keyword

Social Bias Frames

Social Bias Frames is a structured formalism in natural language processing used to identify, represent, and explain the implicit biases, stereotypes, and power dynamics conveyed in human communication. Rather than relying on simple binary classifications to label text merely as offensive or inoffensive, this framework captures the nuanced pragmatic and contextual layers of language. It systematically maps an utterance to multiple explanatory components, including whether a statement is intended to offend, which demographic group is targeted, whether the statement reflects in-group language usage, and the specific underlying stereotype or harmful implication being projected. By formalizing both categorical assessments of harm and the explicit inferences behind implied meanings, Social Bias Frames enable computational models to reason more holistically about subtle prejudice and social inequality in discourse.

1 item

When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks

Eve Fleisig, Rediet Abebe, Dan Klein

Why you should read this

Presents a framework that predicts individual annotator judgments alongside the targeted demographic groups of text to identify when targeted populations disagree with majority-vote labels in offensive content detection.

People often disagree on subjective tasks such as determining what is offensive or toxic online, where each annotator brings their own perspective influenced by factors like culture, identity, and lived experience. In many annotation settings today, however, we ask multiple people — who may have different beliefs — to provide just one label per example, treating majority vote as ground truth. We show how training models to capture individual annotator behavior instead yields better modeling of disagreement patterns among human raters.

Source

https://aclanthology.org/2023.emnlp-main.415.pdf

Added

2026-10-02