Human-annotated rationales and explainable text classification: a survey

TL;DR

This paper reviews how human explanations for classifications can improve data quality and help build more explainable AI models.

Contribution

The paper provides a survey on the collection and use of human-annotated rationales for explainable text classification.

Findings

Human-annotated rationales improve data quality and model performance.

They serve as a benchmark for evaluating model-generated explanations.

Rationales are crucial for advancing explainable artificial intelligence.

Abstract

Asking annotators to explain “why” they labeled an instance yields annotator rationales: natural language explanations that provide reasons for classifications. In this work, we survey the collection and use of annotator rationales. Human-annotated rationales can improve data quality and form a valuable resource for improving machine learning models. Moreover, human-annotated rationales can inspire the construction and evaluation of model-annotated rationales, which can play an important role in explainable artificial intelligence.

Figures