Research

Research Interests

My research centers on AI Sociology - the study of how artificial intelligence systems interact with and impact society. I focus on two primary areas:

AI Behavioral Studies & Human-AI Alignment: Understanding how AI systems behave in social contexts and developing methods to align AI behavior with human values and societal needs.

Machine Learning Applications for Social Impact: Applying ML techniques to address societal challenges, with particular attention to digital disparities, child development, and social dynamics.

Current Research Projects

Digital Disparities and Causal Inference

Representation-Augmented Causal Forests for Individualized Inference of Digital Disparities Using Panel Data
In preparation for submission to AAAI 2026 – Social Impact Track

Developing a causal forest model enhanced with deep representation learning to estimate individualized effects of digital access on youth development. Using China Family Panel Studies (CFPS) data spanning 2010–2022 to understand how digital divides impact individual life outcomes.

Early Childhood Development Mapping

Mapping Early Childhood Development Delays in China

Built machine learning models to predict early childhood development (ECD) delays and applied them to CFPS panel data (2010–2022) for provincial-level estimates. This foundational mapping supports large-scale child health interventions and evidence-based policy development.

LLM Agent Applications

Social Dynamics Simulation Suite & Web Intelligence

Developing LLM-powered applications to explore complex social systems and automate information processing:

• MamaVillage: Modeling intergenerational caregiving dynamics and family support systems
• LLM Agents Play Games: Investigating strategic competition and decision-making in multi-agent environments
• Leviathan: Exploring emergent governance structures and collective decision-making processes
• XPathAgent (Current research project): Utilizing advanced planning and reflective capabilities of LLM agents to generate general-purpose XPaths for efficient web information extraction

These projects demonstrate how LLM agents can model real-world social dynamics, decision-making patterns, and automate complex information extraction tasks.

Applied Research

Commercial AI Applications

Sales Chatbot
Showcased at Techequity-ai 2024 AI Summit

Led development of a virtual sales assistant using LangChain and GPT-4, achieving 98.9% QA accuracy for customer service in beauty products domain. This project demonstrates practical applications of conversational AI in commercial settings.

Research Philosophy

My work bridges theoretical AI research with practical social applications, emphasizing:

• Empirical rigor in studying AI-society interactions
• Causal inference methods for understanding real-world impacts
• Multi-scale analysis from individual to societal levels
• Policy relevance for addressing pressing social challenges

Methodological Expertise

Causal Inference: Causal forests, panel data analysis, treatment effect estimation
Machine Learning: Deep representation learning, predictive modeling, ensemble methods
LLM Applications: Agent-based modeling, simulation design, conversational AI
Data Sources: Large-scale survey data (CFPS), web data extraction, social media analytics

Contact

Email: brycew6m@stanford.edu
GitHub: @brycewang-stanford