Rebuilding Statistics in the Age of AI: Culture, Infrastructure, and Training

• 成果类型: Article; Early Access
• 署名作者: Donoho, David L.; Kang, Jian; Lin, Xihong; Mukherjee, Bhramar; Nettleton, Dan; Nugent, Rebecca; Rodriguez, Abel; Xing, Eric P.; Zheng, Tian; Zhu, Hongtu
• 署名单位: Stanford University; University of Michigan System; University of Michigan; Harvard University; Harvard T.H. Chan School of Public Health; Harvard University; Harvard University; Massachusetts Institute of Technology (MIT); Broad Institute; Yale University; Yale University; Iowa State University; Carnegie Mellon University; University of California System; University of California Santa Cruz; Mohamed bin Zayed University of Artificial Intelligence MBZUAI; Carnegie Mellon University; Columbia University; University of North Carolina; University of North Carolina Chapel Hill; University of North Carolina School of Medicine
• 刊物名称: JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
• ISSN/ISSBN: 0162-1459; 1537-274X
• DOI: 10.1080/01621459.2026.2680161
• 发表日期: 2026-07-09
• 关键词: Data curation and annotation; Empirical modeling and evaluation; Statistics and AI; Uncertainty Quantification; Workforce training and incentives

摘要

This perspective article distills themes from the 2024 JSM town hall, Statistics in the Age of AI, where panelists discussed how the field should evolve as foundation models, large-scale empirical modeling, and data-intensive infrastructures reshape science and society. Rather than a transcript or comprehensive survey, the article offers a forward-looking perspective grounded in the town hall discussion and the authors' reflections. We organize the discussion around five recurring questions: (i) disciplinary culture and incentives, (ii) the role of data curation and annotation, (iii) engagement with modern empirical modeling, (iv) training for large-scale AI applications, and (v) partnerships with key AI stakeholders. We emphasize that statistics already leads in areas such as uncertainty quantification, study design, evaluation, and inference, while arguing that some parts of the field could engage more deeply with data engineering, cloud-scale computation, and end-to-end AI systems. We conclude with pragmatic recommendations for training, incentives, and collaboration so that statisticians can help shape the data-centric future.