Dr. Nari Yoo’s research uses data science and AI to study mental health service access among immigrants and multilingual populations in the United States. She combines natural language processing, large language models, geospatial analysis, machine learning, and survey-based methods to examine language-based and nativity-based differences. Her work includes mapping the availability of mental health providers who speak specific non-English languages using web-scraped directory data, analyzing what factors drive disparities in telemental health use and AI adoption between immigrant and native-born populations, and evaluating how large language models perform when applied to clinical and social service settings. More broadly, she is interested in how AI technologies are being adopted and experienced across diverse service contexts, including how social workers and community-based organizations working with immigrant communities are beginning to use AI tools in their practice. She has developed a continuing education program on AI for mental health for social workers and teaches data visualization and applied Python programming for social work students with no prior coding background.
Please describe one or two of your most interesting projects.
One of my ongoing projects uses web-based provider directories to map where speakers of specific non-English languages can and cannot find mental health providers who speak their language. It sounds like a simple question, but no existing dataset actually tracks this at a national scale, so I built the measurement from scratch using NLP and geospatial methods. I’m also working on a series of studies using nationally representative survey data to understand how immigrants and native-born populations differ in their adoption of telehealth, AI tools, and digital health services, and what structural and demographic factors explain those differences.
How did you end up where you are today? (Your research journey)
I picked up Python as a hobby during my master’s program in social work, with no intention of using it for research. But while working in immigrant-serving settings, I kept noticing the same thing: these communities were severely under-resourced, and there was almost no data to show it. When I started my PhD and tried to study language access in mental health care, the datasets I needed did not exist. Surveys undersampled immigrants, administrative records did not capture provider language capacity, and no one was systematically tracking whether non-English speakers could find care in their language. So I started building the data myself, using web scraping, NLP, and spatial analysis. That became my research. I joined the University of Michigan School of Social Work as an assistant professor in fall 2025.
What makes you excited about your data science and AI research?
The populations I study are usually the ones missing from the data. Immigrants and non-English speakers are undersampled in surveys, underrepresented in health records, and largely invisible in the existing data that shapes how services get planned and funded. What drew me to computational methods is that they let me work with data sources that were not designed for research, like web-based provider directories or social media text, and turn them into evidence. I also think there is a lot of room for data science and AI methods in social work research that has not been filled yet. I believe social work sits on questions that computational tools are well-suited to answer. Being in that space is what keeps me motivated.