(she/her/hers)
Carnegie Mellon University
causal inference, machine learning, algorithmic fairness
Amanda Coston is a PhD student in Machine Learning and Public Policy at Carnegie Mellon University. Her research investigates how to make decision-making systems more reliable and more equitable. A central focus of her work is identifying when algorithms, data used for policy-making, and human decisions disproportionately impact marginalized groups. She is advised by Alexandra Chouldechova and Edward Kennedy. Her research has been featured in the Wall Street Journal and VentureBeat.
Amanda is an NSF GRFP Fellow, K & L Gates Presidential Fellow in Ethics and Computational Technologies, Meta Research PhD Fellow, and Tata Consultancy Services Presidential Fellow. Amanda holds a B.S.E from Princeton University where she was advised by Robert Schapire.
The role of validity in responsible AI
As automated decision systems proliferate high-stakes settings throughout society from healthcare to criminal justice, widespread concerns around the suitability and equity of these systems require urgent attention. Much of the current discourse focuses on fairness and ethics, often overlooking first-order questions of validity. This talk explores issues of data quality and problem formulation, including selection bias, confounding, and missing data. Our empirical analysis in the child welfare and consumer lending domains demonstrates that, when unaddressed, these issues have profound downstream consequences, invalidating fairness assessments and fairness-promoting interventions. Drawing upon doubly-robust techniques from causal inference, we propose novel methods to address these problems.