Causal Inference and Machine Learning in Practice: Use cases for Product, Brand, Policy and Beyond

Schedule

  • Long Beach Convention & Entertainment Center, 300 E Ocean Blvd, Long Beach, CA 90802 (Map)
  • Date: August 7, 2023
  • Time: TBD

Abstract

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.

Paper Submission

Please submit your paper to the CMT portal site, and check the Call for Paper page for details on important dates and submission guidelines.

Outline

Title Speaker Duration Link
Introduction Organizers 10 minutes  
Invited Talk #1 Raif Rustamov 20 minutes  
Invited Talk: The Value of Last-Mile Delivery in Online Retail Ruomeng Cui    
20 minutes      
Paper #1   15 minutes  
Paper #2   15 minutes  
Paper #3   15 minutes  
Paper #4   15 minutes  
Break & Poster Session   30 minutes  
Invited Talk #3 Ang Li 20 minutes  
Invited Talk #4: Towards Automating the Causal Machine Learning Pipeline Vasilis Syrgkanis 20 minutes  
Paper #1   15 minutes  
Paper #2   15 minutes  
Paper #3   15 minutes  
Paper #4   15 minutes  

Invited Speakers

Raif Rustamov, Amazon

Bio

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.

Abstract

Video creatives have a substantial impact on consumer experiences and brand perceptions, but evaluating their effect on shopper behavior remains a complex challenge. In this talk, we introduce the Creative Optimality Gap (COG), a novel metric developed using causal-inferential machine learning methodologies to quantify the relative optimality of video creatives. We provide an example application of this approach in assessing the effectiveness of video creatives for brand advertising at Amazon.

Ruomeng Cui, Emory University/Amazon

Abstract

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.

Ang Li, University of California, Los Angeles

Bio

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.

Abstract

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, Stanford University/EconML

Bio

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.

Accepted Papers

To be updated

Organizers

  • 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