TKTaegyoon Kim

Taegyoon Kim

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Taegyoon Kim has a total of 28 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Epistemic Intolerance Toward Outpartisans

    by Seo-Young Silvia Kim, Taegyoon Kim, Jeong Hyun Kim, Youngdeuk Park

    We identify and operationalize epistemic intolerance, the belief that outgroup members are epistemically incompetent, violate epistemic norms, and are unworthy of epistemic engagement. In particular, we measure epistemic intolerance toward outpartisans using an original 18-item battery fielded in two U.S. surveys (December 2025 and April 2026; N = 3,386, of whom 2,525 partisans are analyzed). We show that epistemic intolerance is widespread and distinct from affective polarization and that, when measured with the full battery, it predicts democratic norm erosion, outparty dehumanization, and anti-intellectualism toward outparty elites, even when benchmarked against both affective polarization and political sectarianism. Moreover, its overall level is bipartisan, and prevalent even among leaners across both the Democratic and Republican parties. These findings identify epistemic intolerance as a consequential and previously underappreciated dimension of partisan hostility, with downstream implications for designing effective depolarization strategies.

  • Elected officials’ online anti-vaccination statements respond to online engagement

    by Taegyoon Kim, Xinyu Wang, Sarah Rajtmajer, Jeffrey J. Harden, Frederick J. Boehmke, Bruce A. Desmarais

  • Open Access

    Elected officials' online anti-vaccination statements respond to online engagement

    by Taegyoon Kim, Bruce A. Desmarais, Xinyu Wang, Jeff Harden, Frederick J. Boehmke, Sarah Rajtmajer

    State legislators play a central role in U.S. vaccination policymaking, yet little is known about how social media engagement dynamics reinforce elected officials' online anti-vaccine messaging. We examine whether engagement with anti-vaccine posts is associated with legislators' subsequent production of anti-vaccine content. Using a large-scale digital trace dataset of tweets from 2,619 state legislators in 2021, we identify 83,130 vaccine-related tweets and 3,314 anti-vaccine tweets via transformer-based text classification. We model anti-vaccine posting at the legislator-month level with an integrated set of individual, state, and temporal predictors, including partisan time trends and state pandemic conditions. We find that greater engagement with prior anti-vaccine tweets is associated with a higher probability of posting anti-vaccine content in subsequent months. We further show that replies tend to mirror the stance of legislators' vaccine-related tweets, suggesting a reply environment that can amplify perceived audience agreement and sustain continued anti-vaccine communication. Also, anti-vaccine tweets are more likely than other vaccine-related tweets to link to low-credibility sources, indicating that engagement-driven reinforcement may also facilitate the circulation of misleading information. Together, these findings illustrate how the engagement architectures of social media platforms shape elite political communication and the informational politics of public-health governance.

  • Open Access

    A benchmark dataset for evaluating gender sensitivity in Korean political discourse with large language models

    by Sunkyoung Park, Eunbi Cho, Chan Young Jung, Woo Chang Kang, Taegyoon Kim, Eunah Park, Sanghoun Song

    Large language models are increasingly applied to political discourse, but their ability to detect culturally grounded gender sensitivity remains underexplored. We introduce KOGENT, a benchmark dataset of 1,222 transcripts from the Korean National Assembly, annotated for gender sensitivity across 6,024 utterances. Each utterance is labeled as high or low in gender sensitivity, based on contextual indicators of bias, discrimination, or inclusion, and tagged for the target group. KOGENT spans Korean legislative sessions from 1948 to 2024. Annotation reliability was ensured through dual coding and adjudication, yielding high intercoder agreement. When tasked with labeling utterances by gender sensitivity, GPT-4.1 achieved F1-scores of 87.5% (zero-shot) and 91.2% (18-shot), while GPT-4o reached 90.4% and 91.1%, respectively. While incorporating in-domain examples enhanced model performance, limitations in distinguishing between criticisms and reinforcements of inequality, culturally specific terminology, and extended contexts were observed for both models. Our results demonstrate KOGENT’s utility as a robust benchmark for analyzing gender sensitivity in Korean political speech and evaluating multilingual LLMs’ sociocultural alignment.


