Keynote at ECAF 2026 on Tensions and Trade-offs in Algorithmic Fairness

29 May 2026

On September 3, Prof. Eirini Ntoutsi gave a keynote at the European Conference on Algorithmic Fairness (ECAF 2026) in Ghent, Belgium. ECAF brings together researchers from computer science, law, philosophy, and the social sciences to advance interdisciplinary research on algorithmic fairness and responsible AI.

In her keynote, “On the Tensions and Trade-offs of Algorithmic Fairness: From Metric Incompatibility to Cross-Pillar Trade-offs,” Eirini discussed fundamental tensions that arise when operationalizing fairness in AI systems. Beyond within-fairness tensions, such as incompatibilities between fairness notions, the talk highlighted cross-pillar trade-offs between fairness and other objectives of responsible AI, including utility, robustness, privacy, and resilience to data quality issues. 

A central message of the keynote was that not all tensions in algorithmic fairness are of the same nature. Some reflect provable incompatibilities, some arise from unstated value choices, and others represent empirical limits that current methods have not yet overcome. Distinguishing between these different types of tensions is essential, as each calls for a different response. 

Slides: https://aiml-research.github.io/files/26.ECAF.Keynote(final).pdf