BiasDora: Exploring Hidden Biased Associations in Vision-Language Models

Chahat Raj, Anjishnu Mukherjee, Aylin Caliskan, Antonios Anastasopoulos, Ziwei Zhu
7/2/2024
cs.CL

Official code: chahatraj/BiasDora (2 stars, last commit 2024-07-04).

Abstract

Existing works examining Vision-Language Models (VLMs) for social biases predominantly focus on a limited set of documented bias associations, such as gender:profession or race:crime. This narrow scope often overlooks a vast range of unexamined implicit associations, restricting the identification and, hence, mitigation of such biases. We address this gap by probing VLMs to (1) uncover hidden, implicit associations across 9 bias dimensions. We systematically explore diverse input and output modalities and (2) demonstrate how biased associations vary in their negativity, toxicity, and extremity. Our work (3) identifies subtle and extreme biases that are typically not recognized by existing methodologies. We make the Dataset of retrieved associations, (Dora), publicly available here https://github.com/chahatraj/BiasDora.

View on arXivView PDF

Code Implementations(1)

chahatraj/BiasDoraOfficial100%

EMNLP 2024

20PythonJun 16, 20242 years ago

Similar Papers

VisBias: Measuring Explicit and Implicit Social Biases in Vision Language Models

89% match1 repo32025

A Unified Framework and Dataset for Assessing Societal Bias in Vision-Language Models

88% match1 repo12024

VIGNETTE: Socially Grounded Bias Evaluation for Vision-Language Models

88% match1 repo12025

Images Speak Louder than Words: Understanding and Mitigating Bias in Vision-Language Model from a Causal Mediation Perspective

88% match2024

Identifying Implicit Social Biases in Vision-Language Models

87% match2024

MultiModal Bias: Introducing a Framework for Stereotypical Bias Assessment beyond Gender and Race in Vision Language Models

87% match1 repo42023

Cite this paper

@article{raj2024biasdora,
  title  = {BiasDora: Exploring Hidden Biased Associations in Vision-Language Models},
  author = {Chahat Raj and Anjishnu Mukherjee and Aylin Caliskan and Antonios Anastasopoulos and Ziwei Zhu},
  year   = {2024},
  eprint = {2407.02066},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2407.02066v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2407.02066.svg)](https://codeofpaper.com/paper/2407.02066)

Discussion