Scholarship

Recursive Reflection Theory proposes that political belief revision occurs not because people simply encounter new information, but because they recursively evaluate the explanations supporting their own beliefs. Belief change begins when individuals conclude that their existing explanatory framework is no longer an acceptable account of why they hold the position they do. To investigate this process, I combine experimental political discussion, AI-mediated conversations, and transcript-based measurement to observe explanatory frameworks as they reorganize in real time. Together, these studies suggest that interventions traditionally understood as persuasion may instead work by prompting citizens to re-evaluate the adequacy of their own explanations rather than by simply transmitting new political beliefs.

Political beliefs are typically studied as individual attitudes or ideological positions. I argue that they are better understood as parts of broader explanatory frameworks that organize values, principles, empirical assumptions, and experiences into coherent understandings of the world. Using transcripts from a large-scale conversational study, I develop a framework for identifying commitments, tensions, reconciliation strategies, and explanatory gaps as they emerge during political discussion. The paper provides the first descriptive account of how people maintain coherence among competing commitments, establishing a new foundation for studying political reasoning through explanatory frameworks.

To what extent does ideological constraint emerge through political conversation? Existing research typically treats ideological constraint as a stable property of individuals' belief systems. Using more than 800 transcripts from experimentally moderated political discussions, this paper examines whether the apparent coherence of political beliefs depends partly on the conversational environments in which they are expressed. Participants discussed police reform, affirmative action, abortion, and climate change with confederates whose partisan identity and ideological consistency were experimentally varied. While participants' policy positions shifted primarily in response to the arguments they encountered rather than the confederate's partisan identity, the apparent constraint among their expressed attitudes also changed systematically. Correlations across political positions were stronger following conversations with ideologically orthodox Democrats and Republicans than after discussions with counter-typical partisans. These findings suggest that ideological constraint is not solely an internal cognitive property but can emerge through interaction with others. More broadly, the study demonstrates the value of combining experimental interventions with conversational data to study the organization of political belief systems.

This project investigates when conventional survey responses accurately capture the reasoning people express during conversation. It contributes to the broader research agenda by examining the relationship between static attitude measures and the explanatory frameworks revealed through dialogue.

Human beings do not simply accumulate beliefs. They organize beliefs, values, experiences, and identities into increasingly encompassing explanations capable of making sense of themselves and the world. This book develops a general theory of these explanatory frameworks. It argues that explanatory frameworks constitute a higher-order cognitive architecture through which people organize experience, generate beliefs, and reorganize their understanding over time. Rather than treating learning as the acquisition of isolated facts or attitudes, the theory reconceptualizes learning as the growth of increasingly encompassing explanatory frameworks. Drawing together evidence from political reasoning, cognitive science, philosophy, education, and artificial intelligence, the book proposes explanatory frameworks as a fundamental object of study for understanding human judgment, learning, and belief revision across domains.

Facing What We Believe argues that political beliefs change not simply because people encounter new information or persuasive arguments, but because they come to recognize that their existing explanations no longer adequately account for what they believe. These moments expose tensions within the explanatory frameworks through which people organize beliefs, values, identities, and experiences, making deeper revision possible. Drawing on original experiments, AI-mediated conversations, and transcript-based analyses of political discussion, the book develops Recursive Reflection Theory as an account of how people evaluate and reorganize their own explanations. It argues that meaningful belief revision depends less on persuasion than on the conditions that prompt people to reconsider the adequacy of the frameworks through which they understand the political world—and explains why those conditions are increasingly rare in contemporary politics.

Human beings do not merely accumulate beliefs. They organize beliefs, values, identities, and experiences into explanatory structures that allow them to make sense of themselves and the world. Although existing research has developed powerful theories of belief formation, schema, narrative identity, cognitive dissonance, and predictive processing, these approaches generally treat beliefs, narratives, or representations as the primary cognitive objects of interest. I argue that these phenomena are more fruitfully understood as components of higher-order explanatory frameworks: cognitive architectures that integrate diverse forms of information into coherent understandings and that are periodically reorganized as new experiences accumulate. The article develops a general theory of explanatory frameworks, identifies their defining properties, distinguishes them from related concepts, and advances a set of theoretical propositions concerning their organization, growth, and revision. Finally, I argue that recent advances in conversational elicitation make explanatory frameworks empirically observable, opening the possibility of studying them directly rather than inferring them indirectly from isolated beliefs or attitudes. Reconceptualizing human reasoning around explanatory frameworks offers a new foundation for research on learning, judgment, belief revision, and political reasoning.

