Research

My research uses computational methods to study core sociological concepts. First, I am interested in how occupation structure in the U.S. evolves overtime, and how such changes can be mapped to the capitalist class structure. Second, I am interested in using computational text analysis to study how discourse and opinions change over time.

• Social Class and the Change of Occupation Structure [Work in progress]
  • Dissertation project
    • Committee Chair: Vivek Chibber, NYU Sociology
  • For my dissertation, I apply computational text analysis and machine learning to understand how occupation structure in the U.S. have changed overtime. In particular, I examine how occupations have evolved overtime along the two key dimensions of social class (i.e. authority and autonomy) as theorized by Erik Olin Wright.
    
    
    
• The Construction of Economic Subjects in Business-related Self-Help Books [Draft available upon request]
  • Co-author with Carly Knight, NYU Sociology
  • Using a novel dataset of New York Times best seller lists since the 1970s, we analyze how the discourse on work, self, and employment change overtime in business-related self-help books. Specifically, we analyze how economic subjects are constructed overtime in parallel with the rise of neoliberalism.
    
    
    
• Chinese Public Sentiments towards the U.S. and Democracy, 2011 to 2020 [Under review]
  • Co-author with Yinxian Zhang, Queens College Sociology, CUNY
  • This project describes the long-term change in public sentiments towards the U.S. among Chinese internet users using political discussion data collected from Zhihu, the most popular online question-and-answer community in China that resembles Quora. Tracking sentiment change among Chinese netizens from 2011 to 2020, this paper provides a long-term overview of Chinese online sentiments in a rapidly changing decade. In addition, by leveraging measurement of the popularity of posts, this study empirically tests whether there is a discrepancy between prevalent versus popular sentiments.
    
    
    
• Red AI? Variation in GPT Models’ Political Knowl- edge and Sentiment in English and Simplified Chinese [Under review]
  • Co-author with Yinxian Zhang, Queens College Sociology, CUNY
  • This research is one of the first studies that systematically investigate the cross-language political biases and inconsistencies in large language models (LLMs). We found that GPT models trained in different languages have sentiment biases that make them more positive toward their “own country” while more negative toward “other countries.” In addition, we found that China-related political issues have significantly higher rates of inconsistency both in terms of content and sentiment, suggesting that Chinese state censorship and US-China geopolitical tensions may have influenced the performance of the bilingual GPT models. Our study brings public attention to the biases and inconsistencies in multilingual LLMs, which bear profound implications for cross-cultural communications.
  • Preprint on ArXiv
    
    
• The Elements of Cultural Power: Novelty, Emotion, Status, and Cultural Capital (2022, American Sociological Review)
  • 2023 Best Student Research Paper Award, ASA Section on Asia and Asian America
  • 2023 Best Student Research Paper Award (Honorable Mention), ASA Section on Communication, Information Technologies and Media Sociology
  • Why do certain ideas catch on? What makes some ideas more powerful than others? In this article, I examines key predictors of cultural power—novelty, emotion, status, and linguistic features—using an innovative diachronic word-embedding method. The study finds a curvilinear relationship between novelty and resonance, as well as a positive relationship between status and cultural power. Contrary to theoretical expectations, moderate emotions, whether positive or negative, are found to be more effective in evoking resonance than more intense emotions, possibly due to the mediating effect of the forum’s “group style.” The study also finds significant effects of linguistic features, such as lexical diversity and the use of English in Chinese discussions. This suggests a Bourdieusian “cultural capital signaling and selection” path to cultural power, which has not been considered in most studies of resonance.
  • Accepted Version available for download.
    
    
• Child and Youth Well-being in China (2019)
  • Co-authored with Lijun Chen (Chapin Hall at the University of Chicago), Dali Yang (the University of Chicago), and Qiang Ren (Peking University)
  • Using data from the longitudinal Chinese Family Panel Studies (CFPS) survey, this book analyzes the well-being of Chinese children and youth from multiple dimensions. We not only pay attention to the economic, physical, psychological, cognitive, and attitudinal development of children in China, but also analyze how social and institutional context (such as migration and parental absence) affects child development.