EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
3311. From fair predictions to just decisions
Invited abstract in session MB-27: On Mathematical Optimization for Explainable and Fair Machine Learning, stream Mathematical Optimization for XAI.
Monday, 10:30-12:00Room: 047 (building: 208)
Authors (first author is the speaker)
| 1. | Elif Ersoy
|
| Industrial Engineering, Ozyegin University | |
| 2. | Enis Kayis
|
| Industrial Engineering Department, Ozyegin University |
Abstract
Data-driven decisions often raise concerns about fairness and justice, which are frequently confused and used interchangeably. This study investigates the relationship between these two definitions: Fairness is generally considered during the prediction phase of decision-making, yet justice mostly relates to the equality or equity of the resulting decisions. Thus fairness in predictions may not automatically yield just decisions. We propose metrics for quantifying distributive justice in decision-making and investigate how fair predictions relate to just decisions. Two main questions guide our research: Do fair predictions translate into just decisions? Secondly, what is the financial impact of making just decisions? We propose justice metrics and developed mathematical models to improve the quality of the decisions from this viewpoint. We focus on a credit lending application. Our results underscore the distinction between fairness and justice, demonstrating that fair predictions do not guarantee just decisions and may even reduce the overall profit for the lending firm.
Keywords
- Machine Learning
Status: accepted
Back to the list of papers