CSTAT attends the 2025 ISU–NISS Conference on AI and Statistics
Drs. Wenjuan Ma and Sichao Wang attended the 2025 ISU–NISS Conference on AI and Statistics at Iowa State University. This three-day event brought together leading researchers to explore cutting-edge developments at the intersection of artificial intelligence and statistical science, opening with an insightful workshop on reinforcement learning and featuring recent advances in online statistical inference, sequential decision-making, and causal machine learning.
The conference showcased exciting applications across numerous fields, including forensic science, survey statistics, spatiotemporal modeling, and hazard analysis. Among the many compelling sessions, one talk focused on "Reinforcement Learning for Respondent-Driven Sampling," which introduced a novel adaptive design for surveying hard-to-reach populations. By using reinforcement learning to dynamically tailor incentives, this method promises to significantly improve the efficiency and cost-effectiveness of public health and social science studies, while tackling the complex statistical challenges of making valid inferences from adaptively collected network data.
Equally compelling was the presentation on "Bridging Causality and Deep Learning with Causal Generative Models." This talk addressed a core challenge in modern AI: while generative models show incredible creativity, they often lack fundamental causal reasoning skills. The researchers outlined an ambitious effort to merge deep learning's expressiveness with the rigor of statistical causality, aiming to build models that can discover causal structures directly from complex, unstructured data. A key focus involved constructing interpretable causal factors from the hidden units of deep learning models, moving beyond black-box representations. This work is pivotal for developing autonomous, intelligent systems that not only predict but truly understand the world.
Beyond these deep dives, the conference featured a public conversation with Nate Silver on the future of AI in forecasting and risk.
Our team returned energized by the potential of these innovative methodologies. Building on CSTAT’s established and expanding AI capabilities, we welcome new collaborations to leverage these cutting-edge tools in AI and machine learning to advance scientific discovery.