UVA Darden LaCross AI Institute Awards Fellowships in AI Research

The University of Virginia LaCross Institute for Ethical Artificial Intelligence in Business recently announced the recipients of its 2025 Fellowships in AI Research (FAIR).

The Fellowship program serves as vehicle for the LaCross AI Institute to pursue and support collaboration in ethical AI research across multiple disciplines, and this year’s recipients exemplify that collaborative spirit.

The first 2025 Fellowship in AI Research was awarded to UVA Assistant Professor of Computer Science Ferdinando “Nando” Fioretto, collaborating with Max Biggs, Assistant Professor of Business Administration at the UVA Darden School of Business. Their proposal is titled “Privacy and Fairness in AI Pipelines: From Data Collection to Decision-Making.”

The second Fellowship was awarded to Darden Assistant Professor of Business Administration Samuel Levy and Chirag Agarwal, Assistant Professor of Data Science at the UVA School of Data Science. Their proposal is titled “Privacy-First Advertising and Machine Unlearning: Scaffolding Consumer Trust with Ethical AI.”

“The proposals indicate the significant amount of AI-related research activity at UVA, including in topics at the intersection of ethics and business,” said Marc Ruggiano, Director of the LaCross AI Institute. “We recognize the effort, expertise and passion that were evident in all of the proposals. There were many deserving ones, and we hope to see all of them progress in the near future.”

The LaCross AI Institute received proposals from across six UVA schools, many of them proposing collaboration between two or more researchers, representing 10 disciplines, and three universities. The proposals addressed each of the Institute’s focus areas: data privacy and ethical AI, human-AI performance, bias and misinformation, analytical leadership, and healthy choices. All proposed projects included substantial student involvement.

Privacy and fairness in AI pipelines

The proposal developed by Biggs and Fioretto seeks to explore ways for policymakers and data users to increase the accuracy of data gathered for public uses while minimizing the possibility of introducing bias into the work and while protecting privacy of those participating.

The researchers examined how large entities like government agencies and global businesses use large-scale surveys to gather data that is then used to build AI models and often inform funding and policy decisions. One example is the U.S. Census Bureau’s American Community Survey. Such surveys might only sample 1% of a massive population and incorporate methods to account for accuracy, reduce bias and protect privacy, but the researchers have found such work still includes issues such as unintentional biases on underrepresented populations in the sample.

Their work is intended to identify where and why the inaccuracies and biases occur and recommend approaches that address them.

Privacy-first advertising and machine unlearning

Levy and Agarwal have identified an opportunity to conduct research that can advance the state of digital advertising and “lay the groundwork for an ethical, privacy-forward future.”

Their project recognizes the need for digital advertising to evolve in ways that respect increasing consumer awareness and build trust amid stricter privacy regulations. Levy and Agarwal aim to integrate the concept of “machine unlearning,” which would permit an advertiser to “forget” a user’s data and give the user more control over their information.

“Our proposed methodologies ensure that privacy-preserving techniques do not compromise marketing effectiveness, paving the way for more resilient, ethical advertising ecosystems,” their proposal states.