State of the Art Open Source Software for AI Fairnessevaluation
summary
Jurity implements a method that infers protected-group fairness without collecting individual demographics using surrogate classes, built from ZIP‑code or census probabilities, and scores metrics via a bootstrap linear regression under the hood. Open source on GitHub and sponsored by Fidelity Investments, it shows comparable results on simulated data and an HMDA mortgage case study, requires roughly 100 surrogate classes with 20–30 individuals each, and reports when inputs or measurement error make estimates unstable.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:11Fairness testing importance
- 3:58Missing demographic data
- 5:48Surrogate classes and fairness
- 7:10Jity library setup
- 10:19Interpreting Jity metrics
- 12:05Visualizing fairness results
- 16:58Q&A
- 23:32Closing remarks and call for feedback
related talks
35mOpen Source Fairness
48mResponsible AI
01h 21mEmpirical methods fairer AI - Auditing a Dutch public sector risk profiling algorithm
36mMachine Learning Fairness: Lessons Learned
14mBuilding with the Responsible AI toolkit
40mEthical AI: Addressing Bias & Algorithmic Fairness
32mOperationalizing Responsible AI
11mAddressing Privacy, Bias, and Fairness in AI Systems