Ph.D. Candidate · Political Science & Social Data Analytics · Penn State

YiMin Huang

I study how authoritarian regimes—China in particular—exert influence within democracies, and why some democracies are more resilient than others.

Parliament House of Finland, Helsinki — from my parliament travels

Introduction

About

YiMin Huang
YiMin Huang Department of Political Science
Pennsylvania State University
ym.h@psu.edu

Hello, welcome to my page. I’m a Ph.D. candidate at Penn State, pursuing a dual-title degree in the Department of Political Science and the Center for Social Data Analytics (SoDA). I study how authoritarian powers—China in particular—influence democracies, and why some publics and institutions prove more resistant than others. I work mainly with survey experiments, causal inference, and text and computational methods.

My dissertation, The Authoritarian Playbook, examines three channels of authoritarian influence in democracies: propaganda that reshapes how citizens recognize democracy, skepticism that erodes public trust in security allies, and financial ties that reach into democratic legislatures—with empirical work centered on Taiwan and Australia. A related project carries these questions into generative AI, asking how Chinese open-weight language models handle politically sensitive content.

Prior to Penn State, I received my master’s degree in Political Science from National Taiwan University and a bachelor’s degree in Political Science from National Cheng Kung University.

For more information, you can view my CV or email me.

What I work on

Research

My research asks how authoritarian regimes—China in particular—exert influence within democracies, and why some democracies prove more resilient than others. I build the original data my questions require: population-based survey experiments in Taiwan, two decades of linked donation and parliamentary records in Australia, and large-scale audits of Chinese AI models. Drafts are available upon request.

Dissertation — The Authoritarian Playbook: Propaganda, Skepticism, and Foreign Donation in Democracies

Under reviewJob market paper

Framing Democracy: China’s Propaganda and Its Influence on Taiwanese Democratic Perceptions

Marginal treatment effects on perceived democracy of Taiwan's authoritarian past and democratic present, by authoritarian personality and traditionalism
Treatment effects on perceived democracy of Taiwan’s authoritarian past (top) and democratic present (bottom), strongest among the most predisposed respondents.

Research on authoritarian influence largely asks whether propaganda changes citizens’ attitudes toward particular actors, institutions, or policies. This paper asks whether it can move something more fundamental: the standards citizens use to recognize democracy itself. China’s 2021 white paper, China: Democracy That Works, rebrands the hallmarks of authoritarian rule—order, unity, harmony, and effective governance—as the criteria by which governance should be judged democratic. In a survey experiment in Taiwan built on the white paper’s own text, I find that exposure narrows the perceived distance between Taiwan’s authoritarian past and democratic present, with the largest effects among respondents most predisposed toward authoritarian values.

Penn State Best Political Science Master’s Essay Award · Richard L. Walker Best Paper Award · Supported by the Taiwan Government Fellowship for Studying Abroad (Ministry of Education, Taiwan)

Expected submission 2026

Democratic Education and Resilience to Foreign Propaganda: Evidence from Taiwan’s Curriculum Reform

Regression discontinuity plot for the control group showing a negligible discontinuity at the 2006 reform cutoff
Without propaganda, the discontinuity at the reform cutoff is negligible.

What protects citizens against propaganda? Taiwan’s 2006 curriculum reform brought the history of democratization into national schooling. Using respondents’ birth months in a regression discontinuity design, I find the reform buffers post-reform cohorts against precisely this propaganda: without propaganda the discontinuity is negligible, but under exposure post-reform cohorts maintain their evaluations of democracy—a result that identifies democratic education as a source of democratic resilience.

Companion paper to the job market paper, drawn from the same experiment.

