Year: 2025
Type: dissertation
Abstract: Predominant theories of public opinion often portray citizens as unskilled reasoners, unable to logically think through political issues and reliant on heuristics, stereotypes, and biases to make minimally informed decisions through a process of directionally motivated cognition. I push back on the notion that bias and laziness in political reasoning makes citizens unskilled reasoners. I flip the motivated nature of political reasoning on its head, arguing that human reasoning likely evolved in an environment that made bias and laziness jointly beneficial. I propose that one such environment is one in which argumentation enabled humans to solve coordination problems and agree on mutually beneficial arrangements in everyday life. The same genre of discussion—argumentation—should therefore encourage similar attitude changes in present-day interactions. I show that Americans are committed to attitude change as a desirable normative good, and that they are willing to attempt to persuade others about politics, especially those with whom they do not share close preexisting relationships. Argumentativeness as a psychological trait is relatively evenly distributed throughout the population, and citizens need not possess high levels of political knowledge to reap the benefits of argumentation. Testing this type of theory across a wide range of interactions requires new computational tools for the study of persuasion in political speech. I leverage recent advances in deep learning to build computational models designed to classify texts on several argumentative characteristics. These deep argument mining models improve upon previous state-of-the-art metrics, offer scholars a quick and inexpensive method of accurately extracting useful information from political text, and show the value of incorporating deep learning approaches into similar models. I then use a series of survey experiments to mimic aspects of argumentation. I show that features common to argumentative exchanges can decrease the strength with which citizens hold their attitudes about political issues, move those attitudes in the direction of an interlocutor’s attitudes, and make attitudes less extreme as a result. I show how the magnitude of these effects partially depends on issue salience and the amount of accessible information available to individuals, and I show that while reductions in attitude extremity are driven by engagement with counterattitudinal viewpoints, reductions in attitude strength are driven by providing reasons supporting one’s own viewpoints. Finally, I examine the effects of argumentation in an online forum. Online environments are not often considered the type of setting where attitude change is likely to occur. However, I show that when attitude change does occur, it is more likely to be in response to comments that employ argumentative characteristics. Moreover, the effectiveness of those argumentative characteristics increases when they are processed by individuals who are more cognitively engaged in the discussion. These results carry implications for the design of interventions to persuade citizens, encourage attitude moderation, and reduce political polarization.
DOI: 10.17615/rbn8-2a48
Author: Mehlhaff, Isaac D.
Institution: University of North Carolina at Chapel Hill
Language: English
Subfield: Philosophy
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7077515996
Cites:
Cited by:
Related to: