Choice Basis, a Model for Multi-attribute Preference: some more Evidence

Abstract

Several models for riskless choice involving the concept of bounded rationality, have been presented and tested under various conditions. In the special case of a binary choice between alternatives characterized on several attributes, present models include essentially the maximin and maximax rules, the dominance rule, the conjunctive and disjunctive rules, the majority and the weighted set of differences rules, the choice by greatest attractiveness rule, the lexicographic, the minimum difference lexicographic and the lexicographic semi-order rules, the addition of utility differences and the sequential accumulation of differences rules. To account for expert’s information processing in a binary choice task, Barthelemy and Mullet (1986) proposed and tested a slightly more complex and flexible model, inspired by the work of Montgomery (1983). This model, called the moving basis heuristics, coordinates four types of rules: 1°) lexicographic rules, 2°) threshold rules, 3°) conjunctive rules, 4°) disjunctives rules. It builds on the principle that the dominance rule is used as a major one and that all the other rules are justused to obtain dominance structure as quickly as possible. Three basic principles are in fact involved in the model: (i) a parsimony principle, (ii) a reliability principle, (iii) a decidability principle. Empirical data supporting the model have been presented and discussed previously (Barthelemy and Mullet, 1986). In this paper we discuss the model in relation to other models and we present the results of three other experiments. The first two replicate the basic experiments; in the third one, experts’ verbal justifications are analyzed.