Discrete Subgraph Sampling for Interpretable Graph based Visual Question Answering

TL;DR

This paper introduces a graph-based visual question answering system that uses discrete subgraph sampling to generate interpretable explanations intrinsically, balancing interpretability with answer accuracy, and validated through human evaluation.

Contribution

It integrates discrete subgraph sampling methods into VQA to produce intrinsically interpretable explanations, a novel approach compared to post-hoc explanation methods.

Findings

Effective mitigation of interpretability-accuracy trade-off.

Strong correlation between token co-occurrence and human preferences.

Public availability of source code for reproducibility.

Abstract

Explainable artificial intelligence (XAI) aims to make machine learning models more transparent. While many approaches focus on generating explanations post-hoc, interpretable approaches, which generate the explanations intrinsically alongside the predictions, are relatively rare. In this work, we integrate different discrete subset sampling methods into a graph-based visual question answering system to compare their effectiveness in generating interpretable explanatory subgraphs intrinsically. We evaluate the methods on the GQA dataset and show that the integrated methods effectively mitigate the performance trade-off between interpretability and answer accuracy, while also achieving strong co-occurrences between answer and question tokens. Furthermore, we conduct a human evaluation to assess the interpretability of the generated subgraphs using a comparative setting with the extended Bradley-Terry model, showing that the answer and question token co-occurrence metrics strongly correlate with human preferences. Our source code is publicly available.