- Long Beach Convention & Entertainment Center, 300 E Ocean Blvd, Long Beach, CA 90802 (Map)
- Date: August 7, 2023 (Monday)
- Time: 1:00 - 5:00 PM Pacific Time
The increasing demand for data-driven decision-making has led to the rapid growth of machine learning applications in various industries. However, the ability to draw causal inferences from observational data remains a crucial challenge. In recent years, causal inference has emerged as a powerful tool for understanding the effects of interventions in complex systems. Combining causal inference with machine learning has the potential to provide a deeper understanding of the underlying mechanisms and to develop more effective solutions to real-world problems.
This workshop aims to bring together researchers and practitioners from academia and industry to share their experiences and insights on applying causal inference and machine learning techniques to real-world problems in the areas of product, brand, policy, and beyond. The workshop welcomes original research that covers machine learning theory, deep learning, causal inference, and online learning. Additionally, the workshop encourages topics that address scalable system design, algorithm bias, and interpretability.
Through keynote talks, panel discussions, and contributed talks and posters, the workshop will provide a forum for discussing the latest advances and challenges in applying causal inference and machine learning to real-world problems. The workshop will also offer opportunities for networking and collaboration among researchers and practitioners working in industry, government, and academia.
Please submit your paper to the CMT portal site, and check the Call for Paper page for details on important dates and submission guidelines.
| Title | Speaker | Time (Duration) | Link |
|---|---|---|---|
| Introduction | Organizers | 1:00 - 1:10 PM (10 minutes) | |
| Invited Talk: COG: Creative Optimality Gap for Video Advertising | Raif Rustamov (Amazon) | 1:10 - 1:30 PM (20 minutes) | Slides |
| Invited Talk: The Value of Last-Mile Delivery in Online Retail | Ruomeng Cui (Emory) | 1:30 - 1:50 PM (20 minutes) | Slides |
| Leveraging Causal Uplift Modeling for Budget Constrained Benefits Allocation | Dmitri Goldenberg, Javier Albert (Booking.com) | 1:50 - 2:05 PM (15 minutes) | Slides |
| Ensemble Method for Estimating Individualized Treatment Effects | Kevin Wu Han, Han Wu (Stanford) | 2:05 - 2:20 PM (15 minutes) | Slides, Paper |
| A Scalable and Debiased Approach to Dynamic Pricing with Causal Machine Learning and Optimization | Nicolò Cosimo Albanese, Fabian Furrer, Marco Guerriero (Amazon AWS) | 2:20 - 2:35 PM (15 minutes) | Slides, Paper |
| An IPW-based Unbiased Ranking Metric in Two-sided Markets | Keisho Oh, Naoki Nishimura (Recruit Co), Minje Sung, Ken Kobayashi, Kazuhide Nakata (Tokyo Institute of Technology) | 2:35 - 2:50 PM (15 minutes) | Slides, Paper |
| Break & Poster Session | 3:00 - 3:30 PM (30 minutes) | ||
| Invited Talk: Unit Selection Based on Counterfactual Logic | Ang Li (UCLA) | 3:30 - 3:50 PM (20 minutes) | Slides |
| Invited Talk: Towards Automating the Causal Machine Learning Pipeline | Vasilis Syrgkanis (Stanford/EconML) | 3:50 - 4:10 PM (20 minutes) | Slides |
| Power and Pre-treatment Fit: Optimizing Synthetic Control Method for Quasi-experiments | Ali O Polat (Shipt) | 4:10 - 4:25 PM (15 minutes) | Slides, Paper |
| Dynamic Causal Structure Discovery and Causal Effect Estimation | Jianian Wang, Rui Song (NCSU) | 4:25 - 4:40 PM (15 minutes) | Slides, Paper |
| Hierarchical Clustering As a Novel Solution to the Notorious Multicollinearity Problem in Observational Causal Inference | Yufei Wu, Zhiying Gu, Alex Deng, Jacob Zhu (Airbnb) | 4:40 - 4:55 PM (15 minutes) | Slides, Paper |
Raif Rustamov is a Senior Applied Scientist at Amazon where he focuses on brand advertising science including relevance modeling, representation learning, and causal inference. He previously worked as a Principal Inventive Scientist in AI and Data Science at AT&T Labs conducting research on recommender systems, customer segmentation, identity for cross-device advertising, and location analytics. Raif has a PhD in Applied and Computational Mathematics from Princeton University and has taught at Purdue and Drew Universities, as well as worked as a research associate at Stanford University.
