Affective Trajectories under Authoritarian Precarization: A Multi-LLM Analysis of Chinese Workers’ Protest Experiences

• In Event: Labor and Labor Movements Section Refereed Roundtables
  In Refereed Roundtable (60 minute): Roundtable 7: Labor Movements, Protest, and Political Contestation

Abstract

This paper empirically examines the scholarly debate on empowerment versus precarization experienced by China's workers in the past two decades, while exploring the affective dimension of workers' protest experiences and institutional encounters using a multi-LLM ensemble pipeline. Scholars remain divided on whether Chinese workers are experiencing empowerment through demographic shifts, labor shortages, and increased rights consciousness (Zhang, 2017; Lee, 2016), or continued precarization despite nominal legal protections due to weak enforcement and absent independent unions (Lee, 2007; Pun, 2005). While existing research has primarily analyzed collective protests and state responses, individual workers' subjective experiences and emotional trajectories through institutional encounters remain underexplored. Understanding how workers experience and emotionally process repeated institutional failures is crucial for explaining why growing rights consciousness has not translated into sustained collective empowerment.

Using 6,624 individual help-seeking posts and 18,428 collective protest cases from China Labour Bulletin datasets spanning 2013 to 2025, I examine two interrelated dimensions: workers' material encounters with employers and institutions, and their affective experiences as expressed through narrative meaning-making. The findings support the precarization thesis: institutional responses predominantly consist of repression through police deployment, arrests, and violence, or complete neglect and abandonment. These encounters produce systematic emotional trajectories from hope and determination toward fear, desperation, and resignation that undermine workers' capacity for sustained collective resistance.

Methodologically, this study deploys an ensemble of five Large Language Models with theoretically-grounded prompts, demonstrating how computational methods can capture nuanced emotional experiences and meaning-making processes that traditional sentiment analysis and topic modeling cannot access. The multi-model consensus approach provides transparent validation while revealing how workers narratively construct meaning around institutional failures. This research contributes to understanding how authoritarian precarization operates not only through material constraint but through systematic cultivation of affective states that internalize powerlessness and fragment collective consciousness, offering new computational tools for analyzing emotions as social facts at scale.

Author

• Yinan Xu, Boston College