Zum Einfluss von Seltenheit auf Zusammenhangsurteile - Implikationen für die Entstehung von Vorurteilen und für Bayesianische Modelle menschlicher Urteile
Zusammenfassung der Projektergebnisse
When judging the relation between events, humans are remarkably sensitive to statistical correlations. At the same time, when events can be either present or absent, for example a migraine or a stressful day, the co-occurrence of the events being present has a higher impact on judgments than their joint absence. From a statistical point of view, this is a bias because statistical indices require events to be weighed equally. The present work provides the hitherto most comprehensive meta-analytic summary of correlation judgment studies. This summary shows the pervasiveness of the unequal weighting of specific events. To conceptually clarify which events are weighed more strongly, the distinction between features (being present or absent) and dimensions (always present at varying levels) is introduced and shown to be predictive of the strength of the unequal weighting. The meta-analytic findings and novel experimental evidence suggest that features being present receive higher weight in correlation judgments than features being absent. This differenfial weighting vanishes for dimensions. Finally, a rational explanation of this unequal weighting in terms of Bayesian information integration is not supported. This explanation would have required that a) events being present are perceived as more rare than events being absent and b) that events expected to be rare have a higher impact on judgments than those expected to be frequent. In separate experiments, both of these assumptions remained unsupported. Thus, much of the unequal weighting of events in human correlation judgments remains to be explained.
