The processing of deviant information in prediction and evaluation

Abstract

Analyses of information integration and of retention were used to examine the processing of deviant information in prediction and evaluation tasks. Sets of test scores were presented serially for a group of hypothetical students, and subjects were asked to evaluate the performance of each student or predict each student's performance on a comprehensive final exam. An averaging model with greater weight for the more recent scores than for the earlier scores was supported for both types of task, but the recency was more pronounced in the prediction task. Weighting of deviant scores differed in the prediction and evaluation tasks. Significant discounting iunderweighting) of deviant scores was obtained only in the prediction task. The ability to recall deviant scores on uncued tests of retention was higher in the prediction task than in the evaluation task. Prediction of future performance based on inconsistent measures of past performance thus appears to be an active process involving the discovery and discounting of unrepresentative information.A common theme in recent reviews of human judgment research is the move from normative theory. with its emphasis on comparisons of actual to optimal behavior, to descriptive models, with concomitant emphasis on psychological interpretations of observed behavior (Anderson, 1974; Kahneman & Tversky, 1973;Slovic, Fischhoff, & Lichtenstein, 1977; Slovic & Lichtenstein, 1971). An interesting question in the formulation of models of human judgment is how deviant pieces of information are processed. The present study uses descriptive models to provide a framework for examining the processing of deviant information in situations involving prediction and evaluation. Kahneman and Tversky (1973) studied "the psychology of prediction" and concluded that people violate the logic of statistical prediction by disregarding factors such as information reliability and prior outcome probability, while relying almost exclusively on the heuristic of "representativeness." According to these authors, people predict by selecting the outcome that is most representative of the input, as if the input information were error free. In their studies, predictions and evaluations tended to coincide. But what if the input information contains inconsistencies that call to question the reliability of the information? The present hypothesis is that the integration of information under such circumstances will differ for predictions and evaluations.The current prediction task is one in which subjects are given a set of test scores for each of a series of hypothetical students and are asked to predict each Requests for reprints should be addressed to Irwin P. Levin,

Information variability and reliability effects in evaluating student performance.

“…In particular, there may be a tendency to uncritically accept numerical scores because of their perceived objectivity. Furthermore, evidence suggests that judges attend more to extreme scores (Levin, Ims, Simpson, & Kim, 1977). Because unreliable sources are more apt to provide such scores, these two tendencies could result in undue weight being given to information provided by unreliable sources.…”

A selective summary of visual averaging research and issues up to 2000

“…Numbers in real-world context Pollard (1984) provides extensive treatment of the contributions of Levin (1974aLevin ( , 1974bLevin ( , 1975 and Levin, Ims, Simpson, and Kim (1977), so only the highlights of those articles are given here. Levin's approach was to use symbolic stimuli (numerals) but to provide a context by instructing the observers to think of the numbers as, for example, price increases, IQ scores, or academic test results (see also Malmi & Samson, 1983).…”

Information integration in price-quality tradeoffs: The effect of missing information

“…Empirically, averaging processes are identified by showing that the effect of a particular stimulus decreases as the number of other dimensions with which it is combined increases (Anderson, 1981). Support has been provided in situations as varied as personalityimpression formation (Anderson, 1965), attitude formation (Anderson, 1971), psychophysical and intuitive statistical judgments (Levin, 1975), clinical assessments (Anderson, 1972), evaluating student performance (Levin, Ims, Simpson, & Kim, 1977), and consumer judgments (Troutman & Shanteau, 1976).…”