AI Seminar: What does it mean for an algorithm to be fair?
Sune Hannibal Holm, associate professor at the Department of Food and Resource Economics, University of Copenhagen.
As algorithms are used more widely in society there are growing concerns that they are biased and discriminate against certain groups. A much discussed case concerns the use of predictive algorithms in decisions about granting bail, but fairness concerns also arise in many other contexts of using algorithms such as hiring and medical diagnostics. How to operationalize algorithmic fairness will in large part depend on how the problem is conceptualized. Is an algorithm fair if certain sensitive features are excluded? Is algorithmic fairness a matter of equality of probability of some outcome? As it happens, similar questions have been the subject of intense study in philosophical ethics. In this talk we will gauge the prospects of clarifying the challenges for operationalising algorithmic fairness by drawing on recent work on fairness in ethical analysis.
AI Seminar Series
This seminar is a part of the AI Seminar Series organised by SCIENCE AI Centre. The series highlights advances and challenges in research within Machine Learning, Data Science, and AI. Like the AI Centre itself, the seminar series has a broad scope, covering both new methodological contributions, ground-breaking applications, and impacts on society.