Loading...

Media is loading
 

Paper Type

Complete

Paper Number

1482

Description

Artificial Intelligence (AI) holds great promise in beneficial, accurate, and effective predictive and real-time decision-making in a wide range of use cases. However, there are concerns regarding potential risks, harm, trust, and fairness issues arising from some AI algorithms' opacity and potential unfairness because of their un-explainability and concern with objectivity. This study proposes a framework for evaluating a machine learning model that incorporates explainability for AI fairness as currently, no such framework exists. We evaluate its applicability with a classification problem using multiple classifiers. The experimental case study demonstrates the successful application of the performance-explainability-fairness framework to the classification problem. The framework can guide means for improving fairness in machine learning models.

Comments

SIG ODIS

Share

COinS
Top 25 Paper Badge
 
Aug 10th, 12:00 AM

Towards a Performance-explainability-fairness Framework for Benchmarking ML Models

Artificial Intelligence (AI) holds great promise in beneficial, accurate, and effective predictive and real-time decision-making in a wide range of use cases. However, there are concerns regarding potential risks, harm, trust, and fairness issues arising from some AI algorithms' opacity and potential unfairness because of their un-explainability and concern with objectivity. This study proposes a framework for evaluating a machine learning model that incorporates explainability for AI fairness as currently, no such framework exists. We evaluate its applicability with a classification problem using multiple classifiers. The experimental case study demonstrates the successful application of the performance-explainability-fairness framework to the classification problem. The framework can guide means for improving fairness in machine learning models.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.