I’m a CS PhD student at MIT, advised by David Sontag. My research lies at the intersection of causality and machine learning, and is generally motivated by technical challenges that arise when applying machine learning to healthcare data. During my PhD, I’ve focused on a few complementary themes:

  • Robustness to distribution shift: The performance of a model learned in one environment (e.g., a particular hospital) often degrades when used in other contexts. However, we often have some sense for the nature and degree of plausible shifts (e.g., differences in certain unobserved factors, like social determinants of health). This has motivated my recent work in learning models that maintain performance across changes in unobserved factors, when noisy proxies are available [ICML 2021].
  • “Debugging” causal models: Retrospective healthcare data is often used to learn better policies for treating disease, when experimentation is infeasible: However, this requires strong causal assumptions, and not all policies can be reliably evaluated. This has motivated my work on developing methods to help domain experts assess the plausibility of causal models [ICML 2019, MS Thesis], and get interpretable characterization of subpopulations where a given policy can be evaluated [AISTATS 2020].

These methodological problems are informed by my applied work with clinical collaborators, such as learning antibiotic treatment policies [Science Trans. Med. 2020] and debugging reinforcement-learning algorithms for sepsis management [AMIA 2021].

Conference Papers

Finding Regions of Heterogeneity in Decision-Making via Expected Conditional Covariance
Justin Lim*, Christina X. Ji*, Michael Oberst*, Saul Blecker, Leora Horwitz, David Sontag
Neural Information Processing Systems (NeurIPS), 2021
[paper], [video], [code]

Regularizing towards Causal Invariance: Linear Models with Proxies
Michael Oberst, Nikolaj Thams, Jonas Peters, David Sontag
International Conference on Machine Learning (ICML), 2021
[paper], [video], [slides], [poster], [code]

Trajectory Inspection: A Method for Iterative Clinician-Driven Design of Reinforcement Learning Studies
Christina X. Ji*, Michael Oberst*, Sanjat Kanjilal, David Sontag
AMIA Virtual Informatics Summit, 2021
[paper], [video], [code]

Characterization of Overlap in Observational Studies
Michael Oberst*, Fredrik D. Johansson*, Dennis Wei*, Tian Gao, Gabriel Brat, David Sontag, Kush R. Varshney
International Conference on Artificial Intelligence and Statistics (AISTATS), 2020
[paper], [video], [code]

Treatment Policy Learning in Multiobjective Settings with Fully Observed Outcomes
Sooraj Boominathan, Michael Oberst, Helen Zhou, Sanjat Kanjilal, David Sontag
ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2020
[paper], [video], [code], [dataset]

Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models
Michael Oberst, David Sontag
International Conference on Machine Learning (ICML), 2019
[paper], [slides], [poster], [video]

Journal Papers

A decision algorithm to promote outpatient antimicrobial stewardship for uncomplicated urinary tract infection
Sanjat Kanjilal, Michael Oberst, Sooraj Boominathan, Helen Zhou, David C. Hooper, David Sontag
Science Translational Medicine, 2020
[article], [code], [dataset]

Predicting Human Health from Biofluid-Based Metabolomics using Machine Learning
Ethan D. Evans, Claire Duvallet, Nathaniel D. Chu, Michael Oberst, Michael A.Murphy, Isaac Rockafellow, David Sontag, Eric J. Alm
Scientific Reports, 2020


AMR-UTI: Antimicrobial Resistance in Urinary Tract Infections
Michael Oberst, Sooraj Boominathan, Helen Zhou, Sanjat Kanjilal, David Sontag
[Homepage], [PhysioNet Page]

Invited Talks

Regularizing towards Causal Invariance: Linear Models with Proxies
Online Causal Inference Seminar
Stanford, March 29th, 2022
[video], [slides]

Primer: Learning Treatment Policies from Observational Data
Models, Inference, and Algorithms Seminar
Broad Institute, September 23rd, 2020
[video], [slides]

Other Publications

Counterfactual Policy Introspection using Structural Causal Models
Michael Oberst
M.S. Thesis in EECS, MIT
[full text], [note on errata]