/socialbias-dataset-construction-biases

Dataset Construction Biases of Social Bias Benchmarks

Primary LanguagePython

Dataset Construction Biases of Social Bias Benchmarks

This repository contains the all datasets and pre-processing/evaluation code for our ACL 2023 paper titled The Tail Wagging the Dog: Dataset Construction Biases of Social Bias Benchmarks. Here is a link to the preprint.

Abstract

  • How reliably can we trust the scores obtained from social bias benchmarks as faithful indicators of problematic social biases in a given model?
  • In this work, we study this question by contrasting social biases with non-social biases that stem from choices made during dataset construction (which might not even be discernible to the human eye).
  • To do so, we empirically simulate various alternative constructions for a given benchmark based on seemingly innocuous modifications (such as paraphrasing or random-sampling) that maintain the essence of their social bias.
  • On two well-known social bias benchmarks (Winogender and BiasNLI), we observe that these shallow modifications have a surprising effect on the resulting degree of bias across various models and consequently the relative ordering of these models when ranked by measured bias.


Bias measures on (a) Winogender (percentage M-F mismatch, log-scale) and (b) BiasNLI (accuracy as percentage neutral, log-scale), across a variety of dataset constructions and models.

- We hope these troubling observations motivate more robust measures of social biases!

Usage

  • All the datasets are publicly available. We include fully preprocessed datasets (corresponding to the various alternate constructions) in this repository.
  • All the models used are publicly available. They are extremely simple to use, and we include sample scripts used to run relevant experiments using these models.
  • Please refer to the folders winogender and biasnli and their corresponding READMEs for further details.