About

Phebe Vayanos is an Associate Professor of Industrial & Systems Engineering and Computer Science at the University of Southern California. She holds the Andrew and Erna Viterbi Early Career Chair in Engineering and is a Co-Director of the CAIS Center for Artificial Intelligence in Society as well as the ORAI interdisciplinary PhD Program in Artificial Intelligence and Operations Research. Her academic background includes a PhD in Computing (Operations Research) and an MEng in Electrical & Electronic Engineering, both from Imperial College London. Prior to USC, she was a lecturer at MIT Sloan School of Management and a postdoctoral research associate at MIT's Operations Research Center. Her research focuses on advancing integer, stochastic, and robust optimization, and their interface with machine learning, causal inference, and economics. She aims to develop predictive and prescriptive models that are robust, interpretable, and fair, particularly for deployment in high-stakes settings. Her work is driven by a desire to make a positive societal impact, especially supporting underserved and marginalized communities. Vayanos has received numerous awards including the NSF CAREER award, the Imperial College Emerging Alumni Leader Award, and the USC Viterbi Junior Research Award. She is a TED Speaker and has participated in the Time 100 Impact Dinner for Extraordinary Women Shaping the Future of AI. In addition to her research, she serves as an Associate Editor for several prominent journals and is actively involved in professional societies, currently serving as Chair of the Committee on Stochastic Programming within the Stochastic Programming Society. Her research is supported by various organizations including the National Science Foundation, the U.S. Army Research Laboratory, and several foundations dedicated to social impact and diversity. Her students have earned prestigious awards, and she has been recognized for her contributions to diversity, equity, and inclusion in her field.

Research topics

• Computer Science
• Political Science
• Machine Learning
• Law
• Theoretical computer science
• Business
• Transport engineering
• Risk analysis (engineering)
• Mathematics
• Engineering
• Psychology
• Mathematical optimization
• Microeconomics
• Economics
• Operations research

Selected publications

• Distributionally robust optimization with decision-dependent information discovery

  Mathematical Programming · 2026-04-20

  preprintOpen access
  PublisherOA PDFDOI

• Robust Offline Policy Learning Under Distribution Shifts with Application to Homeless Services Delivery

  Open MIND · 2026-01-01

  otherOpen access1st authorCorresponding
  OA PDFDOI

• Couch-Surfing and HIV Risk Behavior Among Young Adults Experiencing Homelessness

  AIDS and Behavior · 2026-01-27

  articleOpen access

  Young adults experiencing homelessness (YAEH) are significantly more likely to engage in HIV risk-related sexual behavior relative to their stably housed peers, and their social support networks can influence their engagement in these behaviors. However, few studies have investigated HIV risk behaviors among YAEH who are "couch-surfing," a highly prevalent network-based survival strategy that involves cycling through temporary, informal housing arrangements. The current study utilizes survey data collected from 461 YAEH accessing drop-in center services in Los Angeles, California, between September 2016 and October 2018. Egocentric network analysis was used to examine associations among couch-surfing, sources of social support, and HIV risk and prevention behaviors. The potential moderating effect of social support on the relationship between couch-surfing and specific sexual risk behaviors was also tested. Compared to street- and shelter-based youth, couch-surfing YAEH reported the highest rates of recent transactional sex (18.0%) and concurrent or serial sexual partners (38.2%). Relative to residing in emergency shelter or transitional housing programs, couch-surfing was associated with over twice the odds of engaging in recent transactional sex (OR = 2.52, p = .023, 95% CI 1.13-5.62)-as was living unsheltered (OR = 2.06, p = .029, 95% CI 1.08-3.95). While social support was individually associated with several HIV risk-related sexual behaviors, its effect was ultimately eclipsed by homeless situation in the final model. Findings underscore the need for individual- and structural-level interventions that attend to the unique socioenvironmental contexts of couch-surfing YAEH.

