Research overview

Social emotional processes across individuals and groups

The bulk of my research asks how people express their emotions and interpret others' emotion expressions, with an eye toward understanding the consequences for well-being, relationships, and society. I'm particularly interested in how these social emotional processes shape and are reshaped by emerging AI systems as they become more integrated into our social lives. Below are three broad streams of research that I am currently pursuing, along with a brief overview of my approach and methods.

The link between emotion expression (or the lack thereof) and individual well-being

For over two decades, suppressing the outward display of emotion has been considered an "unhealthy" emotion regulation strategy, finding links between suppression and various well-being costs, such as increased loneliness and decreased life satisfaction. My research challenges the dominant "all suppression is unhealthy" view, foregrounding the role of emotion valence in how suppression relates to well-being.

This line of works shows cultually generalizable and temporally stable patterns: suppression of positive emotions is associated with lower well-being, while suppression of negative emotions is not (Yu et al., 2023, 2026).

Current research questions:
  • Why does positive emotion suppression link to well-being (e.g., Yu et al., 2025)? Is positive emotion suppression a cause of lower well-being or a consequence of it?
  • What specific emotions do people tend to express or suppress across different contexts and why?
  • How can we measure emotion expression and suppression in naturalistic settings? Do behavioral indicators and self-report measures capture the same underlying processes?

The nuanced social effects of negative emotion expression on social functioning

My first line of research suggests that understanding the social effects of negative emotion expression may require a more contexaulized approach that considers how others interpret the expressed emotion. In this line of work,

Focusing on moral outrage expression in online political discouse, I showed that people's inferences about the expresser's motives critically shapes whether they would be willing to have an open-minded conversation with the expresser, even when they hold opposing political views (Yu & Brady, 2025).

Current research questions:
  • Why does online political discouse become concentrated among a few vocal and emotional actors? Can we make the discouse more represensative of the views of the broader population? (Yu & Brady, in prep.)
  • Do emotions motivate offline forms of political participation beyond its online forms? Is more intense emotion expression always more mobilizing?
  • Do people habituate to a discourse filled with negative emotions, and how does it shape their perceptions of politics, partisans, and social media in general?
  • How are emotions on social media regulated, and who are the key actors and what determines the emotions’ trajectories and impacts on both the community and its members? What are the necessary conditions that allow for negative emotions to be expressed while also maintaining the civility of the discourse?

Social emotional processes in the age of AI

I am interested in extending questions about emotion expression and interpretation beyond human interactions to include AI systems. As AI systems become more deeply embedded in our social lives, they increasly play a role in shaping how we express and interpret emotions. People are now using AI systems for emotion regulation, social support, and companionship, and online social interacitons are largely mediated by AI systems. Understanding how people interact with and through AI systems is therefore critical to ensure that these systems are designed to support human well-being and social functioning.

Towards this end, I am exploring the following research questions:
  • Can AI communicate the distribution and heterogeneity of human beliefs rather than disproportionately surfacing extreme, stereotypical, or otherwise unrepresentative ones (Brady, Yu, et al., 2026)?
    • Can an AI provide flexible emotion regulation or does it homogenize its suggestions on how people should regulate emotions?
    • Can AI present a diversity of political perspectives and does it help reduce pluralistic ignorance and polarization?
  • How does AI represent emotion concepts, and how do these representations shape both the system’s behaviors and users’ short- and long-term well-being?
    • Do people outsource their emotion regulation to AI and become less skilled with regulation themselves or less willing to connect with other people?
    • What are the network-level consequences if emotion expressions are gradually reserved for AI?
  • Can AI successfully simulate humans beyond cogntive, but also emotional reasoing?
  • How should social scientists approach the study of AI and human interaction? What normative questions should we consider? What ethical guardrails should be in place?

Social science and social implications

I am also deeply interested in the social implications of social science research and how it can inform policy and practice. In one project, I analyzed large corpora of U.S. policy documents and social science research to examine the extent to which psychological research is cited in U.S. policy documents. I found that it is increasingly cited, especially for education, health, and social justice issues (Yu et al., 2026). These science-of-science and meta-sceicne research questions are very interesting to me and I would like to continue exploring them in the future.

How I work

Approach and methods

I use quantitative survey methods, longitudinal designs, experiments, and computational analyses of language and behavioral data. My work combines confirmatory tests with exploratory modeling, and I value both predictive performance and explanations that remain legible to theory.

Below is a non-exhaustive list of the methods I have used in my work:
Experiments Social media field study Survey methodology Longitudinal analysis Large-scale archival data analysis Embeddings Topic modeling LLM-based classification and content analysis Network analysis Agent-based modeling