Fairness is a controversial but important concept in automated decision making. Accuracy is not sufficient when analyzing a ML classification algorithm. Fairness metrics are used to quantify and mitigate biases in classification predictions. This module will get into a deeper analysis of techniques and its main uses in AI for classification purposes.Additionally, we will review the most relevant existing tools to operationalize fairness in a real use case.
Videos:
Arvind Narayanan (2018): 21 Fairness definitions and their politics
Solon Barocas (2017) : NIPS 2017 Tutorial on Fairness in Machine Learning
Printed materials:
Case study: Titanic predictive model