  PublisherOA PDFDOI

• Toward AI Matching Policies in Homeless Services: A Qualitative Study with Policymakers

  ArXiv.org · 2025-08-10

  preprintOpen accessSenior author

  Artificial intelligence researchers have proposed various data-driven algorithms to improve the processes that match individuals experiencing homelessness to scarce housing resources. It remains unclear whether and how these algorithms are received or adopted by practitioners and what their corresponding consequences are. Through semi-structured interviews with 13 policymakers in homeless services in Los Angeles, we investigate whether such change-makers are open to the idea of integrating AI into the housing resource matching process, identifying where they see potential gains and drawbacks from such a system in issues of efficiency, fairness, and transparency. Our qualitative analysis indicates that, even when aware of various complicating factors, policymakers welcome the idea of an AI matching tool if thoughtfully designed and used in tandem with human decision-makers. Though there is no consensus as to the exact design of such an AI system, insights from policymakers raise open questions and design considerations that can be enlightening for future researchers and practitioners who aim to build responsible algorithmic systems to support decision-making in low-resource scenarios.

  PublisherOA PDFDOI

• Responsible Machine Learning via Mixed-Integer Optimization

  2025-10-01

  book-chapterSenior author
  PublisherDOI

• Toward AI Matching Policies in Homeless Services: A Qualitative Study with Policymakers

  Proceedings of the AAAI/ACM Conference on AI Ethics and Society · 2025-10-15

  articleOpen accessSenior author

  Artificial intelligence researchers have proposed various data-driven algorithms to improve the processes that match individuals experiencing homelessness to scarce housing resources. It remains unclear whether and how these algorithms are received or adopted by practitioners and what their corresponding consequences are. Through semi-structured interviews with 13 policymakers in homeless services in Los Angeles, we investigate whether such change-makers are open to the idea of integrating AI into the housing resource matching process, identifying where they see potential gains and drawbacks from such a system in issues of efficiency, fairness, and transparency. Our qualitative analysis indicates that, even when aware of various complicating factors, policymakers welcome the idea of an AI matching tool if thoughtfully designed and used in tandem with human decision-makers. Though there is no consensus as to the exact design of such an AI system, insights from policymakers raise open questions and design considerations that can be enlightening for future researchers and practitioners who aim to build responsible algorithmic systems to support decision-making in low-resource scenarios.

  PublisherOA PDFDOI

• Robust Optimization with Decision-Dependent Information Discovery

  Management Science · 2025-06-20 · 23 citations

  articleOpen access1st authorCorresponding

  Robust optimization (RO) is a popular paradigm for modeling and solving two- and multistage decision-making problems affected by uncertainty. In many real-world applications, such as R&D project selection, production planning, or preference elicitation for product or policy recommendations, the time of information discovery is decision-dependent and the uncertain parameters only become observable after an often costly investment. Yet, most of the literature on robust optimization assumes that the uncertain parameters can be observed for free and that the sequence in which they are revealed is independent of the decision-maker’s actions. To fill this gap in the practicability of RO, we consider two- and multistage robust optimization problems in which part of the decision variables control the time of information discovery. Thus, information available at any given time is decision-dependent and can be discovered (at least in part) by making strategic exploratory investments in previous stages. We propose a novel dynamic formulation of the problem and prove its correctness. We leverage our model to provide a solution method inspired from the K-adaptability approximation, whereby K candidate strategies for each decision stage are chosen here-and-now and, at the beginning of each period, the best of these strategies is selected after the uncertain parameters that were chosen to be observed are revealed. We reformulate the problem as a finite mixed-integer (resp. bilinear) program if none (resp. some) of the decision variables are real-valued. This finite program is solvable with off-the-shelf solvers. We generalize our approach to the minimization of piecewise linear convex functions. We demonstrate the effectiveness of our method in terms of usability, optimality, and speed on synthetic instances of the Pandora box problem, the preference elicitation problem with real-valued recommendations, the best box problem, and the R&D project portfolio optimization problem. Finally, we evaluate it on an instance of the active preference elicitation problem used to recommend kidney allocation policies to policy-makers at the United Network for Organ Sharing based on real data from the U.S. Kidney Allocation System. This paper was accepted by Chung Piaw Teo, optimization. Funding: This work was supported primarily by the Operations Engineering Program of the National Science Foundation under NSF Award No. 1763108. The authors are grateful for this support. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2021.00160 .

