The Structure of Reasoning: Inferring Conceptual Networks from Short Text

• In Event: That's Just, like, Your Opinion

Abstract

Behavioral models of political engagement typically neglect the structure of human reasoning, assuming instead that opinions represent random samples from some collection of retained information. Yet, scholarship in a number of fields has long indicated that cognitive processes as diverse as reasoning, arguing, remembering, and learning are best modeled as conceptual networks in which connections between similar ideas facilitate the storage and retrieval of relevant information. This structural dimension of reasoning has the potential to significantly influence how an individual samples from and acts on their available beliefs – some people may be prone to constantly return to one central idea, others may jump freely from topic to topic, and others may struggle to see how an issue is relevant to their interests all. This suggests that models of political behavior need to better integrate cognitive models of individual-level reasoning structure. What types of personality traits lead to what types of reasoning structures? How might a tendency towards different structures influence political behavior? Informed by work in political behavior and psychology, this project presents a generative model of individual reasoning in which latent personality traits encourage the activation of different reasoning structures. Using the longitudinal Indianapolis-St. Louis Election Study as well as survey data related to the Affordable Care Act, I demonstrate that individual reasoning structure can be meaningfully inferred from free response text and find that these structures correlate with validated personality and ideology measures. Ultimately, this work presents a collection of archetypes of individual reasoning processes which serve to better inform our understanding of political behavior.

Author

• Sarah Shugars, Northeastern University