Trigger Warnings Are Grounded in a Shared Vocabulary: A Corpus Analysis with User-Generated Labels

Abstract

Trigger warnings advise of potentially disturbing content. On that note: This document discusses abuse. But can we trust trigger warnings? For a warning to be credible, independent authors must have a shared understanding of the type of content that advises caution. We investigate for the first time whether trigger warnings are aligned with the vocabulary of texts written by uncoordinated authors. To quantify the lexical alignment of trigger warnings, we conduct a series of statistical tests on the texts of fan fiction authors who used warnings relating to emotional, physical, or sexual abuse. We find that the vocabulary of texts with these warnings is aligned with a curated dictionary of terms related to abuse. However, a high frequency of a term in texts with a warning does not necessarily indicate a semantic relation.

Resources

Details