2nd Workshop on Causal Inference and Machine Learning in Practice

Schedule

  • Room 116, Centre de Convencions Internacional de Barcelona (CCIB), Plaça de Willy Brandt, 11-14, Sant Martí, 08019, Barcelona, Spain (Map)
  • Date: Monday, August 26
  • Time: 9:00 AM – 1:00 PM (CEST)

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.

Program

Title Speaker Time (Duration) Link
Introduction Roland Stevenson 9:00-9:10am (10m)  
Invited Talk #1 Ruocheng Guo 9:10-9:30am (20m) Slides
From Causality to Incrementality: Estimating Marketplace Inter-product Cannibalisation Effects at Expedia Group Florian Florian*, Catriona Mitchison (Expedia) 9:30-9:45am (15m) Paper
DISCO: constrained bandits for personalized discount allocation within fashion e-commerce Jason Zhang (TripAdvisor), Benjamin M Howson (Imperial College London), Panayiota Savva, Eleanor Loh (ASOS)* 9:45-10:00am (15m) Paper, Slides
Industrial-Grade Smart Troubleshooting through Causal Technical Language Processing: a Proof of Concept Alexandre Trilla*, Ossee Yiboe, Nenad Mijatovic (Alstom), Jordi Vitria (Universitat de Barcelona) 10:00-10:15am (15m) Paper, Slides
Practical Marketplace Optimization at Uber Using Causally-Informed Machine Learning Zijun Qin*, Siyu Chen, Jason Dowlatabadi, Yu Xuan Hong, Vinayak Iyer, Uday Mantripragada, Rishabh Narang, Apoorv Pandey, Abrar Sheikh, Jiaqi Sun, Matthew Walker, Kaichen Wei, Chen Xu, Jingnan Yang, Allen Zhang, Guoqing Zhang, Bobby Chen, Hongtao Sun (Uber) 10:15-10:30am (15m) Paper, Slides
Break & Poster Session #1   10:30-11:00am (30m)  
Invited Talk #2 Rumen Illiev 11:00-11:20am (20m) Slides
Enhancing Uplift Modeling in Multi-Treatment Marketing Campaigns: Leveraging Score Ranking and Calibration Techniques Yoon Tae Park*, Ting Xu (Bestbuy) 11:20-11:35pm (15m) Paper, Slides
Optimization and Control Applications through Repeated Interventions Based on Structural Causal Models Daigo Fujiwara*, Tomonori Izumitani (NTT), Shohei Shimizu (Shiga University & RIKEN) 11:35-11:50am (15m) Paper, Slides
Understanding Product Quality with Unstructured Data: An Application of LLMs and Embeddings at Airbnb Zhiying Gu*, Jacob Zhu, Brad Li, Linsha Chen (Airbnb) 11:50-12:05pm (15m) Paper, Slides
Sensitivity Analysis for Causal ML: A Use Case at Booking.com Philipp Bach (University of Hamburg), Victor Chernozhukov (MIT), Carlos Cinelli (University of Washington), Lin Jia (Booking.com), Sven Klaassen (University of Hamburg), Nils Skotara (Booking.com), Martin Spindler (University of Hamburg) 12:05-12:20pm (15m) Paper, Slides
Poster Session #2   12:20-12:50pm (30m)  
Closing   12:50-1:00pm (10m)  

Invited Speakers

Rumen Illiev, Toyota Research Institute

Bio

Rumen Iliev is a Staff Research Scientist in the Human-Centered AI division at the Toyota Research Institute. His work is focused on understanding and modeling human behavior in the broader context of carbon neutrality. Rumen has a Ph.D. degree in Cognitive Psychology from Northwestern University. Before joining TRI, he was a postdoctoral fellow at the Ford School of Public Policy at the University of Michigan, a data analyst at PERTS/Stanford University and a data scientist at Uber.

Title

Light Duty Vehicles Decarbonisation: A Causal Modeling Perspective

Abstract

One of the major problems that humanity faces in the decades to come is climate change. Transportation is a leading source of GHG emissions and light duty vehicles are associated with a large portion of those emissions. Accordingly, many of the efforts to achieve carbon neutrality are focused on decarbonizing the production and usage of light duty vehicles. Deciding on what changes to make, however, is not a trivial task, since there are multiple possible interventions and in many cases the effects of those interventions are not straightforward to predict. Furthermore, different solutions might fit different contexts. In this talk I will describe the development of a causality-focused decision support system which will allow decision makers to compare a wide range of “what-if” scenarios. Causal modeling provides a natural framework for comparing the effects of different hypothetical interventions, and combining it with other modeling methods, such as supply chain modeling and lifetime analysis, can help us reach better decisions on essential environmental questions.

Ruocheng Guo, ByteDance Research

Bio

Ruocheng Guo is a senior machine learning researcher at ByteDance Research, London, UK. Previously, he served as an Assistant Professor in Data Science at City University of Hong Kong. Guo’s research focuses on causal inference, conformal prediction, and responsible AI with applications to recommendation systems, graph mining, and large language models. With over 60 published papers and more than 2700 citations on Google Scholar, Guo’s work has been featured in top AI venues such as NeurIPS, ICLR, KDD, and The Web Conference etc. Guo also contributes to the AI research community by serving as a program committee member for prestigious conferences including NeurIPS, ICML, ICLR, SIGKDD, WWW, WSDM, AAAI, UAI, CLeaR, AISTATS, ACL, and EMNLP, and as a reviewer for journals such as TKDE, TKDD, TIST, and TMLR.

