Being accurate but not necessarily Bayesian: Comments on Christensen-Szalanski and Beach

Abstract

J. J. J. Christensen-Szalanski and L. R. Beach (Organizational Behavior and Human Performance, 1982, 29, 270–278) presented experimental subjects with a representative sample of cases each having two characteristics: either D or $\text{D}$ and either H or $\text{H}$. Subjects estimated P(H|D) relatively accurately. The authors interpreted this result as indicating that under such conditions subjects use Bayes' theorem properly; specifically, they do take account of the base rate. We disagree with this interpretation and claim instead that the way the information was presented enabled subjects to make a direct estimate of the relative frequency appropriate for P(H|D), the percentage of H in the set of D. Under these circumstances, the question of whether or not people correctly utilize all the component's of Bayes' theorem is completely irrelevant.