Choisir la langue :

I. Valera (Max Planck): Fairness in Algorithmic Decisions: From definitions to mechanisms

Institutional tag: 
Thematic tag(s): 

Algorithmic decision making systems are ubiquitous across a wide variety of online as well as offline services. These systems rely on complex learning methods and vast amounts of data to optimize the service functionality, satisfaction of the end user and profitability. However, there is a growing concern that these automated decisions can lead, even in the absence of intent, to a lack of fairness, i.e., their outcomes can disproportionately hurt (or, benefit) particular groups of people sharing one or more sensitive attributes (e.g., race, sex).
In this talk, I will provide an overview of the state of the art on fairness in classification. In particular, in the first part of the talk I will review the main definitions of fairness exiting in the literature, which are based on parity and then I will describe our approach to train fair classifiers (under a given definition). Finally, I will cover my most recent work, in which we introduce, formalize and evaluate new notions of fairness that are inspired by the concepts of fair division and envy-freeness in economics and game theory. These new notions of fairness allow us to relax the parity-based notions, which may be too stringent, precluding more accurate decisions.

Tuesday, November 28, 2017 - 11:00 to 12:00
Inria B11
Isabel Valera
Max Planck Institute for Software Systems