Log in / Register
Home arrow Economics arrow Bounded rational choice behaviour
< Prev   CONTENTS   Next >

5.1. Motivation

The modeling of individual decisions in transportation has dominantly relied on rational choice models. The best-known example is the discrete choice models based on random utility theory (e.g. McFadden, 1974). Although random utility theory has proven its value in an impressive number of academic and applied research projects, the underlying assumption of fully rational behaviour may not be particularly valid in all application contexts. This seems especially true for complex decision problems that involve many choice alternatives, and combinations of multiple sub-decisions. Under these circumstances, the concept of bounded rationality seems more appealing as it Takes into account the cognitive limitations of the decision-maker – limitations of both knowledge and computational capacity' (Simon, 1987).

Inspired by Simon, theories based on the principle of bounded rationality have been formulated in different disciplines and take on quite different forms. Heuristic models are among them, proposed mainly from psychology, such as conjunctive rules, disjunctive rules (e.g. Dawes, 1964), lexicographic rules (e.g. Fishburn, 1974) and elimination-by-aspect rules (e.g. Payne, Bettman, & Johnson, 1988; Tversky, 1972). These heuristics explicitly model search behaviour for information, stopping search, and deciding with limited information, to get satisfactory rather than optimal results. Many of these heuristic models imply non-compensatory information processing mechanisms as opposed to compensatory mechanisms that are widely used in discrete choice models, which assume that people trade off attribute utilities.

A few researchers have attempted to infer what heuristics people use in decisionmaking by observing whether some attributes are dominant for the decision in the sense that other attributes do not affect the decision outcomes at all (e.g. Hess, Rose, & Polak, 2007; Scott, 2002). However, such methods can only be successful in specific experimental settings with limited number of attributes in order to exactly infer the use of heuristics, which considerably limits the generality and applicability of the methods in usually more complex real-world situations and experiments. In contrast, approximation models seem to be a more practical way of estimating heuristics, usually based on compensatory utility models. For example, Parker and Srinivasan (1976) developed dedicated software, LINMAP, to estimate whether a decision process is single-staged or multi-staged (lexicographic) by analyzing the parameters of a conventional linear utility function. Swait (2001) showed that incorporating attribute cutoffs and phased utility functions into conventional logit models can approximate disjunctive and conjunctive rule. However, the limitation of such models is that information search processes cannot be identified and the attribute cutoffs were self-explicatory. Jedidi and Kohli (2008) showed how to infer lexicographic rules from linear compensatory utility functions. The limitation of their study is that they constrained the parameter space in order to derive the parameterizations representing lexicographic rules. Then, they used an iterative algorithm to assign each observed choice behaviour to the most likely lexicographic rule.

This chapter proposes a modeling framework that attempts to model both an assumed mental representation, information processing and the resulting preferences for attribute profiles. The framework is based on the assumption that people do not necessarily process all factors, and will simplify the decision task to reduce mental burden by using cognitive thresholds to build a mental representation of the decision problem. The use of thresholds in choice models has a long history going back to Manski (1977), with more recent work conducted by Gillbride and Allenby (2004) and Cantillo and de Dios Orthzar (2006). Much of this work has been conducted in the context of identifying choice sets or in the context of developing hybrid compensatory, non-compensatory decision rules (e.g. Cantillo & de Dios Ortuzar, 2005; Timmermans, Borgers, & Veldhuisen, 1986). As will become evident in the remainder of this chapter, the proposed framework is more elaborate.

Second, the mental representation results in preference structures, which are assumed to be the source from which heterogeneous decision heuristics that are potentially applied by the decision-makers can be exactly inferred. Third, a decision outcome is modeled as a latent-class structure with the decision-maker choosing a decision heuristic first, followed by making the decisions based on the heuristic. The outcome of this choice process is assumed to be influenced by mental effort, risk perception and expected outcome of the heuristics. This is in line with, for example, Swait and Adamowicz (2001) and Arana, Leon, and Hanemann (2008), but the explanatory factors and heuristic complexity are defined differently. Fourth, the priority and sequence of information search can be inferred from the estimated probabilities of heuristic use, which has not been attempted in existing research.

The modeling framework is implemented into two types of typical decision problems. The first is the satisficing problem in which people make binary judgment (e.g. accept & reject) on an object. The second is the comparative decision in which people pick out an object from several alternatives. The models are illustrated using a dataset about pedestrians' shopping behaviour in shopping streets.

The organization of the chapter is as follows: The modeling framework and the two model implementations are introduced first. Then, the models are applied to two decision problems: one is the decision of going home; the other is the decision of walking direction choice. Conventional multinomial logit models (MNL) are also estimated for comparison. The last section concludes the chapter with future directions.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science