Title

Conformal Prediction: a new way to understand and exploit uncertainty in counterfactual inference and recommendation

Abstract

Point estimate is often not enough for decision-making. With conformal prediction, predictive intervals can be inferred in a distribution-free fashion. In contrast to traditional methods that rely on strong assumptions on the data distribution. For conformal counterfactual inference, existing methods that construct confidence intervals for counterfactuals either rely on strong ignorability, or access to un-identifiable bounds calibrating the difference between observational and interventional data. To overcome these limitations, we propose a novel approach to provide predictive intervals for counterfactual outcomes with marginal coverage guarantees, even under hidden confounding. With less restrictive assumptions, our approach uses a fraction of interventional data to account for the covariate shift from observational to interventional data. Theoretical results demonstrate the conditions under which our method is strictly advantageous to the naive method that only uses interventional data. Experiments across synthetic and real-world data verifies the superiority of our methods compared with state-of-the-art baselines. In addition, we developed novel conformal prediction methods for recommendation systems: the first inference-time stochastic ranking method for fairness utility trade-off and a novel method for fine-tuning pre-trained sequential recommenders.

Accepted Papers

For Oral Presentation

  1. From Causality to Incrementality: Estimating Marketplace Inter-product Cannibalisation Effects at Expedia Group; Florian Florian*, Catriona Mitchison (Expedia); Paper
  2. DISCO: constrained bandits for personalized discount allocation within fashion e-commerce; Jason Zhang (TripAdvisor), Benjamin M Howson (Imperial College London), Panayiota Savva, Eleanor Loh (ASOS)*; Paper
  3. Industrial-Grade Smart Troubleshooting through Causal Technical Language Processing: a Proof of Concept; Alexandre Trilla*, Ossee Yiboe, Nenad Mijatovic (Alstom), Jordi Vitria (Universitat de Barcelona); Paper
  4. Practical Marketplace Optimization at Uber Using Causally-Informed Machine Learning; Zijun Qin*, Siyu Chen, Jason Dowlatabadi,Yu Xuan Hong, Vinayak Iyer, Uday Mantripragada, Rishabh Narang, Apoorv Pandey, Abrar Sheikh, Jiaqi Sun, Matthew Walker, Kaichen Wei, Chen Xu, Jingnan Yang, Allen Zhang, Guoqing Zhang, Bobby Chen, Hongtao Sun (Uber); Paper
  5. Enhancing Uplift Modeling in Multi-Treatment Marketing Campaigns: Leveraging Score Ranking and Calibration Techniques; Yoon Tae Park*, Ting Xu (Bestbuy); Paper
  6. Optimization and Control Applications through Repeated Interventions Based on Structural Causal Models; Daigo Fujiwara*, Tomonori Izumitani (NTT), Shohei Shimizu (Shiga University & RIKEN); Paper
  7. Understanding Product Quality with Unstructured Data: An Application of LLMs and Embeddings at Airbnb; Zhiying Gu*,Jacob Zhu, Brad Li, Linsha Chen (Airbnb); Paper
  8. Sensitivity Analysis for Causal ML: A Use Case at Booking.com; Philipp Bach (University of Hamburg), Victor Chernozhukov (MIT), Carlos Cinelli (University of Washington), Lin Jia (Booking.com), Sven Klaassen (University of Hamburg), Nils Skotara (Booking.com), Martin Spindler (University of Hamburg); Paper

For Poster Presentation

  1. Shipping Speed Elasticity Estimation Using Causal Inference Machine Learning Techniques; Aparupa Das Gupta*, Ajinkya More, Weijie Yuan, Ajay Kumar (Walmart); Paper, Poster
  2. Understanding guest demand and optimizing two-sided marketplaces: insights from Airbnb; Yufei Wu*, Daniel Schmierer (Airbnb); Paper
  3. Using LLMs for Explaining Sets of Counterfactual Examples to Final Users; Arturo Fredes*, Jordi Vitria (Universitat de Barcelona); Paper
  4. CHEF: Causal Effect Estimation under Hidden Confounding Using an End-to-End Framework with Data Fusion; Chuan Zhou (Peking University)*, Yaxuan Li (Harbin Institute of Technology), Chunyuan Zheng (UCSD), Haiteng Zhang (University of Chinese Academic of Sciences), Haoxuan Li (Peking University), Mingming Gong (University of Melbourne)

Organizers

  • Hasta Vanchinathan, Amazon
  • Jonathan Grotts, Snap - main contact on the day of the workshop
  • Yingfei Wang, University of Washington
  • Zeyu Zheng, UC Berkeley, Amazon - main contact

The CausalML Team

  • Huigang Chen, Meta
  • Jeong-Yoon Lee, Uber - main contact
  • Jing Pan, Snap
  • Paul Lo, Snap
  • Roland Stevenson, Consultant - main contact on the day of the workshop
  • Totte Harinen, AirBnB
  • Yifeng Wu, Uber
  • Zhenyu Zhao, Roblox

Past Workshop