How do people maintain coherence among political beliefs that do not naturally fit together? Existing research has devoted considerable attention to why individuals hold particular political attitudes and when those attitudes change, yet we know remarkably little about the organization of the explanatory frameworks that hold those attitudes together. Using transcripts from a large-scale conversational study, this paper describes recurring patterns in the structure of political reasoning by examining how participants connect principles, values, empirical assumptions, identities, and lived experience into coherent explanations. Rather than revealing a single form of coherence, the analyses identify multiple organizational patterns through which explanatory frameworks accommodate competing commitments, absorb new information, or leave tensions unresolved. These findings suggest that political disagreement is often rooted not in isolated differences over individual beliefs, but in differences in how explanatory frameworks organize and prioritize those beliefs. By providing the first descriptive account of the internal organization of explanatory frameworks, the paper establishes a foundation for studying political reasoning as a dynamic process of explanatory integration rather than a collection of independent attitudes.

How do social norms surrounding racial attitudes differ among white Democrats and Republicans? Do these norms differ from the prevailing ‘norm of equality,’ and how do they affect civic discourse? Using a novel video-based discussion experiment, I show how co-partisan discussion among white Republicans (versus talking to a white Democrat) increases their racial resentment, as measured both in post-treatment surveys and the rhetoric subjects use during discussion. White Democrats do not significantly change their racial attitudes or rhetoric according to the partisan identity of their discussion partner. Among white Republicans, racially resentful language increases by more than 10% when discussing politics with a co-partisan compared to conversing with a white Democrat. This result is driven by the partisan identity of the discussion partner rather than the content of their speech. Even when a co-partisan takes a position on race-related policy that is counter-typical to their party (such as a Republican supporting affirmative action), white Republicans still use significantly more racially resentful language than when they talk to a Democrat, even one who ‘sounds like’ a Republican on race-policy. Findings highlight polarization as a democratic social ill, suggesting consequences for American civic discourse around race.

Overcoming America’s deep partisan polarization poses a unique challenge: Americans must be able to sharply disagree on who should govern while agreeing on more fundamental democratic principles. We study one model of depolarization—reciprocal group reflection—inspired by marital counseling and implemented by a non-governmental organization, Braver Angels. We randomly assigned undergraduates at four universities either to participate in a Braver Angels workshop or simply to complete three rounds of surveys. The workshops substantially reduced polarization according to explicit and implicit measures. They also increased participants’ willingness to donate to programs aimed at depolarizing political conversations. These effects are consistent across partisan groups, though some dissipate over time. Using qualitative data, and building on contact and deliberative theories, we argue that depolarization is especially effective when it includes both informational and emotional components, such that citizens who are moved to empathize with outgroup members become more likely to internalize new information about them.

In 2020, support for Joe Biden among Latina/o/x voters was 8 percentage points lower than support for Hillary Clinton in 2016, the largest drop of any racial/ethnic group. While much media and academic attention has focused on understanding the impact of misinformation, COVID-19 concerns, and racial animus on Latino voters in 2020, we take a step back and clarify the demographic and core ideological characteristics of Latino voters who voted for Donald Trump in 2020. Using a mix of national survey data, precinct returns, and voter file records, we disaggregate components of electoral change. We find evidence of an increasing alignment between Latinos’ ideology, issue positions, and vote choice. Correspondingly, we observe significant pro-Trump shifts among working-class Latinos and modest evidence of a pro-Trump shift among Latinos closer to the immigration experience. These findings, coupled with an analysis of the 2022 CES, point to a more durable Republican shift than currently assumed.

Addressing the question of how to tackle gender inequalities, we test whether women who perceive women’s representation in politics as achievement are empowered. We embed an experiment in an online survey conducted on a sample of 4,545 U.S. women, varying whether the number of women in Congress is presented as achievement or reason for concern. We find that women who are presented with the ‘achievement’ frame are more likely to consider job opportunities for a range of jobs. Heterogeneity based on pre-treatment moderators, particularly political partisanship, which correlates with a host of gender attitudes like feelings of marginalization, helps explain variation in treatment effects. The study broadens our understanding of how descriptive representation matters for non-political outcomes, showing that it empowers women to consider applying to more jobs.

