A Bayesian analysis of subjective probabilities of interpersonal relationships

Abstract

The present study investigated Ss' perceptions of the likelihood of several interpersonal relationships in hypothetical two- and three-person groups. Ss were asked to judge the probability that two people were joined by a specific interpersonal relationship, given various amounts and kinds of prior information regarding the incidence of that relationship in the group. Data were analyzed by assuming that the Bayesian likelihood ratio, computed from initial probability levels and amount of change due to the information, measures the impact of the information on the S. The likelihood ratios provided fairly strong verification for the predictions of balance theory, but they also suggested that other cognitive biases also function in Ss' processing of information about social relationships. It is suggested that the likelihood ratio provides an appropriate and convenient measure of the ‘informativeness’ of a stimulus in any situation in which probabilities can be estimated, and that the consideration of humans as Bayesian processors of social information may have important theoretical and methodological implications for research in the area of belief change.