Modeling Judgment Errors in Naturalistic Numerical Estimation

AbstractWe quantitatively modeled and compared two types of errors in numerical estimation for naturalistic judgment targets: mapping errors and knowledge errors. Mapping errors occur when people make mistakes reporting their beliefs about a particular numerical quantity (e.g. by inflating small numbers), whereas knowledge errors occur when people make mistakes using their knowledge about the judgment target to form their beliefs (e.g. by overweighting or underweighting cues). In two studies, involving estimates of the calories of common food items and estimates of infant mortality rates in various countries, we found that knowledge error models predicted participant estimates with very high out-of-sample accuracy rates, significantly outperforming the predictions of mapping error models. The knowledge error models were also able to identify the objects and concepts most associated with incorrect estimates, shedding light on the psychological underpinnings of numerical judgment.


Return to previous page