Rodriguez co-edits special issue on intersection of AI and social work

By Peter Murphy

Artificial intelligence’s impact on applied and industrial fields — including technology, medicine, and manufacturing — is well documented. However, AI and computational methods are affecting human services as well. University at Buffalo researchers are exploring what role these tools play in one of the most crucial societal professions: social work.

Maria Rodriguez, assistant professor in the Department of AI and Society, is co-editor with JoAnn Lee from the UB School of Social Work on a special issue of Research on Social Work Practice that describes how these methods are being used in social work, providing examples of the emerging field of computational social work.

“I tend to define social work as mitigating market failure. Folks who markets don’t work for, social work as a profession catches,” Rodriguez says. “Maybe someone’s abilities prohibit them from engaging in a market, so maybe they need some support. What are the jobs they can still do? Or maybe there is some supportive service adjacent to a market. We can connect them with — or even help them build it.”

Rodriguez was an assistant professor in the School of Social Work and a faculty member in the Department of Computer Science and Engineering before joining UB’s new Department of AI and Society in 2025. She and Lee are both interested in how advanced computational methods — like machine learning — are used in social work practice. They compiled seven papers detailing how researchers in social work and human services use advanced computational methods like network analysis, system dynamics and large language models.

“One of the so-called promises of AI writ large — maybe large language models (LLM) in particular — is the hope that they will help us deal with the more tedious parts of our work,” Rodriguez says. “For example, one paper asked if a large language model could actually help us conduct a systematic literature review.”

According to the paper, using an LLM via researcher programming or a commercial chat interface — in this case ChatGPT — to review literature produced comparable results. The limitations section of the article discusses the benefits and drawbacks of using an off-the-shelf foundation model to systematically review literature. For example, the misclassification of scholarship based on abstracts and/or titles is real concern, according to the researchers.

Another paper focused on using system dynamics in social work, a method that could be complimentary to the Bronfenbrenner Ecological Model, a foundational theory used in social work practice and research. The model explains an individual’s development in terms of several interrelated systems and chronosystems. It states that each person is an individual within a variety of systems, e.g., their immediate family, community, city, or state.

“Let’s say you have some issue, and we have a program that might help you. In whatever way we offer you that program, it will never be in a vacuum outside of these systems,” Rodriguez says. “If we’re trying to determine if the program works, we need to consider all those environmental contexts. The systems dynamics paper included in the special issue asked, what if we don’t run away from the fact that humans are complex, and try to account for some of this complexity instead?”

Rodriguez has been incorporating computational methods into her research in social sciences since she was a PhD student at the University of Washington. Back then, she needed to analyze 10 years’ worth of U.S. House of Representatives hearings on housing policy. She worked with John Wilkerson, professor in the Department of Political Science at the University of Washington, who suggested the idea of an automated content analysis.

“At the University of Washington School of Social Work, we took all of our statistics courses outside of our department. We had to learn R, which was controversial at the time because most folks were using SAS or STATA. Ironically, we quickly had to let go of R and learn Python shortly thereafter. Our professors in social work were surprised to learn we were programming,” Rodriguez says.

Rogriguez’s background using R — a statistical programming language — and then Python helped her learn some fundamental natural language processing for her qualifying exam.

After completing her degree, Rodriguez worked with the New York City Administration for Children’s Services as part of a training grant, where she began thinking about algorithmic decision-making concerning questions regarding equity. Rodriguez also participated in an early National Science Foundation trustworthy algorithms workshop.

Rodriguez believes the special issue of Research on Social Work Practice aligns with the vision of the Department of AI and Society, a department committed to solving problems using a transdisciplinary approach.

“I think this department aspires to build an intellectual space where we might bring the best of our respective disciplines in order to solve some of our most pressing problems,” Rodriguez says. “In order to do that, I think we need to build roads that allow scholars to work in teams. If we were on a two-lane highway before, we’re on a four-lane highway now.”