NDSU researchers developing tool to deliver data-driven health interventions

The current healthcare system faces many challenges, including protecting the integrity of health data. To help deliver more effective, data-driven health interventions for areas such as North Dakota’s rural and underserved communities, NDSU researchers are developing StatWiseAI, an AI-powered educational and decision-support tool.

StatWiseAI differs from traditional AI in that it serves as a methodological mentor. Juan Li, NDSU computer science professor, leads the research team developing the program.

“StatWiseAI was born from the vision of creating an AI partner that doesn't just 'do' the math, but teaches the researcher how to navigate data with confidence and scientific integrity,” Li said.

“It helps users understand dataset organization, interpret variables, and, crucially, recognize pitfalls like data leakage,” Li added. “By grounding Large Language Models (LLM) in specific codebooks and data schemas, we ensure the AI provides evidence-based guidance rather than generic advice.”

The work, funded by the National Institutes of Health, features a collaboration with the University of Texas, led by Professor Bo Xie. It stems from observations that seasoned researchers often struggle with a ‘data barrier’ when working with large public datasets, due to the steep learning curve required to navigate complex data structures and underlying methodological assumptions. StatWiseAI was conceived from the vision of creating an AI partner that not only performs statistical analyses but also guides researchers in navigating data with confidence and scientific integrity.

Li’s focus is on human-centered AI and how it can be designed to serve as an “educational partner,” not to replace human judgment but to enhance it.

“We are researching how to bridge the gap between a researcher's intent and the technical execution,” Li said. “By identifying where analysts typically stumble when using large-scale public health data, we can design AI interventions that prevent errors and promote more trustworthy research outcomes.”

Biplab Kumar Sarkar is an NDSU doctoral student who has served as the lead developer for StatWiseAI and played a key role in developing the program's core architecture.

"Building StatWiseAI has been an incredible journey in bridging the gap between complex AI models and the practical needs of health researchers,” he said. “As the lead developer, my focus has been on ensuring the AI doesn't just provide answers, but actually understands the underlying data structure and guides the user toward more rigorous science. It’s rewarding to see our work at NDSU creating a tool that can help researchers turn raw data into meaningful health insights."

One of the central research findings from Li’s team is that general-purpose AI systems, such as the standard version of ChatGPT, are often confident but wrong when handling the nuances of specific datasets. They may generate plausible-sounding but inaccurate explanations or overlook complex sampling designs.

“Our research has proven that for AI to be useful in science, it must be grounded,” Li said. “We’ve successfully developed prototypes that anchor the AI’s logic in curated dataset knowledge. We are now moving into pilot studies to measure exactly how much this grounding improves a user’s analytical rigor and trust in the results.”

Li believes that StatWiseAI can help boost health care in rural states like North Dakota.

“Our researchers often rely on massive public datasets to study rural health disparities, aging populations and chronic disease trends,” she said. “However, smaller institutions or rural health departments may lack the massive data science teams found at larger coastal hubs. StatWiseAI democratizes access to high-level data expertise. By lowering the barrier to rigorous analysis, we empower local researchers to uncover insights that can lead to better resource allocation and more effective, data-driven health interventions for our rural and underserved communities.”

Li has also emphasized involving NDSU students in the testing and development phases, with a larger goal in mind.

“We aren't just building a tool,” Li said, “we are training the next generation of North Dakota’s workforce to lead at the intersection of AI, data science and health care.”

“Every one of our students in STEM fields needs to be exposed to the possibilities of the future of science,” said NDSU interim vice president for research and creative activity Heidi Grunwald. “Professor Li’s collaborative project aligns well with the current administration’s focus on the nexus of big data, the need for topic-based AI models and workforce initiatives. I’m excited about this project in particular because it is the future of STEM education, but it also cautions us about using AI without guardrails for identifying hallucinations and false positives and negatives.”