Yajie Xu

Intellectual Humility and Collective Wisdom

An IRB-approved study testing whether an AI training platform, integrating Socratic dialogue and digital storytelling, can strengthen users' intellectual humility (IH). A follow-up agenda will examine how enhanced IH shapes belief updating and the emergence of collective wisdom in networked decision-making. Target venue: CHI 2027.

Collaborators: Raiyan Abdul Baten , Mohammad Ratul

Timeline: AI-Assisted Cooperative Storytelling

A multiplayer reading game in which students sequence AI-generated images to reconstruct the story arc of a narrative text, building high-level reading literacy and cooperative skills. The pipeline uses Text-Narrative-To-Images (TNT-I), a system designed to preserve narrative continuity and discontinuity across panels, grounded in the situation-model theory of comprehension. Proposal submitted to AECT 2026; pilot data collection planned for summer 2026.

Collaborators: Glenn Smith , Maurício Segundo , Gene Kim

Endogenous R&D Collaboration Networks & Immigrant Executives

Developing game theory microfoundation and Bayesian structural models to identify endogenous network effects in collaboration networks, linking firm-level innovation data and immigrant executive information to measure spillover effects, uncover heterogeneity, and study how immigrant-led firms reshape R&D alliances and innovation outcomes.

Collaborators: Murat Munkin , Kazuma Takakura

Human-Centered NLP for Constructive Online Conversations

Combining NLP and social computing to study how people use rhetorical strategies to challenge stereotypes and promote constructive dialogue online. Develops Mandarin datasets and explores AI systems that can better understand and support healthy conversational norms.

Collaborator: Michael Bennie

Dynamic Style Rotation Strategies in Chinese Equity Markets

This project constructs a style-rotation index for Chinese equities and evaluates three timing strategies. The best-performing approach—rotation with market-exit rules—achieves a total return of 51.99 versus the benchmark’s 2.58 (≈20.2× higher) and lowers max drawdown from −0.719 to −0.249. A hedged variant also delivers stable, near–market-neutral performance with positive excess returns.

Celadon • Hugo Theme

An open-source portfolio theme exploring the equilibrium between academic rigor and creative design. Features a configuration-driven one-page architecture, native KaTeX support for econometrics, and a bespoke Morandi color system.

Skin Scan

AI-powered skin cancer detection platform that combines real-time webcam analysis, XGBoost image classification, and secure patient management to support early diagnosis and improve healthcare accessibility.