A Study of Nationality Bias in Names and Perplexity using Off-the-Shelf Affect-related Tweet Classifiers
This work addresses bias in NLP systems for social media analysis, which is critical for fairness in automated content moderation, though it is incremental as it applies existing bias detection methods to new contexts.
The study quantified biases in affect-related tweet classifiers by creating counterfactual examples with perturbations, revealing that country names can influence predictions by up to 23% in hate speech detection and 60% in negative emotion prediction, with biases linked to pre-trained language model training data.
In this paper, we apply a method to quantify biases associated with named entities from various countries. We create counterfactual examples with small perturbations on target-domain data instead of relying on templates or specific datasets for bias detection. On widely used classifiers for subjectivity analysis, including sentiment, emotion, hate speech, and offensive text using Twitter data, our results demonstrate positive biases related to the language spoken in a country across all classifiers studied. Notably, the presence of certain country names in a sentence can strongly influence predictions, up to a 23\% change in hate speech detection and up to a 60\% change in the prediction of negative emotions such as anger. We hypothesize that these biases stem from the training data of pre-trained language models (PLMs) and find correlations between affect predictions and PLMs likelihood in English and unknown languages like Basque and Maori, revealing distinct patterns with exacerbate correlations. Further, we followed these correlations in-between counterfactual examples from a same sentence to remove the syntactical component, uncovering interesting results suggesting the impact of the pre-training data was more important for English-speaking-country names. Our anonymized code is [https://anonymous.4open.science/r/biases_ppl-576B/README.md](available here).