Multimodal Explanations: Justifying Decisions and Pointing to the Evidence

TL;DR

This paper introduces a multimodal explanation approach combining textual and visual justifications for deep models, demonstrating improved localization and interpretability in activity recognition and visual question answering tasks.

Contribution

It proposes a novel multimodal explanation model, introduces two new datasets, and shows that combining modalities enhances explanation quality and evidence localization.

Findings

Training with textual explanations improves textual justification quality.

Multimodal explanations outperform unimodal approaches in evidence localization.

Visual explanations can be more insightful than textual ones in certain cases.

Abstract

Deep models that are both effective and explainable are desirable in many settings; prior explainable models have been unimodal, offering either image-based visualization of attention weights or text-based generation of post-hoc justifications. We propose a multimodal approach to explanation, and argue that the two modalities provide complementary explanatory strengths. We collect two new datasets to define and evaluate this task, and propose a novel model which can provide joint textual rationale generation and attention visualization. Our datasets define visual and textual justifications of a classification decision for activity recognition tasks (ACT-X) and for visual question answering tasks (VQA-X). We quantitatively show that training with the textual explanations not only yields better textual justification models, but also better localizes the evidence that supports the decision. We also qualitatively show cases where visual explanation is more insightful than textual explanation, and vice versa, supporting our thesis that multimodal explanation models offer significant benefits over unimodal approaches.