Effect of Feedback on Base-Rate Utilization

The effect of direct training with feedback on a base-rate problem of the engineers-lawyers type is assessed. Analysis indicates that feedback leads to adjusted probability estimates closer to the Bayesian norm than those in the no-feedback training-only condition. For a second base-rate problem (the so called “divorce problem”), however, no effect of training was shown, even though thinking-aloud protocols for this problem showed that subjects mentioned base-rates more often in the feedback than in the no-treatment condition.