Are Shortest Rationales the Best Explanations for Human Understanding?

TL;DR

This paper questions whether the shortest rationales are the most understandable for humans, introducing a flexible self-explaining model and conducting a user study to evaluate the impact of rationale length on human prediction accuracy.

Contribution

It presents LimitedInk, a self-explaining model allowing variable rationale lengths, and empirically investigates how rationale length affects human interpretability and prediction performance.

Findings

Short rationales do not improve human prediction over random masks.

LimitedInk achieves comparable task performance and rationale agreement.

Longer rationales are more helpful for human understanding.

Abstract

Existing self-explaining models typically favor extracting the shortest possible rationales - snippets of an input text "responsible for" corresponding output - to explain the model prediction, with the assumption that shorter rationales are more intuitive to humans. However, this assumption has yet to be validated. Is the shortest rationale indeed the most human-understandable? To answer this question, we design a self-explaining model, LimitedInk, which allows users to extract rationales at any target length. Compared to existing baselines, LimitedInk achieves compatible end-task performance and human-annotated rationale agreement, making it a suitable representation of the recent class of self-explaining models. We use LimitedInk to conduct a user study on the impact of rationale length, where we ask human judges to predict the sentiment label of documents based only on LimitedInk-generated rationales with different lengths. We show rationales that are too short do not help humans predict labels better than randomly masked text, suggesting the need for more careful design of the best human rationales.