Abstract Number:
1436
Submission Type:
Invited Paper Session
Participants:
Aaditya Ramdas (1), Aaditya Ramdas (1), Dae Woong Ham (2), Molly Offer-Westort (3), Koulik Khamaru (4), Kelly Zhang (5)
Institutions:
(1) Carnegie Mellon University, Pittsburgh, PA, (2) Harvard University and Netflix, Cambridge, MA, (3) University of Chicago, Chicago, IL, (4) Rutgers University, New Brunswick, NJ, (5) Columbia University, New York, NY
Chair:
Session Organizer:
Speaker(s):
Session Description:
With the increased digitization of our lives, more and more adaptive algorithms, including reinforcement learning algorithms, are used for automated decision making and experimentation in a variety of areas, from online advertising and recommendation systems, to political surveys and digital health. Adaptive learning algorithms, characterized by their ability to learn and dynamically adjust randomization probabilities over time, offer a powerful means of experimental design. Adaptive learning algorithms can be used to provide users with better, more personalized digital experiences (minimize regret), as well as to optimize the precision of treatment effect estimates (maximize power). However, standard statistical methods which assume treatments are assigned independently fail to hold when adaptive learning algorithms are used, and standard normal approximations and randomization test approaches can be invalid on this type of data. Moreover, there is a lack of a comprehensive framework for statistical inference on data collected with adaptive learning algorithms that can be used for a variety of adaptive designs (early stopping, policy learning, incorporating prediction models) to address various statistical questions (hypothesis testing, evaluating an estimated optimal policy, etc.)
The primary goal of this proposed session is to shed light on both the theoretical and practical challenges for statistical inference for data collected with adaptive and reinforcement learning algorithms. These will include practical aspects of executing adaptive experiments, specifically in internet companies and social sciences. Additionally, the talks will discuss the statistical challenges associated with different types of experimental designs and inferential questions, including a) early stopping, b) policy learning and evaluation from data collected with adaptive algorithms, and c) observational data settings in which the treatment propensities are not available to the data analyst.
We anticipate that this session will be of great interest to many JSM attendees. Adaptive experimentation and reinforcement learning are growing both in terms of the statistical methods being developed and practitioners who are using those methods. Last year, JSM featured several invited sessions on adaptive experimentation, and we anticipate those who engaged in these sessions last year will find our upcoming session equally compelling.
The format of the session will be four presentations by the speakers, each about 20 minutes long, followed by a 20-minute panel discussion about the open challenges and new directions in adaptive experimentation and reinforcement learning led by the chair. All four speakers are from diverse backgrounds and experiences.
Speaker presentations:
(1) "Design-Based Confidence Sequences for Adaptive Experiments" by Dae Woong Ham (Harvard University and Netflix), Michael Lindon (Netflix), Martin Tingley (Netflix) & Iav Bojinov (Harvard University)
(2) "Data Driven Experiment Designs for Policy Learning and Evaluation" by Molly Offer-Westort (University of Chicago)
(3) "Learning with High Dimensional Observational Data" by Licong Lin (UC Berkeley), Mufang Yaing (Rutgers), Suvrojit Ghosh (Rutgers), Koulik Khamaru (Rutgers), and Cun-Hui Zhang (Rutgers).
(4) "Policy Learning and Evaluation after Adaptive Sampling" by Kelly W. Zhang (Columbia University), Yash Nair (Stanford University), & Lucas Janson (Harvard University).
Sponsors:
Association for the Advancement of Artificial Intelligence 2
IMS 3
Section on Statistical Learning and Data Science 1
Theme:
Statistics and Data Science: Informing Policy and Countering Misinformation
Yes
Applied
Yes
Estimated Audience Size
Small (<80)
I have read and understand that JSM participants must abide by the Participant Guidelines.
Yes
I understand and have communicated to my proposed speakers that JSM participants must register and pay the appropriate registration fee by June 1, 2024. The registration fee is nonrefundable.
I understand