Responsible Machine Learning Through the Lens of Causal Inference

Responsible Machine Learning Through the Lens of Causal Inference

Amanda Coston

Machine learning algorithms are widely used for decision-making in societally high-stakes domains from child welfare and criminal justice to healthcare and consumer lending. These algorithms have proved unreliable or inequitable in a number of consequential settings. We take a causal approach to evaluating the responsible use of machine learning in societally high-stakes settings, guided by two pillars: validity and equity. We address data problems that challenge the validity of standard evaluation procedures by developing new counterfactual evaluation methods. We conduct causal audits for bias in the system in which algorithms are used to inform decision-making. Throughout we propose novel methods that use doubly-robust techniques for bias correction. We present empirical results in the child welfare and criminal justice settings using data from Allegheny County’s Department of Human Services and the Stanford Open Policing Project.

Amanda Coston is a PhD student in Machine Learning and Public Policy at Carnegie Mellon University. Her research investigates how to make algorithmic decision-making more reliable and more equitable using causal inference and machine learning. Prior to her PhD, she worked at Microsoft, the consultancy Teneo, and the Nairobi-based startup HiviSasa. She earned a BSE from Princeton in Computer Science with a certificate in Public Policy. Amanda is a Meta Research PhD Fellow, K & L Gates Presidential Fellow in Ethics and Computational Technologies, a NSF GRFP Fellow, and has received several Rising Star honors.