Research

Delivery SystemState policy environmentsWorkforce compositionLanguage accessAI adoption and implementationPopulationsMental health and wellbeingHelp-seeking and copingDigital technology and AI usePractitionersCapacity-buildingCultural and linguistic adaptationIntervention developmentImmigrant and Minority Wellbeingin the age of Artificial Intelligence

RESEARCH INTERESTS

I study well-being and service use among immigrant and racial, ethnic, and linguistic minority populations, and how digital technology is reshaping access, delivery, and implementation of care for these groups. My work spans three levels of the care system: (a) the delivery system, including workforce composition, the availability of culturally and linguistically concordant care, and state-level policy environments; (b) the populations, including their well-being, coping, help-seeking, informal care networks, and use of digital and AI-mediated resources; and (c) the practitioners, including capacity-building and intervention development for culturally and linguistically appropriate services. Across these three levels, I approach care delivery as a question of equity and implementation: how interventions reach, adapt to, and are sustained in communities that have been marginalized and underserved.

I combine quantitative and computational methods with community-engaged research approaches. Within social work and migration studies, I am especially interested in natural language processing (NLP) and large language models across research, practice, and education: in research, as methods for studying diverse populations; in practice, as technologies that shape assessment, documentation, and decision-making; and in education, as a growing area of training for the profession.

CURRENT PROJECTS

Language access as a structural determinant of immigrant wellbeing

Map of county-level spatial access to Spanish-language public mental health care across the United States, with service deserts shaded orange
Spatial access to Spanish-language public mental health care, by county, from the US Immigrant & Refugee Support Dashboard.

More than 25 million people in the United States, roughly 9 percent of the population over age five, speak English less than “very well,” and that population has grown substantially since 1990[1]. Language barriers are a documented obstacle to health and social services, including mental health care, for immigrant and refugee populations, alongside cost, insurance, stigma, and the limited availability of language-concordant services[2][3]. Title VI of the Civil Rights Act of 1964 and Section 1557 of the Affordable Care Act require federally funded health programs to take reasonable steps to provide meaningful access to individuals with limited English proficiency, and the National CLAS Standards set expectations for language assistance in health care[4]. The obligation is settled. What is not well described is the distribution of the workforce that would have to satisfy it.

Digital technology and AI use among immigrant and minority populations

A woman in a hijab using a laptop and a phone at home

For immigrants and minority communities, digital technology and AI carry two possibilities at once. They can lower barriers that were fixed before, reaching people across distance, cost, and sometimes language in ways an in-person system never did. They can also reproduce exclusion, when a tool is built around a default user and quietly works less well for everyone else[6]. Both are live at the same time; which one prevails in a given case is an empirical question, and the answer turns on how a tool is built and deployed and on who was kept in mind while it was.

AI literacy among practitioners

Hands arranging a stack of labels reading Workflow, Human Review, Prompting, and AI Literacy

Practitioners are increasingly expected to work alongside AI in assessment, documentation, and decision-making[10], and that expectation is arriving ahead of the training that would let them use these tools well. Practitioners who serve immigrant and minority communities carry an added weight, because a tool that works less well for those communities does its harm precisely where the stakes are already high. Most available ethics guidance assumes an organization with the resources to interpret and apply it[11], which does not describe many of the community settings where this work happens.

Computational social science methods for social work

A tablet showing the python.org homepage with a Fibonacci function in the code sample

Social work has always relied on data, but the questions it cares about have outrun the methods it conventionally uses. Whole populations leave traces in provider directories, administrative records, and public text, and reaching them calls for tools that social work training rarely covers: natural language processing, machine learning, geospatial analysis, and the collection of large digital datasets[13][14]. Running through my projects is a methodological thread that develops and adapts these tools for social work questions, and asks what changes when the profession takes them up.