Video creatives play a crucial role in shaping consumer experiences and brand perceptions, but quantifying their impact on shopper experience remains a complex challenge. In this talk, we introduce the Creative Optimality Gap (COG), a metric developed to assess the relative optimality of video creatives using causal-inferential machine learning methodology. Our main contributions include the development of the COG metric through the use of conditional individual treatment effects projected on interpretable video features, the introduction of a meta-learner for its computation, and the incorporation of model uncertainty to avoid false positives. Our work advances the understanding of video creative effectiveness and provides a valuable tool for optimizing ad performance.
Ruomeng Cui is an Associate Professor of Operations Management at the Goizueta Business School, Emory University (on leave). She currently is a full-time Amazon Visiting Academic at Amazon, working in the supply chain domain. Her research focuses on causal inference, machine learning and data-driven modeling, with applications in retail, supply chains, and platforms. She currently serves as an associate editor for Manufacturing & Service Operations Management and Production and Operations Management. She received her Ph.D. in Operations Management from the Kellogg School of Management, Northwestern University and B.Sc in Industrial Engineering from Tsinghua University.
Last-mile delivery has become increasingly important in the online retail industry. In this study, we study the economic value of last-mile delivery. To do so, we conducted a quasi-experiment in collaboration with Cainiao, Alibaba's logistics subsidiary, where home delivery was launched at some pickup stations in 2021. This allowed us to comprehensively evaluate the causal impact of last-mile delivery. Using a difference-in-differences identification method, we found that last-mile delivery significantly increases sales and customer spending on the retail platform. To optimally prioritize limited delivery capacity, we employed causal machine learning to target the most responsive customers. Our findings suggest that online retailers should carefully weigh the costs and benefits of last-mile delivery and tailor their logistic strategies accordingly.
Dr. Li is set to join the Florida State University Department of Computer Science as an assistant professor in August. He is currently a post-doctoral researcher in the Department of Computer Science at UCLA under the guidance of Prof. Judea Pearl. His primary research area is causal inference, artificial intelligence, and causality-based decision-making, with a focus on building causal models that estimate treatment effects (interventions) and evaluating what would have happened if an individual had taken a treatment (counterfactuals). He is also interested in decision-making modeling using knowledge of treatment effects and counterfactuals. Prior to his post-doc, Dr. Li obtained his Ph.D. at UCLA with Prof. Judea Pearl and his M.S. degree at the University of Minnesota Twin Cities.
The unit selection problem aims to identify a set of individuals who are most likely to exhibit a desired mode of behavior, which is defined in counterfactual terms. A typical example is that of selecting individuals who would respond one way if encouraged and a different way if not encouraged. Unlike previous works on this problem, which rely on ad-hoc heuristics, we approach this problem formally, using counterfactual logic, to properly capture the nature of the desired behavior. This formalism enables us to derive an informative selection criterion which integrates experimental and observational data. We show that a more accurate selection criterion can be achieved when structural information is available in the form of a causal diagram. We further discuss data availability issue regarding the derivation of the selection criterion without the observational or experimental data. We demonstrate the superiority of this criterion over A/B-test-based approaches.
