Make Up Your Mind! Adversarial Generation of Inconsistent Natural Language Explanations

TL;DR

This paper reveals that neural models generating natural language explanations can produce inconsistent outputs and introduces an adversarial framework to detect and analyze these inconsistencies, highlighting flaws in model reasoning.

Contribution

The authors propose a simple adversarial framework to identify and analyze inconsistencies in natural language explanations generated by neural models, including addressing full sequence adversarial attacks.

Findings

Neural explanation models can produce mutually inconsistent explanations.

The framework effectively detects inconsistencies in state-of-the-art models.

Models are capable of generating a significant number of inconsistent explanations.

Abstract

To increase trust in artificial intelligence systems, a promising research direction consists of designing neural models capable of generating natural language explanations for their predictions. In this work, we show that such models are nonetheless prone to generating mutually inconsistent explanations, such as "Because there is a dog in the image" and "Because there is no dog in the [same] image", exposing flaws in either the decision-making process of the model or in the generation of the explanations. We introduce a simple yet effective adversarial framework for sanity checking models against the generation of inconsistent natural language explanations. Moreover, as part of the framework, we address the problem of adversarial attacks with full target sequences, a scenario that was not previously addressed in sequence-to-sequence attacks. Finally, we apply our framework on a state-of-the-art neural natural language inference model that provides natural language explanations for its predictions. Our framework shows that this model is capable of generating a significant number of inconsistent explanations.