Berk Ustun

Interests

Areas: Machine Learning, Optimization, Human-Centered Design

Topics: Safety, Interpretability, Algorithmic Fairness, Personalization, Governance

Applications: Medicine, Consumer Finance, Criminal Justice, Physical Sciences


Recent Papers

Prediction without Preclusion: Recourse Verification with Reachable Sets
Avni Kothari, Bogdan Kulynych, Lily Weng, Berk Ustun
ICLR • International Conference on Learning Representations, 2024

Spotlight Paper at ICLR 2024

PDF arXiv Code Reviews
Classification with Conceptual Safeguards
Hailey Joren, Charles T Marx, Berk Ustun
ICLR • International Conference on Learning Representations, 2024
PDF Code Reviews
Operationalizing the Search for Less Discriminatory Alternatives in Fair Lending
Talia Gillis, Vitaly Mersault, Berk Ustun
FAccT • ACM Conference on Fairness, Accountability, and Transparency, 2024
PDF Code Publisher’s Version
Participatory Systems for Classification
Hailey Joren, Chirag Nagpal, Katherine Heller, Berk Ustun
NeurIPS • Neural Information Processing Systems, 2023

Spotlight Paper at NeurIPS 2023

Oral Presentation at NeurIPS Workshop for Participatory Approaches for Mental Health

PDF arXiv
When Personalization Harms Performance: Reconsidering the Use of Group Attributes in Prediction
Vinith M. Suriyakumar, Marzyeh Ghassemi, Berk Ustun
ICML • International Conference on Machine Learning, 2023

Oral Presentation at ICML 2023

PDF arXiv
Predictive Multiplicity in Probabilistic Classification
Jamelle Watson-Daniels, David Parkes, Berk Ustun
AAAI • Association for Advancement in Artificial Intelligence, 2023

Oral Presentation at AAAI 2023

PDF arXiv
Optimized Risk Score to Predict Mortality in Patients with Cardiogenic Shock in the ICU
Eric Yamga, Sreekar Mantena, Darin Rosen, Emily Bucholz, Robert Yeh, Leo Celi, Berk Ustun, Neel Butala
Journal of the American Heart Association, 2023
PDF Publisher’s Version

Machine Learning Methods

Providing Fair Recourse over Plausible Groups
Jayanth Yetukuri, Ian Hardy, Yevgeniy Vorobeychik, Berk Ustun, Yang Liu
AAAI • Association for the Advancement of Artificial Intelligence, 2024
PDF Publisher’s Version
Algorithmic Censoring in Dynamic Learning Systems
Jennifer Chien, Margaret Roberts, Berk Ustun
EAAMO • ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization, 2023
PDF Publisher’s Version
On the Epistemic Limits of Personalized Prediction
Lucas Monteiro Paes, Carol Long, Berk Ustun, Flavio Calmon
NeurIPS • Neural Information Processing Systems, 2022
PDF Publisher’s Version
Rank List Sensitivity of Recommender Systems to Interaction Perturbations
Sejoon Oh, Berk Ustun, Julian McAuley, Srijan Kumar
CIKM • Conference on Information and Knowledge Management, 2022
PDF arXiv
Learning Optimal Predictive Checklists
Haoran Zhang, Quaid Morris, Berk Ustun, Marzyeh Ghassemi
NeurIPS • Neural Information Processing Systems, 2020

Spotlight Presentation at 2022 ICML Workshop for Interpretable ML in Healthcare

PDF arXiv Publisher’s Version Code
Predictive Multiplicity in Classification
Charles Marx, Flavio du Pin Calmon, Berk Ustun
ICML • International Conference on Machine Learning, 2020

Oral Presentation at 2019 NeurIPS Workshop for Human-Centric ML

PDF arXiv Publisher’s Version Code
Actionable Recourse in Linear Classification
Berk Ustun, Alexander Spangher, Yang Liu
FAT* • ACM Conference on Fairness, Accountability and Transparency, 2019
PDF arXiv Publisher’s Version Code Video @ FAT19 Workshop Paper @ FAT/ML 18 WIRED Article
Learning Optimized Risk Scores
Berk Ustun, Cynthia Rudin
JMLR • Journal of Machine Learning Research, 2019

Accompanies Winning Entry for 2019 INFORMS Innovative Applications in Analytics Award

