Amanda Coston

Amanda Coston is an Assistant Professor in the Department of Statistics at the University of California, Berkeley. Her research focuses on addressing challenges in algorithmic decision support systems and data-driven policy-making, emphasizing equity, validity, and reliability. She earned her PhD in Machine Learning and Public Policy from Carnegie Mellon University, advised by Alexandra Chouldechova and Edward H. Kennedy, and completed a postdoc at Microsoft Research's Machine Learning and Statistics Team. Her work spans causal inference, machine learning, and nonparametric statistics, with applications in criminal justice, healthcare, and public policy.

Education:

• PhD in Machine Learning and Public Policy, Carnegie Mellon University (2019-2022)
• MS in Machine Learning, Carnegie Mellon University (2019)
• Bachelor of Science in Computer Science, Princeton University (2013)

Research Interests:

Her research investigates how algorithms and data systems can perpetuate or mitigate disparities in high-stakes domains. Key areas include counterfactual audits of racial bias in policing, fairness in predictive models, and validity in algorithmic decision-making. She develops methodologies to ensure equitable outcomes in applications like healthcare resource allocation and criminal justice risk assessments.

Awards & Honors:

• 2024 Schmidt Sciences AI 2050 Early Career Fellowship
• 2023 FAccT Best Paper Award (Counterfactual Prediction Under Outcome Measurement Error)
• 2023 SaTML Best Paper Award (A Validity Perspective on Evaluating the Justified Use of Algorithms)
• 2022 Meta Research PhD Fellowship

Teaching & Mentorship:

She teaches Causal Inference (STAT 156/256) at Berkeley and has mentored students through programs like AI4ALL. Her teaching emphasizes ethics, fairness, and societal impacts of AI.

Service:

• Referee for journals including Nature Human Behaviour, JASA, and Transactions on Machine Learning Research
• Steering Committee Member for ML4D Workshop (NeurIPS)
• Program Committee Member for FAccT and AAAI

Labs & Collaborations:

Amanda collaborates with interdisciplinary teams on projects involving policy design, algorithmic fairness, and healthcare equity. She co-organized the ML4D workshops at NeurIPS 2018-2019 and leads the FEAT reading group at CMU.