Proceedings of the Annual Meeting of the Cognitive Science Society

Bayesian reasoning tasks require processing data in proba-bilistic situations to revise risk estimations. Such tasks are difficult when data is presented in terms of single-event probabilities; the multiplicative combination of priors and likelihoods often is disregarded, resulting in er-roneous strategies such as prior neglect or averaging heu-ristics. Proportions (relative frequencies) are computation-ally equivalent to probabilities. However, proportions are connected to natural mental representations (so-called ra-tio sense). Mental representations of nested proportions (70% of 20%) allow for a mental operation that corre-sponds to a multiplicative combination of percentages. In two studies, we focused on the conceptual understanding underlying Bayesian reasoning by utilizing graphical rep-resentations without numbers (to avoid calculations with percentages). We showed that verbally framing Bayesian tasks in terms of proportions, as opposed to single-event probabilities, increased correct Bayesian judgment, and re-duced averaging heuristics. Thus, we claim, proportions can be regarded as a natural view on normalized Bayesian situations.