Yuanchen (Sophie) Bai

ABOUT

Hi! I am a second-year Ph.D. student in Information Science at Cornell University, working with Prof. Angelique Taylor, Prof. Wendy Ju, and Prof. Tapomayukh Bhattacharjee.

I received my master’s degree from Carnegie Mellon University (CMU). I was fortunate to work with Prof. Sherry Tongshuang Wu, Prof. Motahhare Eslami, and Prof. Haiyi Zhu. I also participated in a remote research internship with Prof. Xiaojuan Ma.

I received my B.S. in Electrical and Computer Engineering from Shanghai Jiao Tong University. I was fortunate to work with Prof. Pradeep Kumar Ray and Prof. Guangtao Zhai. During this time, I also participated in the winter research program at Massachusetts Institute of Technology (MIT) for Robotics.

RESEARCH INTERESTS

I design, build, and evaluate Embodied Agentic AI to improve real-world workflows and empower non-technical communities to meaningfully interact with these systems. I primarily work in healthcare domains, but I am also excited to explore broader contexts.

PUBLICATIONS (* equal contributions)

Towards Considerate Embodied AI: Co-Designing Situated Multi-Site Healthcare Robots from Abstract Concepts to High-Fidelity Prototypes

Yuanchen Bai, Ruixiang Han, Niti Parikh, Wendy Ju, Angelique Taylor

ACM CHI Conference on Human Factors in Computing Systems (CHI 2026)

We design high-fidelity robotic system prototypes that reduce healthcare workers’ burden and enhance care quality across emergency, rehabilitation, and sleep-care settings. Wse developed eight guidelines for creating considerate embodied AI systems, organized across four dimensions: Embodied Needs Grounding, Embodied Constraints & Feasibility, Embodied Literacy Building, and Embodied Design Space Expansion.

[PDF] [Project Page]

From MAS to MARS: Coordination Failures and Reasoning Trade-offs in Hierarchical Multi-Agent Robotic Systems within a Healthcare Scenario

Yuanchen Bai, Zijian Ding, Shaoyue Wen, Xiang Chang, Angelique Taylor

Under Review

We evaluate a hierarchical multi-agent robotic system in a simulated healthcare scenario. Our analysis identifies coordination failure modes, compares the behaviors of reasoning and non-reasoning models, and reveals the autonomy–stability trade-offs.

[PDF]

FARPLS: A Feature-Augmented Robot Trajectory Preference Labeling System to Assist Human Labelers’ Preference Elicitation

Hanfang Lyu, Yuanchen Bai, Xin LIANG, Ujaan Das, Chuhan Shi, Leiliang Gong, Yingchi LI, Mingfei Sun, Ming Ge, Xiaojuan Ma

ACM Conference on Intelligent User Interfaces (IUI 2024)

We propose a feature-augmented robot trajectory preference labeling system that highlights outliers, extracts keyframes, and adapts prompting based on user familiarity, comparison difficulty, and disagreement levels. It enhances preference criteria formation, labeling consistency, and engagement.

[PDF]

A Systematic Literature Review on Equity and Technology in HCI and Fairness: Navigating the Complexities and Nuances of Equity Research

Seyun Kim, Yuanchen Bai, Motahhare Eslami, Haiyi Zhu

28th ACM SIGCHI Conference on Computer-Supported Cooperative Work & Social Computing (CSCW 2025)

This literature review analyzes 202 papers from HCI and Fairness-focused venues. We identify research motivations, equity definitions and frameworks, and key themes on interventions, tensions, and trade-offs. Based on our findings, we propose an equity framework to help researchers address gaps and advance equity in technology.

[PDF]

ChatASD: LLM-based AI Therapist for Autism Spectrum Disorder (ASD)

Xiaoyu Ren*, Yuanchen Bai*, Huiyu Duan*, Lei Fan, Erkang Fei, Geer Wu, Pradeep Ray, Menghan Hu, Guangtao Zhai

International Forum of Digital Multimedia Communication (IFTC 2023)

We build an autism knowledge dataset covering education and real-case diagnostics. We propose ChatASD Therapist, an LLM-based diagnostic and treatment pipeline for autistic patients, supporting bilingual dialogue and facial video generation.

[PDF]

Measuring Adversarial Datasets

Yuanchen Bai*, Raoyi Huang*, Vijay Viswanathan, Tzu-Sheng Kuo, Tongshuang Wu

IJCNLP-AACL 2023 The ART of Safety Workshop

We analyze adversarial NLP datasets by comparing them to their original counterparts using quantifiable metrics across difficulty, diversity, and disagreement. Our findings reveal the intended and unintended impacts of adversarial transformations, offering insights for more transparent evaluation and dataset curation.

OTHER DESIGN

[Press]

Analytical Workspace for Behavioral Intervention Specialists in Supporting People with Intellectual and Developmental Disabilities

An analytical workspace that unifies behavioral patterns, narratives, context, and professional reasoning into a coherent interface. It supports Behavioral Intervention Specialists in organizing and interpreting behavioral data, facilitating collaborative decision-making.

Bayer Healthcare Educational Information Mobile App

A cross-platform mobile application myRadiology360 developed by Flutter which offers educational resources on imaging exams and disease states.

General Motors WeChat Mini Program for Marketing and Customer Care

A full-stack Direct-to-Customer (D2C) Cloud-based WeChat Mini Program for marketing and customer care. Supported by backend data analytics for user profiles extraction and strategies evaluation.