Skip to main content

Fifty Years of Research on Intransitivity of Preferences: A Process Perspective

Date
Date
Wednesday 5 June 2019, 1pm - 2pm
Location
TR.3 Meadows
Speaker
Rob Ranyard
Who can attend
Staff, students, alumni and external guests

Abstract

It is fifty years since Amos Tversky presented evidence of intransitive preferences in his seminal 1969 article. Early replications of his main study, which involved choices between simple monetary lotteries, replicated his findings. However, later research has cast doubt on the strength of evidence of intransitive choices in this task. For example, Regenwetter et al. (2011) concluded that “unambiguous evidence is currently lacking” of “empirical evidence of intransitivity by individual decision makers” (p. 42). In this paper we aim to shed light on this issue, which is at the core of human decision making, by further exploring Tversky’s theoretical insights, which were based on the notion of dimensional processing. The analysis we present here aims to build on Tversky’s work in ways that have previously been neglected. First, previous research has tended to neglect descriptions of choice behaviour, instead focusing on how well theoretical models fit the choice data. Here we present an exploratory descriptive analysis of choice patterns in Tversky’s lottery context. Second, it is surprising that most previous treatments propose alternative models but do not evaluate the goodness of fit of the models proposed by Tversky. Here we develop a two-parameter specification of a nonlinear, additive difference (NLAD) model, and describe the relationship between the model’s parameters and the transitivity or intransitivity of choice in the original lottery choice task.  That is, we identify the conditions under which the two-parameter NLAD model predicts intransitive, or alternatively, transitive preferences. We further show how different specifications of the model correspond to different dimension-based decision strategies and heuristics. We then review and reanalyze previous choice data from six replications of Tversky’s lottery choice experiment, and reconsider decision process evidence from verbal reports. We estimate the NLAD model’s parameters for each individual data set using maximum likelihood estimation, examines the goodness of fit of the model and use likelihood ratio tests to evaluate specific variants. We find that the two-parameter NLAD model has a very good fit to most individual choice patterns reviewed, with many being predictably intransitive. Furthermore, we find that many transitive patterns correspond to the application of simple, one-dimensional ‘take the best’ heuristics. The findings support the view that human decision making is often based on dimensional processing in such a way that evaluations of decision alternatives are relative to the set under consideration.