A test of task influences in uncertainty measurement

Abstract

Three experiments were carried out to test the hypothesized effect of response mode and event-type factors on measured uncertainty for frequentistic events observed within a realistic task setting. Subjects served as emergency vehicle dispatchers for a hypothetical city and gained experience with events (emergency calls) generated by a stationary stochastic process over a number of sessions. Experiment I compared two forms of judgment (frequency and probability estimation) over eight different kinds of observed events. Subsequently it compared subjects' predictive choice performance on selected event pairs with and without the benefit of prior estimation requirement. The results demonstrated a superiority of frequency over probability judgment and a reliable event-type influence on the quality of estimations produced. Moreover, prior judgment (frequency or probability estimation) seemed to facilitate predictive choice performance. Experiment II sought to clarify the relation between judgment and choice by manipulating the quality of prior estimations via combinations of instructional set and response requirement. Despite the differences in quality of frequency and probability judgments, choices were not directly affected by this variable. Rather, estimations served to cue frequentistic information in storage that was accumulated similarly by all experimental groups. This cuing hypothesis was tested further in Experiment III by comparing an estimation group with a directly cued group on predictive choice accuracy. The similarity of performance by the two groups was considered evidence in support of this cuing hypothesis.