Vasilis Syrgkanis is an Assistant Professor in Management Science and Engineering and (by courtesy) in Computer Science, in the School of Engineering at Stanford University. His research interests are in the areas of machine learning, causal inference, econometrics, online and reinforcement learning, game theory/mechanism design and algorithm design. Until August 2022, he was a Principal Researcher at Microsoft Research, New England, where he was a member of the EconCS and StatsML groups. During his time at Microsoft, he co-led the project on Automated Learning and Intelligence for Causation and Economics (ALICE) and was a co-founder of EconML, an open-source python package for causal machine learning. He received his Ph.D. in Computer Science from Cornell University.
- Leveraging Causal Uplift Modeling for Budget Constrained Benefits Allocation, Dmitri Goldenberg (Booking.com)*; Javier Albert (Booking.com); Slides
- Ensemble Method for Estimating Individualized Treatment Effects, Kevin Wu Han (Stanford University)*; Han Wu (Stanford University); Slides, Paper
- A Scalable and Debiased Approach to Dynamic Pricing with Causal Machine Learning and Optimization, Nicolò Cosimo Albanese (AWS)*; Fabian Furrer (AWS); Marco Guerriero (AWS); Slides, Paper
- An IPW-based Unbiased Ranking Metric in Two-sided Markets, Keisho Oh (Recruit Co., Ltd.)*; Naoki Nishimura (Recruit Co., Ltd.); Minje Sung (Tokyo Institute of Technology); Ken Kobayashi (Tokyo Institute of Technology); Kazuhide Nakata (Department of Industrial Engineering and Economics, Tokyo Institute of Technology.); Slides, Paper
- Power and Pre-treatment Fit: Optimizing Synthetic Control Method for Quasi-experiments, Ali O Polat (Shipt Inc.)*; Slides, Paper
- Dynamic Causal Structure Discovery and Causal Effect Estimation, Jianian Wang (North Carolina State Unicersity)*; Rui Song (North Carolina State Unicersity); Slides, Paper
- Hierarchical Clustering As a Novel Solution to the Notorious Multicollinearity Problem in Observational Causal Inference, Yufei Wu (Airbnb)*; Zhiying Gu (Airbnb); Alex Deng (Airbnb); Jacob Zhu (Airbnb); Slides, Paper
- Community Detection-Enhanced Causal Structural Learning, Yuhe Gao (North Carolina State University)*; Hengrui Cai (University of California Irvine); Sheng Zhang (North Carolina State University); Rui Song (North Carolina State University); Poster, Paper
- ACE: Active Learning for Causal Inference with Expensive Experiments, Difan Song (Georgia Institute of Technology)*; Simon Mak (Duke University); C.F. Jeff Wu (Georgia Institute of Technology); Paper
- Evaluate the Impact of Similar Products Ad Group Recommendations with Causal Inference, Jamie Chen (Amazon)*; Zuqi Shang (AmaOn); Raif Rustamov (Amazon); Poster, Paper
- Machine Learning based Framework for Robust Price-Sensitivity Estimation with Application to Airline Pricing, Shahin Boluki (Pros Inc)*; Ravi Kumar (PROS); Poster, Paper
- OpportunityFinder: A Framework for Automated Causal Inference, Huy Nguyen (Amazon)*; Prince Grover (Amazon); Devashish Khatwani (Amazon); Poster, Paper
- Chu Wang, Amazon
- Yingfei Wang, University of Washington
- Xinwei Ma, UC San Diego
- Zeyu Zheng, UC Berkeley, Amazon - main contact
CausalML Team
- Jing Pan, Snap, CausalML
- Yifeng Wu, Uber, CausalML
- Huigang Chen, Meta, CausalML
- Totte Harinen, AirBnB, CausalML
- Paul Lo, Snap, CausalML
- Jeong-Yoon Lee, Uber, CausalML - main contact
- Zhenyu Zhao, Tencent, CausalML
EconML Team
- Fabio Vera, Microsoft Research, EconML
- Eleanor Dillon, Microsoft Research, EconML
- Keith Battocchi, Microsoft Research, EconML