This study uses a novel experimental approach to isolate ‘expressive’ or ‘noninstrumental’ payoffs to voting along identity lines, separating them from ‘substantive’ motivations. Applying social identity theory to the case of gender in U.S. elections, the study answers the question: Do some men and women receive a purely expressive payoff from preferring same-gender candidates? A series of experiments test whether a purely expressive payoff from voting along gender lines is stronger among certain voters. Employing a self-affirmation treatment and measures of group-identity attachment, as well as a voting vignette, the evidence shows that Republican men receive a purely expressive payoff from voting for male over female candidates. That is, Republican men’s preference for a male over a female candidate can be reduced by a self-affirmation treatment that has subjects focus on their individuating characteristics, thereby temporarily ‘detaching’ them from their group attachments. Women and Democratic men are not affected by the self-affirmation treatment. This study deepens our understanding of voter gender bias in the United States, with a particular focus on the implications of ingroup bias among certain partisan voters for the supply of female candidates within party primary elections.

The rising number of women in Congress changes deliberation. Using committee hearing transcripts from 1995 to 2017, we analyze how the gender composition of committees affects group dynamics in committee hearings. While we find limited evidence that increasing proportions of women affects women’s participation, we find that discussion norms within committees change significantly in the presence of more women. Namely, interruptions decrease when there are more women on the committee; with higher proportions of women, men are less likely to interrupt others. Furthermore, committee members are more likely to engage and stay on the same topics in the presence of more women, suggesting a shift in norms toward more in-depth exchange. Overall, our results show that increasing the proportion of women changes discussion dynamics within Congress by shifting norms away from interruptions and one-sided talk in committees, thereby shifting group norms that govern decision-making during an important policy-making stage.

The claim that partisanship has developed into a social identity is one of the dominant explanations for the current rising levels of affective polarization among the U.S. electorate. We provide evidence that partisanship functions as a social identity, but that the salience of partisan identity—in and of itself—does not account for increased affective polarization. Using a two-wave panel survey capturing natural variation in the salience of politics, we find that partisanship contributes more to individuals’ self-concept in times of heightened political salience. We also show that partisans can be detached from their Democratic or Republican identity by having them focus on individuating characteristics through a self-affirmation treatment. However, we find only limited evidence that when partisan social identity is made less salient, either by way of natural variation in political context or through a self-affirmation treatment, partisans are any less inclined to express in-party favoritism and out-party hostility. Taken together, our evidence shows that partisanship does operate as an important social identity, but that affective polarization is likely attributable to more than the classic in-group versus out-group distinction.

Despite widespread belief in the efficacy of statutes prohibiting discrimination in public accommodations, including protections for the use of privately provided yet publicly available services such as transportation, hotels, and restaurants, we lack causal estimates of the impacts of these statutes on the well-being of those they are designed to protect. We leverage the U.S. Supreme Court’s 1883 strike of the public accommodations provisions in the Civil Rights Act of 1875, along with variation in state-level statutes, to identify the impact of the Act’s public accommodations provisions. Using a panel of repeated geo-located medical exams of U.S. Colored Troops and white Union Army veterans, and a series of difference-in-differences, geographic regression discontinuity, and placebo designs, estimates consistently suggest that the Court’s ruling led to meaningful weight losses for U.S. Colored Troops veterans in states without state-level public accommodations statutes. These findings suggest that statutes prohibiting discrimination in public accommodations can have significant positive impacts on the well-being of those they are designed to protect.

The relationship between participation in revolt and individuals’ economic conditions is among the most debated in political science. While conventional economic theory suggests that those who face the poorest economic prospects are most inclined to fight, extant evidence is decidedly mixed. We address this puzzling variation by analyzing the interplay between macro-structural conditions and individuals’ micro-level circumstances. Under conditions of severe group repression, we show how a ‘glass-ceiling’ logic may operate: among the repressed group, those with relatively high productive potential may be most motivated to revolt. We test this with in-depth analysis of participation in the 1993–2003 Burundian insurgency. The data are consistent with numerous implications of the glass-ceiling logic and inconsistent with extant alternative explanations.

Existing theory suggests that individuals who share their identity with elected officials should have more positive attitudes toward government. However, the effect has not been well identified among African Americans, and the hypothesis is underexplored with respect to women. Thus, I examine the effects of Barack Obama and Hillary Clinton’s successes on African American and female political efficacy, using Iowa Electronic Markets contract prices during the 2008 primary to proxy daily shifts in perceptions of Obama and Clinton’s likelihood of winning. I combine this with daily political efficacy measures from the 2008 National Annenberg Election Study. African American efficacy increases with Obama’s perceived probability of success, while white Democrats who prefer Obama are unaffected, suggesting that the effect among African Americans is driven by race and not the success of a preferred candidate. Clinton’s price is not correlated with women’s efficacy, suggesting that descriptive representation’s effects are less pronounced among women.