In this episode we have PhD students Yongsu Ahn and Alex Cabrera to talk about two separate data visualization systems they developed to help people analyze machine learning models in terms of potential biases they may have. The systems are called FairSight and FairVis and have slightly different goals. FairSight focuses on models that generate rankings (e.g., in school admissions) and FairVis more on comparison of fairness metrics. With them we explore the world of “machine bias” trying to understand what it is and how visualization can play a role in its detection and mitigation.
Enjoy the show!
- Alex Cabrera
- Yongsu Ahn
- Google: “Attacking Discrimination with Smarter Machine Learning”
- Nicky Case: “Parable of Polygons”