Search and evaluation strategies in belief revision: psychological mechanisms and normative deviations

The procedures people adopt in order to seek out and use information have been the focus of empirical research since long in psychology, especially so from the late 50s. This dissertation addresses some key questions left unanswered by a series of seminal studies which date back to the 80s and the 90s on information-gathering and information-use strategies. We first dealt with the question-asking preferences that people exhibit in an abstract task of hypothesis testing. Specifically, we pitted against one another the tendencies to ask positive questions, wherein the confirming answer is expected given the truth of the working hypothesis, and to pose asymmetric queries, wherein the anticipated outcomes of a dichotomous question (i.e., “yes” and “no” answers) convey different amounts of information. Finally, we investigated whether or not people prefer either asymmetrically confirming queries (i.e., questions for which the confirming answer weights more than the disconfirming answer) or asymmetrically disconfirming queries (i.e., questions for which the disconfirming answer conveys more information than the confirming answer). We found a robust tendency to ask positive testing, in keeping with the literature, but neither a preference for asymmetric questions, nor a predominant use of symmetric testing. Furthermore, we showed, correlationally, that people are sensitive to the diagnosticity of questions, as some previous studies in the literature pointed out. Finally, it emerged an interaction between the positivity of questions and the confirming valence of asymmetric queries. A close analysis of the latter finding allowed us to undermine the possibility that people would try to maximize the probability of occurrence of the tested feature, while suggesting a less sophisticated strategy based on the consideration of an easily accessible feature, that is, the probability of a feature under the working hypothesis. After further deepening the study of strategies adopted in the testing phase of hypothesis development, we turned to the evaluation stage. Specifically, we addressed the finding emerged in previous studies of the relative insensitivity of people to the different diagnosticity conveyed by different answers (i.e., “yes” and “no”) to the same question in an abstract task. We showed that not only people might exhibit insensitivity but also oversensitivity to differentially informative answers, indicating a more general failure in information use than previously thought. We also addressed the issue of why people are either insensitive or oversensitive to answer diagnosticity. We provided evidence that an explanation based on the use of the feature-difference heuristic, which has been proposed previously in the literature and wherein people’s estimates are influenced by the difference between the likelihoods, seems unable to explain people’s behavior. By contrast, we found that people prefer to rely on an averaging strategies, in particular on the average between the prior probability and the likelihood. Finally, we investigated an aspect emerged but not directly investigated by previous studies on hypothesis evaluation, that is the feature-positive effect, wherein people tend to overestimate the presence of a feature as opposed to its absence. The results of three experiments with abstract tasks strongly confirmed that the hypothesized effect influences both frequency and accuracy of participants’ responses. We also found that participants exhibited some sensitivity to the formal amount of information, although only with respect to the present clues. Overall, the series of experiments presented in this dissertation contributes to better clarify how people search for information, by showing that they might rely both on formally relevant and formally irrelevant properties of the information they have at hand and by putting into question the alleged tendency to hypothesis confirmation, defined as a maximization of the probability of a confirming datum. Furthermore, these experiments help understand how people treat information, by specifying how people misweigh differentially diagnostic answers and showing that a psychologically compelling tendency, namely the feature-positive effect, might, at least in part, account for people’s information use.