UVA Workshop Explores the Transformative Power of Data Science in Education

Photo: Emma Candelier

Earlier this week, educators, researchers, and data scientists from across the University of Virginia and beyond gathered for a full-day workshop titled Data Science & Education Workshop: New Possibilities, Practices, and Perspectives. Organized by Virginia Education Science Training (VEST) and cosponsored by the School of Data Science and the School of Education and Human Development, the event brought to life the University’s interdisciplinary spirit and its commitment to data science for the common good.

Held at the School of Data Science in the Capital One Hub, the event opened with remarks from Dean Stephanie Rowley (School of Education and Human Development) and Quantitative Foundation Associate Dean for Academic and Faculty Affairs Jeffrey Blume (School of Data Science), who both emphasized the importance of collaboration across disciplines.  

“Investing in the intersection of data science and education is critical —especially as we face pressing questions about equity, access, and the role of AI in classrooms,” said Rowley. “This workshop is not only a reflection of years of collaboration between our schools, but also a launchpad for future partnerships that will shape how we teach, learn, and create opportunity.”

Exploring Data Science, AI, and Education at UVA

The theme of interdisciplinary innovation was echoed in the opening session by Brian Wright, Director of the Collaboratory for the Advancement of Education and Data Science and an associate professor at the School of Data Science. Wright's talk, "The Promise of AI and Data Science in Education Research," framed the rapid evolution of the field with both optimism and critical reflection.

“We started asking in 2019: What exactly is data science?” Wright recounted. “And we realized that we needed a new paradigm—one that centers not only on data and computation, but on people and partnerships.”

Wright’s talk blended technical demonstration with thought leadership. He introduced large language models and AI tools capable of coding qualitative data at scale, offering a glimpse into the future of mixed-methods research. Demonstrating a chatbot developed using OpenAI’s API and hosted on GitHub, he showed how tools like Token AI can not only engage with students in real time but also capture valuable interaction data to inform pedagogy and policy.

But Wright also urged caution. “It’s not just about what these tools can do, but how we use them. Our challenge is to apply them in ways that enhance equity, access, and human-centered learning.”

Workshop sessions throughout the day dug deeper into these themes. Ryan Baker from the University of Pennsylvania led a session on using data science for qualitative research, exploring how machine learning models can help code and analyze complex educational narratives.  

In the afternoon, Ted Quinn, Director of Wildfire Labs, discussed the development of technology to enhance teacher observations in Montessori classrooms. The project aimed to address teachers' needs for detailed, real-time data on student behaviors and interactions. Using Raspberry Pi cameras and location sensors, the system tracks students and objects in 3D, providing teachers with a narrative of student activities. This data helps teachers identify patterns, understand student progress, and address specific student needs. The system supports traditional observation methods, allowing teachers to correct and annotate computer-generated observations, enhancing their ability to personalize instruction.

Data Science and the Human Element in College Access and Success

In an afternoon session titled "People Drive Change: A Case for Caution Applying Data Science in Education," Associate Professor of Public Policy and Education Ben Castleman (Batten School of Leadership & Public Policy and School of Education and Human Development) explored how data science can be leveraged to improve college access and success, emphasizing both its potential and its limitations. He highlighted a project that used personalized text messages—powered by machine learning—to boost college enrollment, resulting in a 3% increase among high school students. Castleman also examined large-scale efforts to match students with personalized college information and shared findings from randomized trials showing that generic "nudge" strategies often fall short without the personal touch of trusted messengers.  

Turning to community colleges, he discussed tools like recommendation engines and predictive analytics designed to improve transfer rates and identify at-risk students, though results revealed persistent gaps between algorithmic and human judgment. Throughout the presentation, Castleman underscored the value of embedding data-informed supports—such as dedicated advisors and tutors—while cautioning against overreliance on technology at the expense of human-centered educational interventions.

“I worry we are leading with diagnosis and not the solution,” said Castleman. He went on to challenge the audience with, “Is the focus on data science crowding out investment in more effective interventions?”

Panel Discussion: Novel Data Science Applications in Education

A final panel showcased novel applications of data science across education, featuring faculty from UVA and partner institutions. Panelists included Vivian Wong (UVA School of Education and Human Development), Scott Acton (UVA Engineering & Applied Science), Hong Jiao (University of Maryland, College Park), Kylie Anglin (University of Connecticut), and Aidong Zhang (UVA Computer Science, Biomedical Engineering, and Data Science). The researchers shared insights on how AI and data science are reshaping the educational landscape, with projects ranging from adaptive assessments to the analysis of social networks in learning environments.  

Jiao and Zhang discussed machine learning applications in education, including the use of language models to generate research ideas and enhance interdisciplinary exploration. Acton introduced a pioneering video analysis project designed to understand classroom activity using custom AI models. Other topics included AI teaching assistants, tools to support teacher effectiveness, and methods for ensuring research validity. Panelists emphasized both the transformative potential of AI in supporting educators and the importance of addressing bias, improving interpretability, and fostering collaboration in model development.

Conclusion

The day concluded with a reception—an opportunity for informal conversation and future collaboration. As attendees reflected on the day’s sessions, many expressed enthusiasm for continued partnership across schools and disciplines.

The workshop aligns closely with the School’s mission to lead in data science research, education, and public service. It also reflects UVA’s broader commitment to innovation in education and a belief that data science, when applied responsibly and inclusively, can unlock new pathways to equity and excellence.

As Blume mentioned in the day’s opening remarks, “One of the founding principles of the School of Data Science is to rethink how we educate—from K–12 to higher education—and this moment with AI presents a sea change.” He added, “We’re not just talking about new tools; we’re talking about reimagining how knowledge is delivered, how students learn, and what skills they’ll need to thrive in a world transformed by data.”

As the field of education faces both new challenges and opportunities in the age of AI, the event served as a powerful reminder that it is not technology alone, but also thoughtful collaboration that will shape the future. And at UVA, that future is being built together—across disciplines, across schools, and always in service of the common good.