These tools and metrics are designed to help AI actors develop and use trustworthy AI systems and applications that respect human rights and are fair, transparent, explainable, robust, secure and safe.
Gender-based Illicit Proximity Estimate (GIPE)
This paper proposes a new bias evaluation metric – Gender-based Illicit Proximity Estimate (GIPE), which measures the extent of undue proximity in word vectors resulting from the presence of gender-based predilections. Experiments based on a suite of evaluation metrics show that RAN-Debias significantly outperforms the state-of-the-art in reducing proximity bias (GIPE) by at least 42.02%. It also reduces direct bias, adding minimal semantic disturbance, and achieves the best performance in a downstream application task (coreference resolution).
Trustworthy AI Relevance
This metric addresses Fairness and Transparency by quantifying relevant system properties. Fairness: GIPE directly measures cross-group (gender) associations with a protected-concept class (illicit), making it a bias-detection metric that can reveal disparate representations that may lead to discriminatory outcomes (e.g., profiling, stigmatizing labels, unequal moderation). Tracking GIPE can guide debiasing, equalization, or downstream fairness evaluations.
Related use cases :
Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval
Uploaded on Apr 22, 2024Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval
Uploaded on May 21, 2024Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval
Uploaded on Jun 5, 2024Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval
Uploaded on Jan 14, 2025Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval
Uploaded on Jun 19, 2025About the metric
You can click on the links to see the associated metrics
Objective(s):
Purpose(s):
Target sector(s):
Lifecycle stage(s):
Target users:
Risk management stage(s):
Partnership on AI
