Forskarseminarium: How to be fair? A study of label and selection bias
Seminarium
Datum: fredag 24 november 2023
Tid: 10.00 – 11.00
Plats: Rum L30, DSV, Nodhuset, Borgarfjordsgatan 12, Kista
Välkommen till ett forskarseminarium om snedvridna data, och hur man kan skapa mer rättvisa modeller. Talare är den belgiska forskaren Toon Calders.
24 november besöker Toon Calders Institutionen för data- och systemvetenskap, DSV. Han är professor i datavetenskap vid University of Antwerp i Belgien.
Toon Calders är inbjuden av forskargruppen Data Science Research Group och kommer att hålla i ett forskarseminarium på DSV. Under seminariet delar han med sig av sina senaste forskningsresultat med utgångspunkt i en artikel som har publicerats i den vetenskapliga tidskriften Machine Learning.
Läs artikeln ”How to be fair? A study of label and selection bias”
Seminariet genomförs i DSVs lokaler i Kista. Ingen föranmälan krävs!
Kontakta Franco Rugolon om du har frågor
Kort om seminariet (på engelska)
It is widely accepted that biased data leads to biased and thus potentially unfair models. Therefore, several measures for bias in data and model predictions have been proposed, as well as bias mitigation techniques whose aim is to learn models that are fair by design.
Despite the myriad of mitigation techniques developed in the past decade, however, it is still poorly understood under what circumstances which methods work. Recently, Wick et al. showed, with experiments on synthetic data, that there exist situations in which bias mitigation techniques lead to more accurate models when measured on unbiased data. Nevertheless, in the absence of a thorough mathematical analysis, it remains unclear which techniques are effective under what circumstances.
We propose to address this problem by establishing relationships between the type of bias and the effectiveness of a mitigation technique, where we categorize the mitigation techniques by the bias measure they optimize. In this paper we illustrate this principle for label and selection bias on the one hand, and demographic parity and ‘‘We’re All Equal’’ on the other hand. Our theoretical analysis allows to explain the results of Wick et al. and we also show that there are situations where minimizing fairness measures does not result in the fairest possible distribution.
Senast uppdaterad: 20 november 2023
Sidansvarig: Institutionen för data- och systemvetenskap, DSV