Effects of amount of information on quantitative judgment.

Abstract

Individuals sometimes repeatedly perform a judgment task using a given set of cues. Then one or more other cues are added to that set. This dissertation addresses the effects of such cue-addition on the accuracy and process of making quantitative judgments. Conceptual models and statistical procedures for isolating cue-addition effects are developed and tested with two measures, the achievement index ($r\sb{a}$) and mean squared error (MSE). The empirical results show that conclusions about cue-addition effects depend on the accuracy measure employed. Furthermore, it is demonstrated how the overall accuracy measures, $r\sb{a}$ and MSE, rely on their own component accuracy measures. In terms of the achievement index, $r\sb{a}$, most of its components decline as the amount of information increases, except for the measure of predictability $R\sb{e}$. Specifically, when more information is available, judges have a worse overall fit in their predictions and experience greater difficulty capturing the correct policy in utilizing cues. However, for judges whose original cue weights display inverted ordering, adding information which is highly correlated with the initial cues can be beneficial. Judges are usually unable to reliably anticipate nonlinear/configural deviations of criterion values from the predictions of optimal linear models, regardless of the number of cues presented to them. Therefore, whether increased information yields an increase or a decrease in $r\sb{a}$ depends on the tradeoff between the increased inherent predictive power contributed by extra cues and the extra demands on processing capacity made by those cues. To evaluate cue-addition effects on MSE, two of its components are investigated. In terms of the variability of predictions, judges have less variability in their predictions when more information is available. With regard to the criterion prediction bias, it can be either positive or negative, depending on the subjective criterion mean and cue expectations. The results suggest that judges make their criterion predictions via a mechanism akin to that described by balancing/cumulating principles. Finally, a new mathematical model describing the judge's prediction behavior is proposed on the basis of the observed results.