Measuring the belief system of a person.

Abstract

Theories of belief system structure and dynamics assume that belief systems are a person-level construct. However, measures of belief system structure do not measure the structure of person-level belief systems and instead measure aggregated belief system structure (e.g., the belief system in a particular country). In this paper, I show that a measure of conceptual similarity between attitudes and identities of a belief system works as a valid, reliable, flexible, and efficient measure of person-level belief system structure in the United States. In Studies 1 (N = 387), 2 (N = 389), and 3 (N = 598), I show conceptual similarity judgments are reliable and are related to measures of political engagement, political knowledge, attitude consistency, and preference congruence as predicted by computational models of belief system dynamics. In Studies 4 (N = 981) and 5 (N = 983), I show that conceptual similarity judgments are affected by partisan frames and that changes in conceptual similarity judgments are associated with attitude change as predicted by computational models of belief system dynamics. Conceptual similarity judgments can be used with a variety of attitudes and identities in easy to administer studies. It provides a tool to fill an empirical gap identified by theories of belief system dynamics.

The structure of mass political belief systems: A network approach to understanding the Left-Right spectrum

“…Our comparisons of people according to their self-classification of ideology, which are methodologically similar to previous comparisons between high and low information voters Fishman & Davis, 2022), assume that the three networks are representative of the belief networks of the individuals within those groups. Other techniques, for example comparisons of the way that individuals within each group rate the conceptual similarity between target political constructs (Brandt, 2022), may have utility in future studies designed to verify the observed differences at the level of individuals. Hence, for example, we would expect people who self-classify on the Left or Right to rate policies as more conceptually similar than those who self-classify as belonging to the Center.…”

Inter‐attitude centrality does not appear to reduce persuasion for political attitudes

“…First, although the first experiments used an accepted method for estimating belief system structure, the method estimates the belief system at the sample level and may not appropriately identify the most central and peripheral nodes for each participant (Brandt & Morgan, 2022). Therefore, in this study, belief systems of individuals were estimated using a conceptual similarity task, which has been shown to be a valid measure of a person's belief system (Brandt, 2022; see also Judd et al., 1991). This allows us to identify the most central and peripheral attitudes in each participant's belief system at the pretest and assign them to central and peripheral conditions that are tailored to each participant's belief system structure.…”

“…That said, as we detail in our supplemental materials, we do not have much evidence that the manipulations affected the moral conviction of the attitudes in our study. One additional way to try and capture this idea would be to test if moral reframing messages create new connections in the belief system (e.g., with the relevant moral values) using the conceptual similarity task as an outcome variable (e.g., Brandt, 2022).…”

“…Second, to estimate inter‐attitudinal centrality it is necessary to study multiple attitudes and examine how they are connected to one another. To address this challenge, we leverage network methods to estimate both sample level (Experiments 1 and 2; Brandt et al., 2019) and participant level (Experiment 3; Brandt, 2022) belief systems and then identify the most and least central attitudes in those systems.…”

“…We then randomly assign participants to be persuaded by a moral reframe targeting the central attitude, the peripheral attitude or a control condition (no moral reframe). In Experiment 3, we use a new technique (Brandt, 2022) to identify the most central and peripheral attitudes unique for each participant. We then randomly assign the liberal participants to be persuaded by a moral reframe targeting their own central attitude, peripheral attitude, or a control condition (no moral reframe).…”