Auditing Fairness under Unobserved Confounding

Michael Oberst, Johns Hopkins University

08.08.2024

4:00 pm - 5:30 pm

Online via Zoom

Inequity in resource allocation has been well-documented in many domains, such as healthcare. Causal measures of equity / fairness seek to isolate biases in allocation that are not explained by other factors, such as underlying need. However, these fairness measures require the (strong) assumption that we observe all relevant indicators of need, an assumption that rarely holds in practice. For instance, if resources are allocated based on indicators of need that are not recorded in our data ("unobserved confounders"), we may understate (or overstate) the amount of inequity.

In this talk, I will present work demonstrating that we can still give informative bounds on certain causal measures of fairness, even while relaxing (or even eliminating) the assumption that all relevant indicators of need are observed. We use the fact that in many real-world settings (e.g., the release of a new treatment) we have data from prior to any allocation, which can be used to derive unbiased estimates of need. This result is of immediate practical interest: we can audit unfair outcomes of existing decision-making systems in a principled manner. For instance, in a real-world study of Paxlovid allocation, we show that observed racial inequity cannot be explained by unobserved confounders of the same strength as important observed covariates.

Organized by:

Institute of AI in Management LMU Munich

Link to Representation Learning: A Causal Perspective

©jittawit.21 - stock.adobe.com

AI Keynote Series • 18.07.2024 • Online via Zoom

Representation Learning: A Causal Perspective

Lecture with Yixin Wang from University of Michigan, applying causal insights to create clear, efficient representations using counterfactuals.

Link to Interpretable prediction with missing values

©jittawit.21 - stock.adobe.com

AI Keynote Series • 06.06.2024 • Online via Zoom

Interpretable prediction with missing values

Missing values in healthcare data hinder interpretability and predictions. Fredrik Johansson's talk presents two solutions and suggests future research directions.

Link to Innovating AI Products for Social Good
in the Age of Foundational Models

©jittawit.21 - stock.adobe.com

AI Keynote Series • 23.05.2024 • Online via Zoom

Innovating AI Products for Social Good in the Age of Foundational Models

Professor Qian Yang explores how LLMs necessitate considering societal impacts and innovating for social good in education and mental healthcare.

AI Keynote Series • 08.02.2024 • LMU Institute of AI in Management via zoom

Causal Scoring: A Framework for Effect Estimation, Effect Ordering, and Effect Classification

The presentation introduces causal scoring for decision-making, with interpretations.

AI Keynote Series • 18.01.2024 • LMU Institute of AI in Management via zoom

Machine Learning Who to Nudge: Causal vs Predictive Targeting in a Field Experiment on Student Financial Aid Renewal

Join the presentation of Jann Spiess, Stanford Graduate School of Business, on the nuances of intervention effectiveness on targeting strategies.

Auditing Fairness under Unobserved Confounding

Michael Oberst, Johns Hopkins University

Related