Refreshments are available starting at 10:30 a.m. The seminar will begin at 10:45 a.m.
Abstract
Our everyday choices—from what we eat to how we communicate—carry significant consequences for our health, relationships, and the environment. These actions impact critical societal challenges such as climate change and growing polarization. How could computer science help? While generative AI offers new opportunities to empower prosocial actions at scale, several key challenges limit our ability to leverage AI tools for good. In this talk, Kristina Gligorić will explore how AI can identify opportunities for intervention and deliver scalable solutions to address pressing societal issues.
Gligorić will outline how she carries out this approach in two examples of societal issues fundamental to fostering a healthy and equitable society: civility and sustainability. First, she will discuss how generative AI can support constructive online conversations. She will present novel large language model methods for detecting harmful content (Proceedings of the National Academy of Sciences ’24) and describe a human-in-the-loop system that uses generative AI to assist users in creating constructive posts, developed in an ongoing collaboration with a social media company. In the second part, Gligorić will then shift focus to climate issues and promoting sustainable dietary habits in campus environments, highlighting findings on barriers to sustainability (PNAS Nexus ’24, Frontiers in Nutrition ’24). She will demonstrate how AI tools can assist dining hall chefs and food scientists in revising menus and designing experiments to promote more sustainable eating habits. These interventions have made a positive impact on many people; the work on sustainable dietary behaviors helped shape an on-campus food system that serves thousands of students and staff daily, while online interventions changed the behavior of thousands of users and moderators, with changes persisting over six months.
Finally, Gligorić will outline efforts to develop AI interventions across diverse societal contexts. By combining LLM predictions with expert annotations (arXiv ’24), we can extend AI’s impact across domains like health care, policymaking, and law. Gligorić will conclude by discussing future directions for (1) improving LLMs to address social contexts better, (2) creating tools to assist social scientists, and (3) scaling AI assistance tools and human-in-the-loop interventions to empower experts in tackling complex problems. These directions will enable using AI to address societal challenges at scale.
Speaker Biography
Kristina Gligorić is a postdoctoral scholar in Stanford University’s Computer Science Department. Previously, she obtained her PhD in computer science at the École Polytechnique Fédérale de Lausanne, or EPFL. Her research focuses on computational approaches to address societal issues by developing and applying large language models, data science, and causal inference methods. Her work has been published in top computer science conferences focused on computational social science and natural language processing—such as the Association for Computational Linguistics, the ACM Conference on Computer-Supported Cooperative Work and Social Computing, and the International Association for the Advancement of Artificial Intelligence Conference on Web and Social Media (CWSC)—and broad audience journals including the Proceedings of the National Academy of Sciences, Nature Communications, and Nature Medicine. She is a Swiss National Science Foundation fellow, a Massachusetts Institute of Technology Electrical Engineering and Computer Science Rising Star, and a University of Chicago Rising Star in Data Science. She has received several awards for her work, including the EPFL Thesis Distinction, a CSCW 2021 Best Paper Honorable Mention Award, and the EPFL Best Teaching Assistant Award.