In October 2024, the School of Data Science and Society (SDSS) announced four seed grant awards in support of interdisciplinary data science research projects. The school had previously announced six seed grant awards in 2023.
“We’re excited to help support these interdisciplinary projects across campus,” said Stan Ahalt, dean of the School of Data Science and Society. “At our school, we want to be a hub for collaboration, and we want this funding, and other resources developed at the school to help nurture those intersections.”
In 2024, SDSS awarded seed funding to the following four projects:
An AI-powered decision aid for advancing women’s health choices
Co-PIs: Kandyce Brennan, assistant professor, School of Nursing
Tianlong Chen, assistant professor, department of computer science, College of Arts and Sciences
Rachel McInerney, clinical assistant professor, School of Nursing
Maureen Baker, associate professor, School of Nursing
Nursing faculty are collaborating with faculty in computer science to develop an AI chatbot that provides accessible sexual and reproductive health information, particularly in underserved areas. This chatbot will leverage natural language processing (NLP) to simulate human-like dialogue, promoting patient-centered health care by integrating holistic health care practices and clinical standards. Through this project, this team will establish a high-quality dataset to train similar models in various health care domains, particularly in areas where decision aids are used.
Promoting moral and civic education at Carolina and in the digital public sphere with the Parr Center’s ethical dilemma video series
PI: Michael Vazquez, teaching assistant professor, department of philosophy, College of Arts & Sciences; director of outreach, Parr Center for Ethics
Partnering institutions: department of psychology and neuroscience, department of computer science, College of Arts and Sciences; School of Information Library Sciences
This team will use the seed grant to support the continuation of an ongoing partnership between the Parr Center for Ethics and TED-Ed, producing ethical dilemma videos designed to foster open-mindedness, humility, and an appreciation for diverse perspectives. The video series uses ethical dilemmas as the primary tool for promoting appreciation for moral complexity. Using existing and supplementary materials, the team will prototype a mobile-friendly app or browser interface (dubbed “Parr Heel Academy”) to foster new ways to engage with the growing library of materials. Grant funds will be used to recruit students to test and refine the platform, to launch and create a campus-wide ethics learning initiative utilizing the platform.
Using LLMs to identify variant functions of genome-wide association studies (GWAS) related to psychiatric disorders
Co-PIs: Harlin Lee, assistant professor, School of Data Science and Society
Hyejung Won, associate professor, School of Medicine
Faculty at the School of Data Science and Society and the School of Medicine are teaming up to determine how genetic variants affect the risk of psychiatric disorders, which could help find better-individualized preventions and treatments. Scientists previously discovered genetic risk factors associated with psychiatric disorders through genome-wide association studies (GWAS). Still, their functional impacts must be experimentally validated using massively parallel reporter assay (MPRA) tools. While MPRA offers a high-throughput technique to validate variant function, it cannot be run on roughly 9 million common variants in the human genome. The project uses machine learning (ML) to generalize findings from MPRA to psychiatric disorder variants that are yet to be identified. Characterizing these variants prioritized by machine learning may identify novel insights into the biology of psychiatric disorder.
ChatNetZero: Demystifying net-zero with a fine-tuned large language model chatbot
Co-PIs: Angel Hsu, associate professor, department of public policy, College of Arts and Sciences and the Environment, Ecology, and Energy Program; director, Data-Driven Envirolab (DDL), Institute for the Environment
Shashank Srivastava, assistant professor, department of computer science, College of Arts and Sciences
Leveraging AI-driven large language models, this team has created and is fine-tuning ChatNetZero, a scientifically grounded Q&A platform that analyzes net-zero documents, detects greenwashing and provides transparent, interpretable insights to the public. Incorporating anti-hallucination and anti-greenwashing modules, ChatNetZero aims to combat misinformation, ensure accurate assessment of net-zero communities, and foster effective business and management strategies. This project contributed to the planning for a new Center for Climate Leadership and AI-driven Integrity (CLAIM). Hsu recently received support from the NSF to bring together collaborators to design CLAIM, to address societal and governance implications of generative AI on corporate climate commitments. Read more about the project at the SDSS website at https://go.unc.edu/AngelHsuSeedGrant.
“Innovative interdisciplinary research projects are vital for driving, expanding, and merging data science studies within and across fields, enhancing Carolina’s research quality and impact, and boosting Carolina’s ability to secure external funding,” said Terry Magnuson, chief research and strategy officer at the School of Data Science and Society. “We are so pleased to be supporting these exemplars.”
This year, SDSS launched the Carolina Data Science Researcher Community (DSCo) Hub, designed to grow data science research across Carolina. The DSCo Hub is a researcher support resource open to all faculty and staff data science researchers and collaborators across UNC-Chapel Hill. It consists of a virtual Microsoft Teams space for collaboration and resource sharing paired with in-person events and other staff-supported services.
Membership in the DSCo Hub is free. To join, submit a membership request and agree to the DSCo Hub code of conduct. More information can be found on the DSCo Hub webpage.