Purchase this article with an account.
C. Shawn Green, Charlie Benson, Daniel Kersten, Schrater Paul; Promoting optimal decision making by reducing unexplained variability in outcome. Journal of Vision 2009;9(8):836. doi: https://doi.org/10.1167/9.8.836.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Human behavior in binary choice tasks is notoriously suboptimal. Given repeated choices between two options, one with a higher probability of being the correct option than the other, the obvious optimal solution is to choose only the higher probability option. Interestingly, this optimal strategy is rarely observed. The more typical finding is that subjects sample the options in proportion to their respective probabilities of being correct - a tendency known as probability matching. While standard models in the field posit that subjects in decision making tasks simply collect outcome statistics and base their decisions upon those statistics, we propose that individuals have a natural propensity to not just simply learn the outcome statistics, but instead attempt to build a causal model that can reduce unexplained variability in outcome. Only when this unexplained outcome variability is sufficiently reduced will behavior approach optimal. We tested this hypothesis by comparing subject performance in various conditions that had identical outcome statistics, but differed in the degree to which they fostered the creation of compelling causal models that could explain the statistics. As predicted, subject behavior was significantly nearer to optimal in the condition where a natural causal model existed than in conditions where no such model could be formed.
This PDF is available to Subscribers Only