BanStereoSet: A Dataset to Measure Stereotypical Social Biases in LLMs for Bangla

TL;DR

BanStereoSet is a culturally localized dataset for evaluating stereotypical social biases in Bangla language models, addressing the gap in bias research beyond English and highlighting significant biases in current models.

Contribution

The paper introduces BanStereoSet, a new bias evaluation dataset tailored for Bangla, extending existing bias datasets to a less-studied language and social context.

Findings

Language models exhibit significant biases in Bangla across multiple social categories.

The dataset reveals culturally specific stereotypes prevalent in Bangla-speaking communities.

Bias measurement highlights the need for culturally adapted datasets for fair language technology development.

Abstract

This study presents BanStereoSet, a dataset designed to evaluate stereotypical social biases in multilingual LLMs for the Bangla language. In an effort to extend the focus of bias research beyond English-centric datasets, we have localized the content from the StereoSet, IndiBias, and Kamruzzaman et. al.'s datasets, producing a resource tailored to capture biases prevalent within the Bangla-speaking community. Our BanStereoSet dataset consists of 1,194 sentences spanning 9 categories of bias: race, profession, gender, ageism, beauty, beauty in profession, region, caste, and religion. This dataset not only serves as a crucial tool for measuring bias in multilingual LLMs but also facilitates the exploration of stereotypical bias across different social categories, potentially guiding the development of more equitable language technologies in Bangladeshi contexts. Our analysis of several language models using this dataset indicates significant biases, reinforcing the necessity for culturally and linguistically adapted datasets to develop more equitable language technologies.