Natural frequencies and Bayesian reasoning: the impact of formal education and problem context

Abstract

Research suggests that children and adults are able to solve Bayesian problems when natural frequencies are used, but most studies have involved only students or academics. The present research focuses on random samples from the general population. Experiments 1 and 2 showed that the natural-frequency format resulted in more Bayesian responses than the probability format. Results of Experiment 3 showed that when the task is not framed as a medical problem and the numbers are not too difficult, about 40% of the general population could correctly solve the problems. More importantly, the natural-frequency format resulted in a substantial increase in correct answers in the lower level of formal education group. Experiment 4 suggests that the context of a problem influences people’s ability to solve a Bayesian problem. Participants were twice as likely to solve a social problem compared with a medical problem. The results of the four experiments show that natural frequency resulted in much better estimates than probabilities. It should be emphasized, however, that even when natural frequencies were used, a majority of participants were not able to solve the problems correctly.