Probabilistic information processing: Effects of a biased payoff matrix on choices and bids

Abstract

The purpose of the experiment was to determine how subjects process equivocal information relevant to a choice between two hypotheses when one of the hypotheses is more valuable than the other. Within the context of the theory proposed by Wallsten [Journal of Mathematical Psychology 1976 14 144–185; In Castellan, N.J., Pisoni, D., & Potts, G. R. (Eds.), Cognitive theory (Vol. II) 1977] the two hypotheses should be differentially salient to the subjects, and this should cause biases in the processing of relevant information. Accordingly, on each of several hundred trials over many sessions, six subjects independently observed relevant equivocal information and then chose between two hypotheses which differed in value. Subjects used the Marschak bidding procedure to provide confidence ratings in their decisions. Analysis of the binary choice probabilities showed that four subjects had a choice bias for the more valuable hypothesis, while two subjects demonstrated the reverse bias. Corresponding analyses of the bid data suggested a different bias pattern, which was explainable by considering the subjects' strategies in arriving at the bids. Nevertheless, the ordinal properties of the bid data were investigated in accordance with the models and techniques developed elsewhere by Wallsten 1976, with the results that (i) the algebraic models of information processing described well the data from individual subjects, and (ii) the model parameters derived for each subject demonstrated information-processing biases consistent with the independently observed choice biases. These data in conjunction with the aforementioned processing theory suggest that hypothesis salience can be manipulated in a way that affects information processing and, furthermore, that subjects have choice biases toward more or less risky gambles according to whether the more or less valuable event is more salient to them.