Machine learning is recently used to detect hate speech and other forms of abusive language in online platforms. However, a notable weakness of machine learning models is their vulnerability to bias, which can impair their performance and fairness. One type is annotator bias caused by the subjective perception of the annotators. In this work, we investigate annotator bias using classification models trained on data from demographically distinct annotator groups. To do so, we sample balanced subsets of data that are labeled by demographically distinct annotators. We then train classifiers on these subsets, analyze their performances on similarly grouped test sets, and compare them statistically. Our findings show that the proposed approach successfully identifies bias and that demographic features, such as first language, age, and education, correlate with significant performance differences.

A challenge that many online platforms face is hate speech or any other form of online abuse. To cope with this, hate speech detection systems are developed based on machine learning to reduce manual work for monitoring these platforms. Unfortunately, machine learning is vulnerable to unintended bias in training data, which could have severe consequences, such as a decrease in classification performance or unfair behavior (e.g., discriminating minorities). In the scope of this study, we want to investigate annotator bias -a form of bias that annotators cause due to different knowledge in regards to the task and their subjective perception. Our goal is to identify annotation bias based on similarities in the annotation behavior from annotators. To do so, we build a graph based on the annotations from the different annotators, apply a community detection algorithm to group the annotators, and train for each group classifiers whose performances we compare. By doing so, we are able to identify annotator bias within a data set. The proposed method and collected insights can contribute to developing fairer and more reliable hate speech classification models.

Recently, numerous datasets have been produced as research activities in the field of automatic detection of abusive language or hate speech have increased. A problem with this diversity is that they often differ, among other things, in context, platform, sampling process, collection strategy, and labeling schema. There have been surveys on these datasets, but they compare the datasets only superficially. Therefore, we developed a bias and comparison framework for abusive language datasets for their in-depth analysis and to provide a comparison of five English and six Arabic datasets. We make this framework available to researchers and data scientists who work with such datasets to be aware of the properties of the datasets and consider them in their work.