Bias in estimates of conditional probabilities and betting behavior as a function of relative frequency and validity of cues in a Cue-Probability Learning task

Abstract

In a cue-probability learning task with two nonmetric cues and events effects of distribution of cue probabilities, P( C j ), and cue validities, P( E i / C j ), on estimates of conditional probabilities and betting behavior were studied. The design was factorial with 3 levels of P( C j ) and 2 levels of both P( E i / C j ). Immediately after the learning session subjects made a bet on the outcome of either of the cues and thereafter estimated some of the probabilities in the task. The hypothesis that the validity of the less frequent cue would be underestimated was confirmed. An opposite effect of overestimating the validity of the more frequent cue was also observed. Parallel results were obtained for betting behavior. There was a general tendency of choosing the more frequent cue irrespective of cue validity. This tendency to disregard valid but rate information merits attention, since it promotes suboptimal decisions.