A Survey on Stereotype Detection in Natural Language Processing

TL;DR

This survey maps stereotype detection in NLP, detailing definitions, datasets, methods, and evaluation challenges. It integrates psychology-grounded theories (Stereotype Content Model and Agency-Beliefs-Communion) with computational analyses to trace how stereotypes are represented and mitigated in language technologies. Key contributions include a taxonomy of literature, a curated inventory of stereotype-related datasets across languages, and an overview of debiasing and detection approaches, alongside highlighted challenges in definitions, data quality, and generalizability. The work emphasizes the societal importance of early stereotype detection to prevent escalation to hate speech and advocates for multilingual, intersectional, and ethically grounded research to improve fairness in NLP systems.

Abstract

Stereotypes influence social perceptions and can escalate into discrimination and violence. While NLP research has extensively addressed gender bias and hate speech, stereotype detection remains an emerging field with significant societal implications. In this work is presented a survey of existing research, analyzing definitions from psychology, sociology, and philosophy. A semi-automatic literature review was performed by using Semantic Scholar. We retrieved and filtered over 6,000 papers (in the year range 2000-2025), identifying key trends, methodologies, challenges and future directions. The findings emphasize stereotype detection as a potential early-monitoring tool to prevent bias escalation and the rise of hate speech. Conclusions highlight the need for a broader, multilingual, and intersectional approach in NLP studies.

Figures (7)

• Figure 1: The Pyramid of Hate.
• Figure 2: Pipeline describing the different steps of the survey process.
• Figure 3: Topics extracted with BERTopic from the text of the abstracts.
• Figure 4: Number of publications per field of study.
• Figure 5: Most frequent bigrams in Titles and Abstracts.