Beyond traditional assumptions in fair machine learning

Abstract

This thesis scrutinizes common assumptions underlying traditional machine learning approaches to fairness in consequential decision making. After challenging the validity of these assumptions in real-world applications, we propose ways to move forward when they are violated. First, we show that group fairness criteria purely based on statistical properties of observed data are fundamentally limited. Revisiting this limitation from a causal viewpoint we develop a more versatile conceptual framework, causal fairness criteria, and first algorithms to achieve them. We also provide tools to analyze how sensitive a believed-to-be causally fair algorithm is to misspecifications of the causal graph. Second, we overcome the assumption that sensitive data is readily available in practice. To this end we devise protocols based on secure multi-party computation to train, validate, and contest fair decision algorithms without requiring users to disclose their sensitive data or decision makers to disclose their models. Finally, we also accommodate the fact that outcome labels are often only observed when a certain decision has been made. We suggest a paradigm shift away from training predictive models towards directly learning decisions to relax the traditional assumption that labels can always be recorded. The main contribution of this thesis is the development of theoretically substantiated and practically feasible methods to move research on fair machine learning closer to real-world applications.

PhD Thesis submitted at the University of Cambridge, October 2020. The thesis is based on a number of previous works also available on arxiv (see Chapter 1)

References in corpus (9)

• Equality of Opportunity in Supervised Learning
• Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
• Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence
• On Fairness and Calibration
• On the (im)possibility of fairness
• Counterfactual Risk Minimization: Learning from Logged Bandit Feedback
• What's Sex Got To Do With Fair Machine Learning?
• An Empirical Study on Learning Fairness Metrics for COMPAS Data with Human Supervision
• Exploration in two-stage recommender systems