Edu-MMBias: A Three-Tier Multimodal Benchmark for Auditing Social Bias in Vision-Language Models under Educational Contexts

TL;DR

This paper introduces Edu-MMBias, a comprehensive framework for detecting social biases in vision-language models within educational contexts, revealing biases that bypass text safeguards and highlighting visual modality's role.

Contribution

It presents a novel three-tier bias auditing framework with a generative pipeline and human-in-the-loop verification for holistic bias detection in VLMs.

Findings

Models favor lower-status narratives, indicating class bias.

Deep-seated stereotypes related to health and race are present.

Visual inputs can trigger biases bypassing text safeguards.

Abstract

As Vision-Language Models (VLMs) become integral to educational decision-making, ensuring their fairness is paramount. However, current text-centric evaluations neglect the visual modality, leaving an unregulated channel for latent social biases. To bridge this gap, we present Edu-MMBias, a systematic auditing framework grounded in the tri-component model of attitudes from social psychology. This framework diagnoses bias across three hierarchical dimensions: cognitive, affective, and behavioral. Utilizing a specialized generative pipeline that incorporates a self-correct mechanism and human-in-the-loop verification, we synthesize contamination-resistant student profiles to conduct a holistic stress test on state-of-the-art VLMs. Our extensive audit reveals critical, counter-intuitive patterns: models exhibit a compensatory class bias favoring lower-status narratives while simultaneously harboring deep-seated health and racial stereotypes. Crucially, we find that visual inputs act as a safety backdoor, triggering a resurgence of biases that bypass text-based alignment safeguards and revealing a systematic misalignment between latent cognition and final decision-making. The contributions of this paper are available at: https://anonymous.4open.science/r/EduMMBias-63B2.