[1907.06430] A Causal Bayesian Networks Viewpoint on Fairness

We offer a graphical interpretation of unfairness in a dataset as the presence of an unfair causal path in the causal Bayesian network representing the data-generation mechanism. We use this viewpoint to revisit the recent debate surrounding the COMPAS pretrial risk assessment tool and, more generally, to point out that fairness evaluation on a model requires careful considerations on the patterns of unfairness underlying the training data. We show that causal Bayesian networks provide us with a powerful tool to measure unfairness in a dataset and to design fair models in complex unfairness scenarios.

3 mentions: @csilviavr@fluck_5@balapillai
Keywords: fairness
Date: 2019/10/05 09:48

Referring Tweets

@csilviavr We use CBNs to highlight the necessity of understanding the data underlying the ML system for fairness, and to introduce techniques to deal with complex fairness scenarios. #causalinference More details in the papers: t.co/YMASznGWRd t.co/fBYZiXPEJn t.co/kdzA9uyDQF
@fluck_5 Having said that, I yet have to read the paper, nor I'm an expert on CBNs to give any other kind of judgment on the approach per se. t.co/CU3O6nTNgP
@balapillai Google’s DeepMind onto getting rid of mafa #unrambutan_seed, win-lose, numb unconscious, weak, defensive, unspontaneous and powerplaying Little Napoleans who are fake officers of the court blessed by the Bar Council.... t.co/IGQg79JmhR

Related Entries

Read more Building ML models for everyone: understanding fairness in machine learning | Google Cloud Blog
0 users, 13 mentions 2019/09/25 18:49
Read more Fairness入門 ~AIに倫理観を教える~ | AI-SCHOLAR
20 users, 12 mentions 2019/10/10 08:40
Read more Symposium and Statement on Machine Learning and Fairness | 人工知能学会 倫理委員会
0 users, 2 mentions 2019/12/28 21:51
Read more GitHub - fairlearn/fairlearn: A Python package that implements a variety of algorithms that mitigate...
0 users, 4 mentions 2019/11/06 18:51