@inproceedings{zalkikar-chandra-2025-measuring,
title = "Measuring Social Biases in Masked Language Models by Proxy of Prediction Quality",
author = "Zalkikar, Rahul and
Chandra, Kanchan",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.68/",
doi = "10.18653/v1/2025.acl-long.68",
pages = "1337--1361",
ISBN = "979-8-89176-251-0",
abstract = "Innovative transformer-based language models produce contextually-aware token embeddings and have achieved state-of-the-art performance for a variety of natural language tasks, but have been shown to encode unwanted biases for downstream applications. In this paper, we evaluate the social biases encoded by transformers trained with the masked language modeling objective using proposed proxy functions within an iterative masking experiment to measure the quality of transformer models' predictions and assess the preference of MLMs towards disadvantaged and advantaged groups. We find that all models encode concerning social biases. We compare bias estimations with those produced by other evaluation methods using benchmark datasets and assess their alignment with human annotated biases. We extend previous work by evaluating social biases introduced after retraining an MLM under the masked language modeling objective and find proposed measures produce more accurate and sensitive estimations of biases based on relative preference for biased sentences between models, while other methods tend to underestimate biases after retraining on sentences biased towards disadvantaged groups."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zalkikar-chandra-2025-measuring">
<titleInfo>
<title>Measuring Social Biases in Masked Language Models by Proxy of Prediction Quality</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Zalkikar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kanchan</namePart>
<namePart type="family">Chandra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>Innovative transformer-based language models produce contextually-aware token embeddings and have achieved state-of-the-art performance for a variety of natural language tasks, but have been shown to encode unwanted biases for downstream applications. In this paper, we evaluate the social biases encoded by transformers trained with the masked language modeling objective using proposed proxy functions within an iterative masking experiment to measure the quality of transformer models’ predictions and assess the preference of MLMs towards disadvantaged and advantaged groups. We find that all models encode concerning social biases. We compare bias estimations with those produced by other evaluation methods using benchmark datasets and assess their alignment with human annotated biases. We extend previous work by evaluating social biases introduced after retraining an MLM under the masked language modeling objective and find proposed measures produce more accurate and sensitive estimations of biases based on relative preference for biased sentences between models, while other methods tend to underestimate biases after retraining on sentences biased towards disadvantaged groups.</abstract>
<identifier type="citekey">zalkikar-chandra-2025-measuring</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.68</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.68/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>1337</start>
<end>1361</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Measuring Social Biases in Masked Language Models by Proxy of Prediction Quality
%A Zalkikar, Rahul
%A Chandra, Kanchan
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F zalkikar-chandra-2025-measuring
%X Innovative transformer-based language models produce contextually-aware token embeddings and have achieved state-of-the-art performance for a variety of natural language tasks, but have been shown to encode unwanted biases for downstream applications. In this paper, we evaluate the social biases encoded by transformers trained with the masked language modeling objective using proposed proxy functions within an iterative masking experiment to measure the quality of transformer models’ predictions and assess the preference of MLMs towards disadvantaged and advantaged groups. We find that all models encode concerning social biases. We compare bias estimations with those produced by other evaluation methods using benchmark datasets and assess their alignment with human annotated biases. We extend previous work by evaluating social biases introduced after retraining an MLM under the masked language modeling objective and find proposed measures produce more accurate and sensitive estimations of biases based on relative preference for biased sentences between models, while other methods tend to underestimate biases after retraining on sentences biased towards disadvantaged groups.
%R 10.18653/v1/2025.acl-long.68
%U https://aclanthology.org/2025.acl-long.68/
%U https://doi.org/10.18653/v1/2025.acl-long.68
%P 1337-1361
Markdown (Informal)
[Measuring Social Biases in Masked Language Models by Proxy of Prediction Quality](https://aclanthology.org/2025.acl-long.68/) (Zalkikar & Chandra, ACL 2025)
ACL