Testing and extending a dual-source model of everyday conditional reasoning
Zusammenfassung der Projektergebnisse
The goal of this project was to evaluate and extend the dual-source model of conditional reasoning (DSM). According to the DSM, a formal, probabilistic measurement model, two different types of information are integrated when drawing probabilistic conditional inferences: a knowledge-based component reflecting background knowledge about the underlying probability distribution and a form-based component reflecting the subjective probability with which an inference schema is seen as warranted. In the first two experiments, we validated the model parameters using selective influence manipulations. Manipulating the argument form was captured by the form-parameters while manipulating the expertise with which a conditional rule was uttered was captured by the mixtureweight for integrating both types of information. In the third and fourth experiment (Experiments 3a and 3b), we examined classical suppression effects (Byrne, 1989) and found a dissociation of the processes underlying the effects of disablers and alternatives: While both types of counterexamples suppressed the influence of the form-based components, disablers also suppressed the credibility of the conditional via decreasing the mixture-weight, whereas alternatives decreased the knowledge-based component. In the fifth experiment (Experiment 4) we investigated the role of workingmemory in probabilistic conditional reasoning by measuring participant's working memory capacity and manipulating working memory load in a between-subjects design. Independent of manipulated load, the form-based parameters were more in line with classical logic for participants high in capacity compared to participants low in capacity. Finally, we performed a meta-analysis on data from seven studies, which showed that the dual-source model provided a better account of the data than three Bayesian competitor models. Overall our results provide strong support for the interpretation that distinct cognitive processes contribute independently to probabilistic conditional reasoning, and they show how the dual-source model can be used to provide new answers to important questions in the psychology of reasoning.
