The effect of "irrelevant" variables on decision making: Criterion shifts in preferential choice?

Abstract In decision research, it is widely assumed that decision makers have subjective experiences which indicate definitively, albeit with random fluctuations, whether or not one alternative is preferable to another. Further, it is assumed that subjects' choices reflect this underlying preference order. It follows from this position, as a logical consequence rather than an empirical observation, that if a manipulation affects people's choices, it does so by changing the way people subjectively experience the alternatives (presumably via people's encoding or combination of attribute information). In psychophysics, an analogous theoretical position has been relinquished. Specifically, it is no longer believed that observers in a detection experiment have subjective experiences which indicate definitively, albeit with error, whether a signal has been perceived. Rather, according to signal detection theory, the observer's sensory experience can be portrayed as a quantifiable sensation that must be converted into an overt report of “signal present” or “signal absent” by comparing the sensation with a criterion. In this paper, we describe two models of preferential choice, the lexicographic criterion model and a probabilistic criterion model, that replace the notion of definitive preference with an analogue of signal detection theory. In the models, the decision maker is represented as having quantifiable inclinations, rather than definitive preferences, that must be converted into an overt choice by comparing the inclinations with a criterion. By attributing the effects of experimental manipulations to shifts in the criterion, rather than to changes in people's inclinations, the model offers a testable alternative to the proposition that altered choices imply altered preferences.