Conflict detection across various probabilistic reasoning tasks

Abstract

Many biased reasoners seem sensitive to the conflict between their heuristic, biased responses and logical/probabilistic principles, known as conflict detection (CD). As CD has been found to predict receptiveness to feedback and training, it seems educationally relevant. However, to date, CD has mostly been studied with classic “heuristics-and-biases tasks”, and it is unclear how consistent it is within participants across tasks that share a mindware component. Therefore, we explored whether CD is: (1) found in more complex (Bayesian) reasoning tasks, and (2) consistent across probabilistic reasoning tasks of varying complexity that share a mindware component (i.e. base-rate, inverse-fallacy, and Bayesian-inference-tasks). Results showed that the proportion of conflict detectors decreased with increasing task complexity. Moreover, for the Bayesian-inference-task, CD-presence depended on analysis method (types of errors included). We found little evidence for consistency in CD.