.
The reason for imposing this mechanism is twofold. First, it allows us to represent the optimal behavior of the DM within an easily understandable setting that can be simulated numerically. Second, it accounts for the information capacity constraints with which we endow the DM. That is, a DM can only engage in pairwise comparisons when choosing between his two main options of continuing acquiring information on an observed product or starting acquiring information on a new one.
Note, however, that the optimal behavior exhibited by a DM should be determined by all previously observed characteristics as well as the number of observations remaining to be acquired. If this were the case and the heuristic mechanism was not imposed, then, even in the current bi-dimensional product environment, the DM would have to consider all possible combinations of information acquisition results simultaneously and include them within...