The need for causality to address fairness in ML
Addressing the problem of fairness is crucial to safely use machine learning algorithms to support decisions with a critical impact on people’s lives such as job hiring, child maltreatment, disease diagnosis, loan granting, etc. Several notions of fairness have been defined and examined in the past decade, such as statistical parity and equalized odds. The most recent fairness notions, however, are causal-based and reflect the now widely accepted idea that using causality is necessary to appropriately address the problem of fairness.
The main objective of our research is to measure discrimination as accurately as possible. To this end, we make a distinction between the concepts of “bias” (a deviation of an estimation from the quantity it estimates) and “discrimination” (the unjust or prejudicial treatment of different categories of people on the ground of race, age, gender, disability, etc.).
In this seminar we start by illustrating why causality is essential to reach this objective of accurately measuring fairness. Then, we show how causality can be used to characterise and quantify four types of bias, namely, representation bias, confounding bias, selection bias, and measurement bias. Finally, we briefly present the different research directions we are pursuing at Comète team related to ethical/responsible AI.
Sami Zhioua is an advanced researcher at Comète team (INRIA Saclay-Île-de-France) since September 2021. His current research work focuses on ethical aspects of Machine Learning and AI. In particular, he is working on measuring and mitigating discrimination (fairness) and privacy leaks in automated decision systems using causality. Previously, he worked on privacy enhancing technologies such as Tor, malware analysis, industrial control system (SCADA) security, and using reinforcement learning (RL) to solve software engineering problems. He holds a Ph.D. in Computer Science from Laval University, Québec, Canada and his academic experience includes research and teaching positions at McGill University, KFUPM, and HCT Dubai.