I. The Face of Political Elites

This research agenda examines how the “face” of political elites—understood broadly as visible appearance, affective expression, and embodied signaling—shapes political evaluation, hierarchy, and institutional accountability across regime types.

Across different political contexts, I construct original large-scale multimodal datasets linking visual and emotional cues to elite career outcomes and institutional incentives.

Data Infrastructure

1. Elite Facial Image Dataset (China – Political Selection)
From: Portraits of Power: Facial Appearances and the Tacit Domain of Political Selection in China. (APSR, Conditionally Accepted)

• 4,000+ mid- and senior-level Chinese officials, 2000-2022
• Official portrait photographs collected from government archives
• Linked to detailed career histories, promotion timing, and purge outcomes
• Machine-estimated facial traits (competence, trustworthiness, aggressiveness, attractiveness) using a deep-learning model trained to replicate human evaluations

This dataset enables analysis of how perceived facial traits predict political selection in authoritarian bureaucracies.

2. Elite Affective Expression Dataset (China – Hierarchical Emotional Discipline)
From: The Disciplined Face of Power: Affective Signaling and Hierarchical Rule in China.（Working paper)

• 20,000+ video appearances of Chinese officials
• Frame-level facial affect extraction using deep neural networks
• Measures of emotional calmness, stress signaling, and affective regulation

This dataset examines how emotional discipline functions as an informal signaling mechanism within bureaucratic hierarchies.

3. Legislative Emotional Performance Dataset (South Korea – Electoral Accountability)
From: Spontaneous Outburst or Strategic Signal? Legislative Emotions in South Korea.（Working paper)

• Full universe of plenary sessions (784 sessions; 1,750+ hours) of the Korean National Assembly, 2010–2025
• Automated video-based emotion recognition applied to parliamentary debate recordings
• Integrated with legislative transcripts and legislator-level biographical metadata

This dataset enables analysis of how electoral accountability incentives shape strategic emotional expression in democratic legislative settings.

Visualizations

1. Facial Appearance and Political Selection

From: Portraits of Power: Facial Appearances and the Tacit Domain of Political Selection in China
(American Political Science Review, Conditionally Accepted)

Average Faces by Facial Rating and Gender

The figures below visualize how perceived facial traits correlate with elite political advancement in China. Officials are grouped by percentile rankings of machine-estimated facial traits. Synthetic “average faces” are generated within each percentile group to illustrate systematic perceptual variation associated with career outcomes.

Female Officials

Male Officials

The leftmost and rightmost images represent averages of the highest- and lowest-rated officials for each trait.

Contributions of Facial Regions to Trait Perceptions

The figures below illustrates the contributions of different facial regions to the machine-based ratings of the four facial traits. The facial heat maps are produced by Gradient-weighted Class Activation Mapping (Grad-CAM), a visualization technique for interpreting deep-learning models. Regions highlighted in red indicate significant contributions to the prediction of the trait, while blue regions indicate minimal contributions.

Heat maps are generated using Gradient-weighted Class Activation Mapping (Grad-CAM).

2. Affective Hierarchy and Emotional Discipline

From: The Disciplined Face of Power: Affective Signaling and Hierarchical Rule in China
(Working paper)

Coming soon.

II. LLMs and Social Science Research Methods

This research agenda develops automated frameworks for extracting political facts, coding legal texts, and scaling expert-level annotations using large language models.

Visualizations and datasets forthcoming soon.

III. Political Elites in Global Governance

This research agenda studies elite networks, institutional power, and leadership selection in international organizations and global governance institutions.

Visualizations and datasets forthcoming soon.