Fascist-o-meter: Neo-fascist Discourse Classifier for the United States of America Societal Frame

ACL ARR 2025 February Submission6488 Authors

Abstract:

Neo-fascism is a political and societal ideology that has been having remarkable growth in the last decade in the United States of America (USA), as well as in other Western societies. It poses a grave danger to democracy and the minorities it targets, and it requires active actions against it to avoid escalation. This work presents the first-of-its-kind neo-fascist classifier for digital discourse in the USA societal context. Using notable neo-fascist groups (the forums of Iron March and Stormfront.org) activity on the internet, we collect over 600k posts and messages.
With the help of political science researchers, we develop guidelines to classify neo-fascist online statements. The guidelines are applied to a subset of the collected posts. Through crowdsourcing, we annotate a total of 611 neo-fascist and 389 non-neo-fascist posts.
With this classified data set, we fine-tune and test both Small Language Models (SLMs) and Large Language Models (LLMs). HomophobiaBERT is the top-performing SLM, achieving an F1 score of 83.87\%. Among LLMs, GPT-4o leads with an F1 score of 86.80\% in a few-shot modality, while Gemma2 achieves a comparable F1 score of 83.70\%, indicating its competitive performance.
We find that the prevalence of neo-fascist rhetoric in this kind of forum is ever-present, making them a good target for future research. The societal frame is a key consideration for neo-fascist speech when conducting natural language processing research. Finally, the work against this kind of political movement must be pressed upon and continued for the well-being of a democratic society.

Paper Type: Long

Research Area: Computational Social Science and Cultural Analytics

Research Area Keywords: human behavior analysis, stance detection, language/cultural bias analysis, sociolinguistics, NLP tools for social analysis

Contribution Types: Publicly available software and/or pre-trained models, Data resources, Data analysis

Languages Studied: English

Submission Number: 6488