Keynote presentations

It’s widely known that health disparities are not equally experienced across all populations, races and economic thresholds. Keynote speaker Andrea Parker, associate professor in the School of Interactive Computing at Georgia Tech and adjunct associate professor in the Rollins School of Public Health at Emory University and at Morehouse School of Medicine, shared her work to transform the health of communities through innovations in social computing. 

Parker said there is a large untapped potential to reduce health inequities using AI and social infrastructure, but there is only so much the technology alone can do. Partnering with community organizations to deliver content can help. For example, faith-based organizations were key to reaching unserved and vulnerable populations in the Black community in Parker’s work. 

In one project, Parker’s team created a faith-based app called Church Connect that offered a digital peer health advisor, a verse of the day and a prayer center to support users’ physical, social and spiritual well-being. The app looked like a traditional social media application, while providing a way to foster real-life social connections through engagement with the health tool. 

A second app, called Storywell, encouraged family physical activity through fit bands worn by parents and children that unlocked story chapters families could read together as activity goals were achieved. A three-month study showed Storywell increased bonding moments and supported educational learning and activity among family members, Parker said, illustrating how grounding tools in community organizations can help ensure they meet the intended audience’s needs.

“I think this is a hugely underinvested area,” Parker said.

In a second keynote, Cledar CEO and founder Hubert Niewiadomski discussed data-driven applications in the artificial intelligence (AI) space. The use of AI in the fintech market is expected to experience 17% growth by 2030, he said, opening the door for large language models to play a transformative role in the economy and fintech industry.

Large language models are tools used to analyze large language data sets, and they are gaining traction and adoption in finance for their ability to generalize large amounts of data while making sense of complex, disparate or even weak data. There are tradeoffs, though, Niewiadomski said. Bigger models are more expensive and offer more precision, while smaller models are less precise, but more affordable. And without humans, the models don’t work as well. Take ChatGPT, for instance. ChatGPT is a language model, not a decision model. This means that what comes out of the model is only as good as the data that’s put into it.

“It still requires a human touch to ensure accuracy, efficacy,” Niewiadomski said. “The model does not know material truth, it’s just a mathematical function. It doesn’t know anything about the semantics of the data. This requires building a lot of things around it to ensure accuracy.”

Asked where to put resources, Niewiadomski suggested hardware, infrastructure, people and policy. These are areas where UD’s faculty, researchers and students are already hard at work.

A growing community of data science expertise

The diversity of UD’s data science community has expanded in recent years to include affiliated researchers from more than 65 academic units on campus, from art history to engineering to music, Africana studies, neuroscience, philosophy, marine science and more. Inter-university connections have developed, too, across more than a dozen academic institutions, including Lincoln University and Delaware State University.

“This underscores why the University of Delaware is so great, because there is a community and a structure that allows for this type of exchange,” said Ben Bagozzi, chair of the Data Science Symposium planning committee and an associate director of the DSI. “It also gives data science at UD a nice character, in the sense that it is so open and inclusive with regard to such a broad array of skills and interests.” 

Broadening engagement with external partners was a key goal of this year’s symposium. Industry collaborators were invited to host information tables and participate in lightning talks and posters to foster an open exchange of information that can guide understanding of how to do meaningful data science research. And the response was positive, with participation from companies including DuPont, Center for Accelerating Financial Equity, ChristianaCare, Innovative Precision Health, Tech Impact and Kendal, among others.

“Learning what industries are doing, collaborating and sharing our research with diverse partners is really important in terms of our identifying new interests and expertise in the data science area,” said Bagozzi, who also is an associate professor of political science and international relations at UD.

Lightning talk topics spanned the gamut, enlightening attendees on the range of research underway, from industry collaborators that are leveraging data science to influence supply consistency in manufacturing, to nonprofit efforts to improve the financial health and wellness for underserved, financially vulnerable populations, to academic graduate student research projects focused on understanding how the Earth exhales, and how digital health can be used to optimize patient care.

Other presentations gave attendees a window into other growing initiatives on campus and in the state, including the Institute for Engineering Driven Health, the Research Expanding Access to Child Health (REACH) Center at Nemours Children’s Hospital, the Gerard J. Mangone Climate Change Science and Policy Hub and the Data Science Institute-FinTech Consortium.

AI infrastructure at UD

Artificial intelligence, like machine learning and deep learning, are tools that data scientists use to analyze and understand data. UD recently launched the AI Center of Excellence (AICoE) on campus to help advance multidisciplinary research and enhance responsible AI resources. It’s an important addition that aligns with a broader call nationally for institutes of this type, according to DSI Director Cathy Wu.

“Given the opportunities that the emerging AI technologies can offer on topics spanning health care, climate change, education, economic development, transportation and energy, it’s critical to further invest in AI research, especially when core principles such as safety, trust and security are at stake,” Wu said.

Modern AI is data-driven and computationally intensive. The advancements in data science and AI are both driven by increased computing power, the availability of large datasets and streaming data, and algorithmic advances in machine learning. Complementary in their capabilities, the DSI and AI Center of Excellence at UD aim to activate multidisciplinary research collaborations to transform how the data and computing revolution is harnessed, conduct foundational data science research, and develop leading-edge AI technologies to convert data into actionable knowledge to address critical societal issues and support broad stakeholders.

Providing access to world-class talent

Data science is booming, with U.S. Department of Labor statistics placing it among the top growing employment fields. Job growth in the field is expected to increase by 35% by 2032, faster than most other occupations.

Educating a future workforce, where data science is understandable and available to all, is a major step that can help ensure broader understanding of AI and data science. That said, hurdles remain in recruiting and training a diverse workforce capable of taking up the charge — and its challenges. In other words, educational equity matters.

Claude Tameze, professor of mathematical sciences and founding director of the Mathematics Learning Center at Lincoln University, has been working to increase representation in the community of folks capable of leveraging AI, machine learning and other tools for the benefit of society. 

“You need to have diversity in the room before you even begin,” Tameze said.

UD efforts are underway to build capacity in this growing field, too.  One recent example is the DSI + AICoE hackathon co-hosted by DSI and the AI Center of Excellence last summer that connected members of the UD community with local industry partners to address a series of problem-based challenges around code, data and applications.

One team, called Kendal Transformers, was composed of graduate students and industry members from Kendal, a Quaker-based nonprofit that operates affiliated senior-living communities. The Kendal Transformers team focused on understanding the needs of seniors as they progress through the aging process. The idea was to improve the abilities of health care providers to serve this population better by identifying predictive models to show when and where people will require additional care by classifying them as low- or high-risk for various predictors. Members of the team shared their work and lessons learned from this successful collaboration, alongside several other project teams at the symposium.