Xenophobia Meter: Defining and Measuring Online Sentiment toward Foreigners on Twitter

Abstract

Xenophobia, a form of hatred directed at foreigners, immigrants, and sometimes even people who are just perceived as foreigners, has been flooding social media in recent political climates. In order to capture language related to foreigners and those perceived-as-foreigners (F&PAF) we present the 7-scale Xenophobia Meter, ranging from anti– to pro- F&PAF sentiments with examples and application rationale. We also publish a dataset of over 7,000 tweets labeled according to this meter, from 11 U.S.-based accounts that are on the forefront of defining the rhetoric related to immigration and policy. We apply a number of models to automatically identify xenophobic and F&PAF-related language. We also present findings from qualitative interviews with human annotators about their labeling experiences. While we find xenophobia is a complex social phenomenon to identify by both humans and machine learning algorithms, we hope that our work inspires researchers, policymakers, and the public to learn about xenophobia and to make efforts to shift the rhetoric and policies toward allyship, equity and inclusion.