Image
Postdoctoral Researcher Byunghwee Lee Wins Best in Show at Datapalooza 2025
The School of Data Science is pleased to announce that Postdoctoral Researcher Byunghwee Lee has been awarded "Best in Show" at Datapalooza 2025 for his research poster examining how large language models infer private traits from everyday online discourse. Byunghwee Lee is a postdoctoral research associate working in Professor Yong-Yeol Ahn’s lab.
Lee’s work investigates whether seemingly innocuous social and linguistic cues, such as musical preferences or casual slang, can enable AI systems to infer sensitive attributes like political leaning or sexual orientation. Analyzing online discussions from Debate.org and Reddit, his study shows that large language models can reliably predict an individual’s political alignment and significantly outperform traditional machine learning approaches. The findings highlight both the sophistication of modern AI systems and the potential privacy risks that arise when subtle social or cultural signals are embedded in public data.
Posters were judged by four School of Data Science faculty for clarity, organization, and use of data visualization.