Expected submission 2027

Under the Shifting Sands: Propaganda, Skepticism, and U.S.–Taiwan Relations

Research on alliance credibility has focused on how patrons signal commitment and how allied governments and elites judge those signals—but the mass public of the protected ally remains largely unstudied, even though a commitment like the United States’ to Taiwan must both deter Beijing and reassure Taiwanese society. In a survey experiment, I compare “U.S.-skepticism” narratives about security with narratives about trade: because a trade narrative says nothing about defense, any effect it has on security judgments reveals how doubt spills across domains. Early results show that economic skepticism erodes confidence in U.S. security commitments more than military skepticism does—a finding with direct implications for how democracies defend against narratives that never mention defense.

Expected submission 2027

Silent Conduits: Ties, Donations, and China’s Growing Influence in Australia

We know a great deal about how domestic interest groups buy access in democratic legislatures; we know far less about whether foreign authoritarian donors do the same. Assembling two decades of Australian campaign-finance and parliamentary records, I find that China-linked donations cluster significantly around the foreign-affairs committee network, while timing tests show the money does not chase individual legislators’ committee assignments—a pattern consistent with cultivating durable access rather than transactional exchange. The chapter’s contribution is the idea of democratic mimicry: foreign authoritarian actors replicating the strategies of domestic influence-peddlers inside a democracy’s own open institutions.

Generative AI and Authoritarian Information Control

Authoritarian censorship has always run through infrastructure the state controls—newspapers, platforms, firewalls. This research program studies what happens when it moves inside a product anyone can download: Chinese open-weight language models now anchor local AI deployment worldwide, and because they run offline, any censorship they carry from China must be built into the model itself.

Working draft · submission by December 2026Taiwanese Overseas Pioneers Grant

Strategic Distraction by Other Means: How Chinese Open-Weight Large Language Models Manage Political Risk

Boxplots of the ratio of English to Chinese answer length on sensitive topics, for Chinese-trained and non-Chinese models
English vs. Chinese answer length on sensitive topics, across Chinese-trained and non-Chinese models.

Running twelve Chinese and non-Chinese models locally for over 50,000 replicable responses on sensitive topics, I find that Chinese models rarely refuse: they reframe, recasting politically sensitive questions in safer registers at seven times the rate of non-Chinese models. The strategic distraction that previous scholars documented on Chinese social media now operates from inside model weights.

Working paper

Disappearing the Fallen: The Mechanisms of Authoritarian Information Control inside Open-Weight Language Models

Small-multiple charts showing whether each Qwen model generation acknowledges a purged official, asked in Chinese versus English
Acknowledgment of purged officials across five Qwen generations, asked in Chinese (zh) vs. English (en).

Following the record of China’s purged elites across five successive generations of the Qwen family, I find the models withholding facts they demonstrably possess: later generations conceal officials’ downfalls that earlier generations had inadvertently disclosed, and questions asked in English open a further breach in the wall. Having ruled out insufficient model capacity, I find the erasure is selective, not ignorant. The paper develops an audit design that separates what a model cannot answer from what it chooses not to answer.

Other Ongoing Projects

Pre-analysis plan in progress

How Foreign Disinformation Strategies Shape Democracy Attitudes and Polarization

with Wonjun Song (Hanyang University) and Joseph Wright (Penn State)

A cross-national study of how Chinese Communist Party campaigns promoting “Chinese-style democracy” shape public opinion and partisan polarization in East Asia. Conjoint designs ask citizens to trade off competing democratic attributes.

C-SoDA Research Accelerator Award

Improving Respondent Attentiveness in Online Surveys Through Design Interventions

with Yang Yang (Penn State)

Studies survey attentiveness as a behavior that questionnaire design can improve rather than a fixed trait to screen out, testing visual emphasis and real-time corrective feedback in a factorial experiment.

Expected submission 2027

Legislators in Taiwan’s Popular Committees

How do electoral incentives in Taiwan’s mixed-member system shape competition over legislative committee assignments? A congressional counterpart to the committee politics of my Australia chapter.

In print

Publications

Off the clock

Parliaments

A few photographs from parliaments and capitols I have visited.

World map highlighting countries where I have visited the parliament
Parliament visits so far (updated September 2025).