Yejin Choi

Yejin Choi is an Adjunct Professor at the Department of Linguistics, University of Washington, affiliated with the College of Arts & Sciences. Her work focuses on AI safety, large language model (LLM) reasoning, alignment of AI systems, and ethical challenges in AI. She contributes to advancing multimodal models, cultural knowledge representation, and mitigating biases in AI systems.

Her research spans theory-of-mind reasoning in LLMs, hallucination mechanisms, privacy preservation, and scalable solutions for self-learning agents. Recent studies include analyzing LLM behavior in ethical dilemmas, improving safety through collaborative agent systems, and benchmarking models in real-world scenarios.

Choi is involved in projects like WildVision, WildBench, and SafetyAnalyst, which evaluate AI systems' performance and safety in uncontrolled environments. Her work emphasizes practical applications of AI ethics, including mitigating political bias, ensuring privacy, and enhancing transparency in machine unlearning processes.

She collaborates on foundational challenges in AI alignment, proposing frameworks like Modular Pluralism and exploring cultural perceptions through prompts. Her contributions bridge technical innovation with societal implications, advocating for AI systems that reflect pluralistic human values.