Edited by Randy Gallistel, Rutgers University, Piscataway, NJ, and approved August 13, 2013 (received for review January 9, 2013)
Human decision-making seems fundamentally domain dependent. Sensory-motor decisions (e.g., where to put your feet on a rocky ridge) seem near-optimal, whereas decisions based on numerical information (e.g., choosing between financial options) seem suboptimal. Additionally, when people rely on information gained through experience, they make choices that are often the opposite of those they make when relying on described information. However, comparing results across domains on the basis of past results is difficult, because decision-making is studied very differently in different domains. We compared decision-making performance across domains under precisely matched conditions, finding evidence against the idea that fundamental dissociations exist. In fact, peoples’ ability to make decisions seem rather good, although not perfect, in both sensory-motor and cognitive domains.
Classical studies suggest that high-level cognitive decisions (e.g., choosing between financial options) are suboptimal. In contrast, low-level decisions (e.g., choosing where to put your feet on a rocky ridge) appear near-optimal: the perception–cognition gap. Moreover, in classical tasks, people appear to put too much weight on unlikely events. In contrast, when people can learn through experience, they appear to put too little weight on unlikely events: the description–experience gap. We eliminated confounding factors and, contrary to what is commonly believed, found results suggesting that (i) the perception–cognition gap is illusory and due to differences in the way performance is assessed; (ii) the description–experience gap arises from the assumption that objective probabilities match subjective ones; (iii) people’s ability to make decisions is better than the classical literature suggests; and (iv) differences between decision-makers are more important for predicting peoples’ choices than differences between choice tasks.
Author contributions: A.J., U.H., S.K.R., and P.A.W. designed research; A.J. performed research; A.J. and U.H. contributed new analytic tools; A.J. analyzed data; and A.J., U.H., S.K.R., and P.A.W. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300239110/-/DCSupplemental.
