Rethinking Explainability in the Era of Multimodal AI

TL;DR

This paper emphasizes the importance of developing multimodal explanations that capture cross-modal interactions, criticizing unimodal methods and proposing principles for more faithful and stable explanations in multimodal AI systems.

Contribution

It introduces key principles for multimodal explanations—Granger-style influence, synergistic faithfulness, and unified stability—to improve interpretability of multimodal models.

Findings

Unimodal explanations fail to capture cross-modal influences.

Proposed principles guide the development of more faithful multimodal explanations.

Enhanced explanations can uncover shortcuts and reduce modality bias.

Abstract

While multimodal AI systems (models jointly trained on heterogeneous data types such as text, time series, graphs, and images) have become ubiquitous and achieved remarkable performance across high-stakes applications, transparent and accurate explanation algorithms are crucial for their safe deployment and ensure user trust. However, most existing explainability techniques remain unimodal, generating modality-specific feature attributions, concepts, or circuit traces in isolation and thus failing to capture cross-modal interactions. This paper argues that such unimodal explanations systematically misrepresent and fail to capture the cross-modal influence that drives multimodal model decisions, and the community should stop relying on them for interpreting multimodal models. To support our position, we outline key principles for multimodal explanations grounded in modality: Granger-style modality influence (controlled ablations to quantify how removing one modality changes the explanation for another), Synergistic faithfulness (explanations capture the model's predictive power when modalities are combined), and Unified stability (explanations remain consistent under small, cross-modal perturbations). This targeted shift to multimodal explanations will help the community uncover hidden shortcuts, mitigate modality bias, improve model reliability, and enhance safety in high-stakes settings where incomplete explanations can have serious consequences.