Second Order WinoBias (SoWinoBias) Test Set for Latent Gender Bias Detection in Coreference Resolution

TL;DR

This paper introduces SoWinoBias, a test set designed to detect hidden gender biases in coreference resolution systems, revealing biases that persist despite debiasing efforts.

Contribution

The paper presents a novel test set, SoWinoBias, specifically aimed at uncovering latent gender biases in coreference models, and evaluates current debiasing methods on it.

Findings

Current debiasing methods do not fully eliminate latent gender bias.

SoWinoBias reveals biases not detected by traditional test sets.

Embedding space alterations impact bias detection.

Abstract

We observe an instance of gender-induced bias in a downstream application, despite the absence of explicit gender words in the test cases. We provide a test set, SoWinoBias, for the purpose of measuring such latent gender bias in coreference resolution systems. We evaluate the performance of current debiasing methods on the SoWinoBias test set, especially in reference to the method's design and altered embedding space properties. See https://github.com/hillarydawkins/SoWinoBias.