Incorporating Human Explanations for Robust Hate Speech Detection

TL;DR

This paper explores how incorporating human explanations and a new Stereotype Intent Entailment task can improve the robustness and contextual understanding of hate speech detection models based on large transformer language models.

Contribution

It introduces the Stereotype Intent Entailment task and demonstrates its effectiveness in enhancing model understanding of implicit stereotypes in hate speech detection.

Findings

SIE improves content understanding in hate speech models

Modeling contextually grounded stereotypes is crucial

Challenges remain in capturing implicit intent

Abstract

Given the black-box nature and complexity of large transformer language models (LM), concerns about generalizability and robustness present ethical implications for domains such as hate speech (HS) detection. Using the content rich Social Bias Frames dataset, containing human-annotated stereotypes, intent, and targeted groups, we develop a three stage analysis to evaluate if LMs faithfully assess hate speech. First, we observe the need for modeling contextually grounded stereotype intents to capture implicit semantic meaning. Next, we design a new task, Stereotype Intent Entailment (SIE), which encourages a model to contextually understand stereotype presence. Finally, through ablation tests and user studies, we find a SIE objective improves content understanding, but challenges remain in modeling implicit intent.