Achieving Socio-Economic Parity Through the Lens of EU AI Act

This lecture addresses unfair treatment and discrimination as central ethical problems of AI systems and situates them within the context of the EU AI Act, which aims to promote innovation while simultaneously protecting fundamental rights. Although the Act calls for the use of existing fairness concepts, these often neglect socioeconomic status and thus unintentionally favor economically privileged groups. This deficiency contradicts the principle of equal opportunity, which demands the compensation of disadvantages beyond individual control.

The paper analyzes common fairness concepts using real-world data and demonstrates their inability to effectively address socioeconomic inequalities. As a solution, the new fairness concept of socioeconomic parity (SEP) is proposed, the effectiveness of which is empirically demonstrated and which enables a better alignment of AI fairness with legal and social requirements.

Eirini Ntoutsi is a Full Professor for Open Source Intelligence at the Faculty of Computer Science, University of the Bundeswehr Munich and the Research Institute CODE.