Pseudodiagnosticity and preference hierarchy in a search-only inference paradigm

Bayes' Theorem provides a rationality-standard for information search when there are two mutually exclusive hypotheses and one or more statistical cues pertaining to the likelihoods of the hypotheses. Prior research shows that when people already have a cue pertaining to a hypothesis and are asked to seek additional information to help decide which hypothesis is correct, they tend to exhibit a specific form of pseudodiagnosticity: Rather than seek information that would assess the same cue relative to an alternative hypothesis, they tend to seek information about how a second cue would pertain to the first hypothesis. For example, if people are told that 70% of genuine paintings are landscapes, they then seek to know the percentage of genuine paintings that are watercolor rather than the percentage of fake paintings that are landscapes. However, this response pattern has sometimes been violated in a way that may depend on the cues' numerical values (e.g., 70% vs. 30%), thus raising a question as to the nature of the bias: Does the selection bias characterize the search process per se, or does it reflect the manner in which people utilize already-obtained percentage information? To address these issues, we employed a novel, search-only judgment paradigm in which people were asked to search for cues and to select them without ever obtaining the cues' percentage values. The results confirmed a tendency toward same-hypothesis pseudodiagnosticity both in primary (i.e., most-preferred) and secondary preference, and supported a model in which pseudodiagnosticity can proceed with or without numerical cue data.