GEOFFair: a GEOmetric Framework for Fairness

Alessandro Maggio, Luca Giuliani, Roberta Calegari, Michele Lombardi, Michela Milano

Primary Area: societal considerations including fairness, safety, privacy

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: AI fairness, geometric framework, GEOFFair

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: A proposal of a geometric framework to study the fairness in a generic data-driven AI problem

Abstract:

Fairness is a critical concern in Machine Learning, impacting its applications across domains. Existing fairness analyses often rely on complex mathematics, lacking of intuitive understanding.
In this study, we introduce \emph{GEOFFair}, a Geometric Framework for Fairness. It represents Machine Learning elements as vectors and sets, offering a more intuitive understanding of fairness related concepts. GEOFFair visualizes fairness mitigation techniques as vector projections, it provides a solid base to investigate the bias injection, aiding in constructing proofs, and it enables the study of fairness properties by means of geometric considerations.
The main contribution of the work is to highlight GEOFFair's effectiveness in fairness studies, demonstrating that solely maximizing accuracy based on observed labels may not always be optimal for fairness.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 6542