Causal Inference in the Era of Personalization: Handling Complexity with Simplicity

• Activity Number: 374 - Mental Health Statistics Section P.M. Roundtable Discussion (Added Fee)
• Type: Roundtables
• Date/Time: Tuesday, July 31, 2018 : 12:30 PM to 1:50 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Kosuke Imai* and Xiao-Li Meng and Booil Jo
• Companies: Princeton University and Harvard University and Stanford University
• Keywords: personalized; causal inference; heterogeneity; complexity; accessibility

Abstract

How controlled experiments and causal inference should adapt in response to the availability of big data and growing interest in personalized medicine has been a curious and cautious question. As we shift our interest to more individualized inference, it becomes unavoidable to deal with increased complexity with larger and high-dimensional data that captures individual heterogeneity. This situation of course makes principled approaches of causal inference extremely difficult. Ironically, as we try to cope with ever increasing complexity, we are also faced with growing demands for simple and accessible tools for causal inference, perhaps influenced by widespread use of machine learning and predictive modeling techniques. How far can we go with our desire to learn complex truths with simple and intuitive methods? This session will provide a sorely needed opportunity to discuss this apparently puzzling situation and clarify some potential directions.