Embedded Files
I am an undergraduate student at Yonsei University, majoring in Economics and Computer Science. I am engaging in social AI research at SNU PI Lab, advised by Youngjae Yu.
I aspire to be a researcher who is well aware of the broader implications of her work and strives to contribute to society. As AI drives changes at an unprecedented speed and scale, my mission is to mitigate its negative societal impacts and prevent its malicious uses.
photo by: my lovely sister Shua
credits: my lovely parents
Research Interests
I aim to address potential ethical and social issues by building safer AI systems that can navigate the complexity of real-world human interactions. I believe that advancing human-like cognition is key to building AI systems that are not only capable, but safe and beneficial for society.
I am interested in (1) evaluating how frontier AI models behave in context-rich, interactive scenarios to identify safety risks and (2) integrating interdisciplinary insights from game theory, behavioral economics, and cognitive science to enhance social and cognitive reasoning capabilities of AI. I'm also interested in investigating how bias shapes social impacts of AI, especially in applications like policymaking.
My interests include but are not constrained to the above as I continue to actively explore problems with meaningful ethical and social implications.
Education
Yonsei University
undergraduate (2022-)
B.A. Economics
B.S. Computer Science and Engineering
Chungnam Samsung Academy (CNSA)
Economics & Business, International Studies & Humanities (Dual Diploma)
Bilingual Diploma, Honor Diploma
Publications
Investigating Counterfactual Unfairness in LLMs towards Identities through Humor
Shubin Kim*, Yejin Son*, Junyeong Park, Keummin Ka, Seungbeen Lee, Jaeyoung Lee, Hyeju Jang, Alice Oh, and Youngjae Yu.
Under review
KatFishNet: Detecting LLM-Generated Korean Text through Linguistic Feature Analysis
Shinwoo Park, Shubin Kim, Do-Kyung Kim, and Yo-Sub Han.
ACL 2025 / Paper / Website / GitHub
Despite the significance of detecting LLM-generated text, most detection methods focus on English. With its unique spacing rules, rich morphology, and distinct punctuation patterns, the Korean language requires specialized approaches that English-centric methods fail to accommodate. We introduce KatFish, the first benchmark dataset for LLM-generated Korean text and analyze key linguistic differences between human and LLM-generated text. Using these insights, we propose KatFishNet, a detection method tailored to Korean that leverages linguistic features. Our work provides essential resources, methodologies, and insights for non-English LLM detection.
Honors & Awards
Scholarships
National Academic Excellence Scholarship (Humanities; Major Establishment, Type I) 2024 - Present
* Government scholarship from KOSAF given to 2 students from Yonsei University, a total of 270 students in Korea.
Merit-Based Yonsei Scholarship (Fall 2022, Spring 2023, Fall 2023)
* Partial tuition based on GPA.
Honors
Highest Honors (Spring 2022), Honors (Spring 2023), Honors (Fall 2023)
* Awarded by Yonsei University.
Awards
English
CNSA English Debate Contest (Gold) 2019
CNSA English Speaking Contest (Silver) 2018
CNSA English Writing Contest (Gold) 2018
Chinese
CNSA Chinese Writing Contest (Silver) 2019
CNSA Chinese Writing Contest (Silver) 2018
Acting
Korean Language Play Contest (2nd place, national) 2018
English Play Contest (2nd place, city-wide) 2014
English Play Contest (1st place, city-wide) 2013
Course Projects
LLM-Generated Korean Text Authorship Classification
Course: Theory & Practice of Deep Learning (2024)
Fine-tuned GPT-2 on a custom dataset created using 4 state-of-the-art LLMs. Designed tests to assess its generalization capabilities by conducting experiments in both in-domain and out-of-domain settings. Performed zero-shot experiments with HyperCLOVA X, an LLM trained largely with Korean corpus.
Netflix User Classification Contest
Course: Understanding & Application of Data Science (2022)
My first experience with machine learning and Kaggle. Processed Netflix user data and built an ensemble model using MLP and SVM. Ranked 1st place.
Hunmin Jeomja (훈민점자) & Moonlightype
Course: Problem Solving & Algorithms (2021)
With a focus on raising social awareness, developed an application that interprets the positions of keyboard keys as braille and converts it to Korean / English characters.
Extracurriculars
Compassion Mate
Volunteer Translator & Monitor (2019-2023, 2025- | 4.5 years-)
In 2019, I volunteered as an International Sunday School Teacher. Motivated by my experience of interacting with young children from Nigeria and Uganda, I wanted another consistent way to help young children around the world. I volunteered as a Compassion Mate, translating English letters written by young children (mainly from Southeast Asia, South America, and Africa) to their Korean benefactors. Furthermore, I volunteered as a Compassion Mate Monitor, screening letters translated by new volunteers and providing feedback.
A Midsummer Night's Dream
Alumni Collaboration Musical Production
Head of Sound & Head of Costume (2023)
My first time working behind the scenes in a performing arts production. In a large-scale production performed at the Yonsei Amphitheater, I was responsible for scheduling, sound design, managing technical equipment, and operating sound effects. I gained skills in leadership, adaptability, and hands-on problem-solving, successfully navigating complex team dynamics and timely demands under pressure.
Global Lounge Ambassador
Event Organizer & Host (2019)
I was selected by the Department of Foreign Languages to organize events related to Sustainable Development Goals. I was in charge of the whole process from planning, advertising to hosting. I cultivated experience on how to cleverly identify projects that would draw students' attentions and also effective ways to advertise within budget constraints.
Hello? Daddy!
Actor (2017-2018)
With my passion for sharing my ideas with the public, I was immensely involved in performing arts. Particularly, I was in a teenage musical group based in Seoul and performed Hello? Daddy! in Daehak-ro, Seoul. As I was not living in Seoul at the time, I commuted for 5 hours each day. This experience taught me about the grit and dedication needed to commit to my goals.
Report abuse