Introduction to Fairness in Machine Learning
Abstract
Machine learning models are increasingly used in socially significant and sensitive domains, where an inaccurate model is a problem but an unfair one can be a legal liability. This lecture introduces the ways unfairness is defined and measured in machine learning, shows where bias enters the pipeline, and reviews techniques for mitigating it during data preparation, model training, and decision post-processing.
I delivered an invited lecture at the IEEE Computer Chapter Co-16 Seminar, the long-running seminar series run jointly by the Mathematical Institute SANU, the Faculty of Organizational Sciences, and the IEEE C-16 Chapter.
The lecture was an introduction to fairness in machine learning: how unfairness is defined and measured, where bias enters a modelling pipeline, and what can be done about it at the data preparation, model training, and decision post-processing stages.
I returned to the same seminar series in 2025 with a talk on research on fairness in machine learning models.
Sandro Radovanović
Assistant Professor at University of Belgrade
My research interests include machine learning, development and design of decision support systems, decision theory, and fairness and justice concepts in algorithmic decision making.