NETS 7976 - Causal Inference with Network Interference - Fall 2024

Tuesdays: 2:30 – 4:30pm
September 5 – December 10, 2024
177 Huntington, room 1008

Summary

This class is an introduction to the measurement of causality in complex networks. It builds on the fundamental foundations of causal inference, drawing on techniques in statistics and econometrics and applying them to a network setting. These techniques are essential in today's scientific landscape, where the study of large systems is best modeled as complex networks. Furthermore, rigorous methods to establish causality in networks have critical implications for the design of interventions in complex systems. Students will leave the class with a foundational understanding of causal inference in networks and a survey of the state of the art in the field. Particular emphasis will be placed on epistemic uncertainty about the structure of interference and the sensitivity of existing techniques to these choices. Students completing the course should feel confident approaching the forefront of research in the field and applying these techniques to their own work.

This class is designed in collaboration with Erik Weis, a PhD student in Network Science; it draws from a previous Network Science Special Topics course taught in 2020 by Prof. Donghee Jo, NETS 7983 -- Causal Analysis: Emphasis on Network Settings.

PDF of this syllabus: here.

Coursework, Class Structure, Grading

The course adopts a flipped classroom concept: Students are exposed to the relevant material through readings and outside lectures/assignments. The in-class hours are dedicated to recapitulation of the most important points (via student-prepared presentations), clarification of questions, and open discussion. The content of the course is organized as follows. Weeks 1-5 are dedicated to an in-depth review of the fundamentals of causal inference. Weeks 6-10 are dedicated to a recapitulation of the ideas in the first phase but in the context of networks. Weeks 11-14 cover special topics and recent research, as well as completion of a project. The final class will be presentations of final projects. Evaluation is as follows:

• Active participation in weekly meetings and completion of weekly deliverables (70%): Students produce a deliverable for each class (e.g. a presentation, code, or writeup), designed to demonstrate honest effort and engagement with the material. Students should also actively participate in discussions by posing relevant questions, attempting to answer other students' questions, and connecting the material to real-world problems, e.g. from their own field of research.

• Final project (30%): The final project is to produce a full research proposal for a causal study in which network spillover effects play a central role. Students should draw on the toolbox of methods discussed throughout the course to design a methodology suited to answer the question. Students should clearly specify the assumptions made by their chosen method and justify their use in the particular application. This justification should include an extensive literature review. Finally, they should approach the inference problem pragmatically by considering the feasibility of any additional data collection or experimentation required by the proposed method. Students will give a presentation on the last day of class and submit a final report.

Learning Objectives and Outcomes

By the end of this course, students should have a deep familiarity with causal inference techniques in both network and non-network settings, including parametric and nonparametric techniques. From this, students will be able to articulate the assumptions in a wide variety of causal inference techniques and identify when these assumptions are appropriate for the problem at hand. Finally, students will be able to approach the forefront of causal inference on networks, with particular emphasis on the intersection of causal discovery and estimating unknown interference. Students should leave the course with the necessary knowledge to apply causal inference to their own network datasets and problems. The course is designed to survey the leading edge of the causal inference field, while being:

• Current: While the courses begin with foundational concepts in causal inference, the final portion of the course will provide exposure to the latest research on causal inference in networks.

• Practical: Coding assignments lead students to develop hands-on experience with causal inference techniques by using existing software to reproduce existing results.

• Actionable: The final project allows students to apply their foundational understanding by tailoring the various techniques at their disposal to fit a particular problem.

Materials

There are no required materials to purchase for this course, but we rely on a range of published work on networks and causation, as well as basic textbooks on causal inference:

• Scott Cunningham. Causal Inference: The Mixtape. Yale University Press, 2021.

• Aleksander Molak. Causal Inference and Discovery in Python: Unlock the secrets of modern causal machine learning with DoWhy, EconML, PyTorch and more. Packt Publishing Ltd, 2023.

• Miguel A. Hernan and James M. Robins. Causal Inference: What If. CRC Press, 2024.

• Judea Pearl. Causality. 2nd ed. Cambridge University Press, 2009. isbn: 978-0-521-89560-6.

• Guido W. Imbens and Donald B. Rubin. Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction. Cambridge University Press, 2015. isbn: 978-0-521-88588-1.

Instructor

Brennan Klein is an associate research scientist at the Network Science Institute, with a joint affiliation at the Institute for Experiential AI. He is the director of the Complexity & Society Lab. His research spans two broad topics: 1) Information, emergence, and inference in complex systems — developing tools and theory for characterizing dynamics, structure, and scale in networks, and 2) Public health and public safety — creating and analyzing large scale datasets that reveal inequalities in the United States, from epidemics to mass incarceration. Dr. Klein received a PhD in Network Science in 2020 from Northeastern University and got his BA in Cognitive Science & Psychology from Swarthmore College in 2014. Website: http://brennanklein.com/.

Week 1: Sep 5

Review: Probabilistic Modeling & Representations of Causality

Topics:

• Chapter 1 - Causal Inference Mixtape: Introduction

• Chapter 2 - Causal Inference Mixtape: Probability and Regression Review

• Chapter 3 - Causal Inference Mixtape: Directed Acyclic Graphs

• Chapter 4 - Causal Inference Mixtape: Potential Outcomes Causal Model

Supplementary Materials:

• Miguel A. Hernan and James M. Robins. Causal Inference: What If. CRC Press, 2024 (Chp. 1, 2, 3, 9)

• NETS 7983 Slides - Week 1: Course Introduction and Basics of Causal Analysis

• NETS 7938 Slides - Week 2: Applications: Network-level RCTs, DAGs, & do-calculus

Week 2: Sep 10

Matching Methods

Topics:

• Chapter 5 - Causal Inference Mixtape: Matching and Subclassification

• Chapter 6 - Causal Inference Mixtape: Regression Discontinuity

Supplementary Materials:

• Miguel A. Hernan and James M. Robins. Causal Inference: What If. CRC Press, 2024 (Chp. 4, 15)

• NETS 7983 Slides - Week 3: DAGs Part II & Mediation Analysis

Week 3: Sep 17

Instrumental Variables and Panel Data

Topics:

• Chapter 7 - Causal Inference Mixtape: Instrumental Variables

• Chapter 8 - Causal Inference Mixtape: Panel Data

Supplementary Materials:

• NETS 7983 Slides - Week 4: Mediation Analysis Part II & Instrumental Variables

• Miguel A. Hernan and James M. Robins. Causal Inference: What If. CRC Press, 2024 (Chp. 16)