Robust Evaluation Measures for Evaluating Social Biases in Masked Language Models
This work addresses the problem of robust bias evaluation for researchers and practitioners using masked language models, though it is incremental as it builds on prior measures.
The paper tackles the problem of evaluating social biases in masked language models by addressing the lack of robustness in existing measures, which rely on pseudo-log-likelihood scores without capturing distributional information. It proposes new measures using Gaussian distributions and divergences, showing significantly improved robustness and interpretability on datasets like StereoSet and CrowS-Pairs.
Many evaluation measures are used to evaluate social biases in masked language models (MLMs). However, we find that these previously proposed evaluation measures are lacking robustness in scenarios with limited datasets. This is because these measures are obtained by comparing the pseudo-log-likelihood (PLL) scores of the stereotypical and anti-stereotypical samples using an indicator function. The disadvantage is the limited mining of the PLL score sets without capturing its distributional information. In this paper, we represent a PLL score set as a Gaussian distribution and use Kullback Leibler (KL) divergence and Jensen Shannon (JS) divergence to construct evaluation measures for the distributions of stereotypical and anti-stereotypical PLL scores. Experimental results on the publicly available datasets StereoSet (SS) and CrowS-Pairs (CP) show that our proposed measures are significantly more robust and interpretable than those proposed previously.