  PublisherOA PDFDOI

• Strong Optimal Classification Trees

  Operations Research · 2024-07-31 · 20 citations

  articleSenior author

  Decision trees are among the most interpretable and popular machine learning models that are used routinely in applications ranging from revenue management to medicine. Traditional heuristic methods, although fast, lack modeling flexibility for incorporating constraints such as fairness and do not guarantee optimality. Recent efforts aim to overcome these limitations using mixed-integer optimization (MIO) for better modeling flexibility and optimality, but speed remains an issue. In “Strong Optimal Classification Trees,” Aghaei, Gómez, and Vayanos use integer optimization and polyhedral theory to create an MIO-based formulation with a strong LO relaxation resulting in a 29% speed-up in training time compared with state-of-the-art MIO-based formulations, as well as up to an 8% improvement in out-of-sample accuracy.

  PublisherDOI

• Learning Fair Policies for Multi-Stage Selection Problems from Observational Data

  Proceedings of the AAAI Conference on Artificial Intelligence · 2024-03-24

  articleOpen access

  We consider the problem of learning fair policies for multi-stage selection problems from observational data. This problem arises in several high-stakes domains such as company hiring, loan approval, or bail decisions where outcomes (e.g., career success, loan repayment, recidivism) are only observed for those selected. We propose a multi-stage framework that can be augmented with various fairness constraints, such as demographic parity or equal opportunity. This problem is a highly intractable infinite chance-constrained program involving the unknown joint distribution of covariates and outcomes. Motivated by the potential impact of selection decisions on people’s lives and livelihoods, we propose to focus on interpretable linear selection rules. Leveraging tools from causal inference and sample average approximation, we obtain an asymptotically consistent solution to this selection problem by solving a mixed binary conic optimization problem, which can be solved using standard off-the-shelf solvers. We conduct extensive computational experiments on a variety of datasets adapted from the UCI repository on which we show that our proposed approaches can achieve an 11.6% improvement in precision and a 38% reduction in the measure of unfairness compared to the existing selection policy.

  PublisherOA PDFDOI

• Learning Optimal Classification Trees Robust to Distribution Shifts

  arXiv (Cornell University) · 2023-10-26 · 1 citations

  preprintOpen accessSenior author

  We consider the problem of learning classification trees that are robust to distribution shifts between training and testing/deployment data. This problem arises frequently in high stakes settings such as public health and social work where data is often collected using self-reported surveys which are highly sensitive to e.g., the framing of the questions, the time when and place where the survey is conducted, and the level of comfort the interviewee has in sharing information with the interviewer. We propose a method for learning optimal robust classification trees based on mixed-integer robust optimization technology. In particular, we demonstrate that the problem of learning an optimal robust tree can be cast as a single-stage mixed-integer robust optimization problem with a highly nonlinear and discontinuous objective. We reformulate this problem equivalently as a two-stage linear robust optimization problem for which we devise a tailored solution procedure based on constraint generation. We evaluate the performance of our approach on numerous publicly available datasets, and compare the performance to a regularized, non-robust optimal tree. We show an increase of up to 12.48% in worst-case accuracy and of up to 4.85% in average-case accuracy across several datasets and distribution shifts from using our robust solution in comparison to the non-robust one.

  PublisherOA PDFDOI

Frequent coauthors

• Milind Tambe

  49 shared

• Eric Rice

  University of Southern California

  39 shared

• Aida Rahmattalabi

  28 shared

• Sina Aghaei

  University of Southern California

  26 shared

• Mohammad Javad Azizi

  Sahand University of Technology

  10 shared

• Nathanael Jo

  Massachusetts Institute of Technology

  9 shared

• Andrés Gómez

  University of Southern California

  8 shared

• Sara Marie Mc Carthy

  Southern California University for Professional Studies

  8 shared

Labs

• Phebe Vayanos LabPI

Education

• Ph.D., Computer Science

  University of Southern California

  2009

• M.S., Computer Science

  University of Southern California

  2005

• B.S., Computer Science

  University of Southern California

  2003

Awards & honors

• NSF CAREER award (2021)
• Imperial College Emerging Alumni Leader Award
• USC Viterbi Junior Research Award (2022)
• INFORMS Diversity, Equity, and Inclusion Ambassador Program…