Deliverables relating to the Ethics in Artificial Intelligence & Intelligent Interfaces University Units
-
Updated
Jan 18, 2024
Deliverables relating to the Ethics in Artificial Intelligence & Intelligent Interfaces University Units
Does a classifier trained on COMPAS data still show racial bias without seeing race or COMPAS scores? Fairness audit showing proxy effects (priors_count), threshold sensitivity, and why dropping race redistributes harm rather than removing it.
AI fairness auditing tool that detects multi-hop proxy discrimination chains in tabular datasets. Beats published academic baselines on COMPAS, Adult, and German Credit.
Proxy discrimination auditing for insurance pricing — FCA EP25/2, Consumer Duty, bias metrics
To associate your repository with the proxy-discrimination topic, visit your repo's landing page and select "manage topics."