Data Skeptic cover image

Fairness Aware Outlier Detection

Data Skeptic

00:00

The Shortcoming of Statistical Parity

Group fidelity is a property where two outliers should be equally likely to be assigned higher scores, regardless of their membership. In this paper, we propose an approximation to group fidelity that would preserve the group level ranks from the base models. We tried to preserve this notion that we have for it from the base detector into discrite aliens. The other thing is like one desideratum that we propose bat we cannot enforce, or we could not enforce wor something like base rate preservation. But at least a men achieve astistical parity. So what we say i that in the fair outlie direction, we should be able to preserve this base rate in the flagged population that we have.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app