PDF arXiv Publisher’s Version Code Short Video
Fairness without Harm: Decoupled Classifiers with Preference Guarantees
Berk Ustun, Yang Liu, David Parkes
ICML • International Conference on Machine Learning, 2019
PDF Publisher’s Version Code Video @ ICML19
Repairing without Retraining: Avoiding Disparate Impact with Counterfactual Distributions
Hao Wang, Berk Ustun, Flavio du Pin Calmon
ICML • International Conference on Machine Learning, 2019
PDF arXiv Publisher’s Version Code Video @ ICML19
The Direction of Discrimination: An Information-Theoretic Analysis of Disparate Impact
Hao Wang, Berk Ustun, Flavio du Pin Calmon
ISIT • IEEE International Symposium on Information Theory, 2018
PDF arXiv Code Article in Colloquy
Optimized Risk Scores
Berk Ustun, Cynthia Rudin
KDD • Knowledge Discovery and Data Mining, 2017

Winner of the 2017 INFORMS Computing Society Best Student Paper

PDF Publisher’s Version Code Video
Supersparse Linear Integer Models for Optimized Medical Scoring Systems
Berk Ustun, Cynthia Rudin
MLJ • Machine Learning, 2015

Accompanies Winning Entry for 2016 INFORMS Innovative Applications in Analytics Award

PDF arXiv Publisher’s Version Code Code (matlab)
Supersparse Linear Integer Models for Predictive Scoring Systems
Berk Ustun, Stefano Tracà, Cynthia Rudin
AAAI Late Breaking Track, 2013
PDF arXiv Publisher’s Version
Importance Sampling in Stochastic Programming: An MCMC Approach
Panos Parpas, Berk Ustun, Mort Webster, Quang Kha Tran
INFORMS Journal of Computing, 2015
PDF Publisher’s Version

Applications

Optimized Short-Forms of the Cognitive Distortions Questionnaire
Amanda Morrison, Berk Ustun, Arielle Horenstein, Simona Kaplan, Irismar Reis de Oliveira, Sedat Batmaz, James Gross, Ekaterina Sadikova, Curt Hemanny, Pedro Pires, Philippe Goldin, Ronald Kessler, Richard G. Heimberg
Journal of Anxiety Disorders, 2022
PDF Publisher’s Version
Developing an Optimal Short-Form of the PTSD Checklist for DSM-5
Kelly Zuromski, Berk Ustun, Irving Hwang, Terence Keane, Brian Marx, Murray Stein, Robert Ursano, Ronald Kessler
Depression and Anxiety, 2019
PDF Publisher’s Version
Optimized Scoring Systems: Towards Trust in ML for Healthcare and Criminal Justice
Cynthia Rudin, Berk Ustun
Interfaces, 2018

Finalist for the 2017 Wagner Prize for Excellence in Operations Research Practice

PDF Publisher’s Version Presentation Video
The World Health Organization Adult ADHD Self-Report Scale Screening Scale for DSM-5
Berk Ustun, Lenard Adler, Cynthia Rudin, Stephen Faraone, Thomas Spencer, Patricia Berglund, Michael Gruber, Ronald Kessler
JAMA Psychiatry, 2017
PDF Publisher’s Version JAMA Editorial NPR Article
Association of an EEG-Based Risk Score With Seizure Probability in Hospitalized Patients
Aaron Struck, Berk Ustun, Andres Rodriguez Ruiz, Jong Woo Lee, Suzette LaRoche, Lawrence J. Hirsch, Emily J. Gilmore, Jan Vlachy, Hiba Arif Haider, Cynthia Rudin, M. Brandon Westover
JAMA Neurology, 2017

Accompanies Winning Entry for 2019 INFORMS Innovative Applications in Analytics Award

PDF Publisher’s Version JAMA Editorial Article in Duke News
Interpretable Classification Models for Recidivism Prediction
Jiaming Zeng, Berk Ustun, Cynthia Rudin
Journal of the Royal Statistical Society: Series A, 2016

Winner of the 2016 USRESP Competition

PDF arXiv Publisher’s Version Code
Clinical Prediction Models for Sleep Apnea: Superiority of Medical History over Symptoms
Berk Ustun, M. Brandon Westover, Cynthia Rudin, Matt Bianchi
Journal of Clinical Sleep Medicine, 2015
PDF Publisher’s Version JCSM Editorial

Theses

Simple Linear Classifiers via Discrete Optimization
Berk Ustun
PhD thesis. Massachusetts Institute of Technology, 2017
The MCMC Approach to Importance Sampling in Stochastic Programming
Berk Ustun
Master's thesis. Massachusetts Institute of Technology, 2012