My research connects computational approaches to social questions with the evaluation of AI systems in practice. I examine how technical choices affect the quality of information, how people assess AI outputs, and how policy and organizational settings shape their use.
I use computational methods to study social and public health questions. My work includes text analysis, social network analysis, and machine learning applied to data on behavior, health, and social conditions. I am interested in how these methods can produce useful evidence and in the limits that data sources and analytic choices place on what we can conclude.
I study how AI systems are evaluated and incorporated into work shaped by policy, organizational processes, and professional judgment. My current research focuses on policy-aligned AI for child welfare documentation. It connects the reliability of underlying data and retrieval systems with the ways people review AI outputs and retain responsibility for decisions.
Foundation
Data and Evaluation Infrastructure
I examine how document conversion, data preparation, and retrieval affect the evidence available to an AI system. My work with child welfare policy documents includes evaluating conversion errors and retrieval performance, with attention to whether source material can be located and checked.
Theory
Organizations, Policy, and AI Use
I use a sociotechnical perspective to examine how policies, organizational roles, and work practices shape AI use. This perspective informs questions about where AI assistance fits into existing processes, what users need to assess its outputs, and how responsibility is assigned.
Tools
Policy-Aligned AI Systems
My dissertation focuses on retrieval-augmented generation for policy questions and the review of CPS investigation documentation. The research examines policy alignment and source citations, asking whether AI-generated responses and reviews provide evidence that users can verify.
Oversight
Human Judgment and AI Oversight
I am interested in how practitioners review, question, and act on AI-generated information. My research agenda includes human-in-the-loop and human-on-the-loop oversight, with attention to how these arrangements affect verification, professional judgment, and accountability.
Measuring and Repairing Document Conversion Errors in the Michigan Child Welfare Policy Manual
Examines document conversion errors in child welfare policy materials and approaches to identifying and repairing them.
AI Incidents and Marginalized Populations: A Descriptive Analysis of the AI Incident Database (1983–2026) from a Social Work Perspective
Examines documented AI incidents involving marginalized populations.
An Open Two-Corpus Dataset of Animal Harm Incidents and Criminal Judgments in China (AHID-CN)
Presents a dataset of animal harm incidents and criminal judgments in China.
Searching the Literature by Meaning: Benchmarking Free and Commercial Word Embedding Models on Social Work Text
Compares free and commercial embedding models for semantic retrieval on social work text.
Applying statistical machine learning methods to analyze differences in the severity level of COVID-19 among countries
Application of social network analysis of COVID-19 related tweets mentioning cannabis and opioids to gain insights for drug abuse research