Problem-Structure and Format in Training Conditional and Cumulative Risk Judgments

We describe an experiment evaluating a training program designed to help participants interpret probability and risk information presented in a normalized single-event probability format. Both conditional probability and cumulative probability judgments benefited from training in using tree-structures to represent the problem and the set-subset relations involved. The effects of this training persisted over time regardless of whether the training required them to translate the single-event probabilities into “natural” (non-normalized) frequencies or remain with a probability format. However, training did not generalize and was specific to the problem-type tested. A training program for coping with single-event probabilities is therefore feasible for specific problem (or risk judgment) types but may create difficulties when alternative risk or probability judgments are encountered.