  • Open Access

    Is External Information Useful for Stance Detection with LLMs?

    by Quang Minh Nguyen, Taegyoon Kim

    In the stance detection task, a text is classified as either favorable, opposing, or neutral towards a target. Prior work suggests that the use of external information, e.g., excerpts from Wikipedia, improves stance detection performance. However, whether or not such information can benefit large language models (LLMs) remains an unanswered question, despite their wide adoption in many reasoning tasks. In this study, we conduct a systematic evaluation on how Wikipedia and web search external information can affect stance detection across eight LLMs and in three datasets with 12 targets. Surprisingly, we find that such information degrades performance in most cases, with macro F1 scores dropping by up to 27.9\%. We explain this through experiments showing LLMs' tendency to align their predictions with the stance and sentiment of the provided information rather than the ground truth stance of the given text. We also find that performance degradation persists with chain-of-thought prompting, while fine-tuning mitigates but does not fully eliminate it. Our findings, in contrast to previous literature on BERT-based systems which suggests that external information enhances performance, highlight the risks of information biases in LLM-based stance classifiers. Code is available at https://github.com/ngqm/acl2025-stance-detection.

  • Open Access

    The Effects of Partisan Elites’ Violent Rhetoric on Support for Political Violence

    by Taegyoon Kim

    Violent partisan hostility is a growing concern for American democracy, yet the role of partisan elites in driving support for political violence remains underexplored. This is especially problematic given their significant influence over public opinion. How does partisan elites’ rhetoric threatening violence affect support for violence against opposing partisans? Focusing on social media, where violent messages by partisan elites from both sides can reach a wide audience with potentially devastating effects, I explore how such messages influence support for violence against opposing partisans. Drawing insights from theories of partisan opinion leadership and inter-group conflict, I conduct a two-by-two (violent rhetoric vs. non-violent rhetoric; co-party elites vs. opposing party elites), pre-registered survey experiment on a sample reflecting the national population. I demonstrate that co-party (but not opposing party) elites’ violent rhetoric increases support for political violence and that partisans fail to counteract elites’ violent rhetoric. Furthermore, I show that fear mediates the impact of violent rhetoric, highlighting its role in fostering support for political violence through fearmongering. This paper is among the first to examine the effects of partisan elites’ violent rhetoric on support for political violence, advancing our knowledge in partisanship, emotion and politics, and political communication.


  • Open Access

    Whose Palestine Is It? A Topic Modelling Approach to National Framing in Academic Research

    by Maida Aizaz, Taegyoon Kim, Lanu Kim

    Maida Aizaz, Taegyoon Kim, Lanu Kim. Proceedings of the 9th Widening NLP Workshop. 2025.

  • Open Access

    Is External Information Useful for Stance Detection with LLMs?

    by Quang Minh Nguyen, Taegyoon Kim

    Quang Minh Nguyen, Taegyoon Kim. Findings of the Association for Computational Linguistics: ACL 2025. 2025.


  • Open Access

    The national network of US state legislators on Twitter

    by Ishita Gopal, Taegyoon Kim, Nitheesha Nakka, Frederick J. Boehmke, Jeffrey J. Harden, Bruce A. Desmarais

    Networks among legislators shape politics and policymaking within legislative institutions. In past work on legislative networks, the ties between legislators have been defined on those who serve in the same legislature or chamber. Online information networks, which have been found to play important roles in legislative communication at the national level, are not bounded by individual legislative bodies. We collect original data for over four thousand US state legislators and study patterns of connection among them on Twitter. We look at three types of Twitter networks—follower, retweets, and mentions. We describe these networks and estimate the relationships between ties and salient attributes of legislators. We find that networks are organized largely along geographic and partisan lines and that identity attributes—namely gender and race—exhibit strong associations with the formation of ties.