Keynote Speech 2: Bias and Discrimination in AI Systems: From Single-Identity Dimensions to Multi-Discrimination

AI-driven decision-making has become pervasive in various aspects of our lives, impacting everyone, everywhere, at any time. However, concerns have arisen regarding the discriminatory effects of AI, as evidenced across diverse domains like content recommendation, healthcare, predictive policing, and...

Full description

Saved in:
Bibliographic Details
Published in2023 International Conference on Intelligent Computing, Communication, Networking and Services (ICCNS) p. 2
Main Author Ntoutsi, Eirini
Format Conference Proceeding
LanguageEnglish
Published IEEE 19.06.2023
Subjects
Informatics
Intelligent systems
Learning (artificial intelligence)
Machine learning
Machine learning algorithms
Medical services
Supervised learning
Online AccessGet full text

Cover

More Information
Summary:AI-driven decision-making has become pervasive in various aspects of our lives, impacting everyone, everywhere, at any time. However, concerns have arisen regarding the discriminatory effects of AI, as evidenced across diverse domains like content recommendation, healthcare, predictive policing, and autonomous driving. The field of fairness-aware machine learning aims to address bias and discrimination in AI/ML models, but most existing approaches focus on discrimination related to a single protected attribute, such as gender or race. In reality, human identities are multi-dimensional, and discrimination can arise from multiple protected characteristics, for example a combination of gender, race and age. In this talk, I will explore fairness and discrimination in supervised learning for tabular data, progressing from traditional single-attribute discrimination to addressing the complexities of multi-dimensional discrimination.
DOI:10.1109/ICCNS58795.2023.10193086