Semantic Gradients Interactions in SSD: A Case Study in Racial Identity and Hate Speech
2026-05-26 • Computation and Language
Computation and Language
AI summaryⓘ
The authors extended a method called Supervised Semantic Differential (SSD) to create Interaction SSD, which helps study how the meaning of words changes based on different groups or conditions, like people's traits or backgrounds. They apply this to data on hate speech to see if annotators' racial identities affect how they judge hateful comments. Their model finds that while there is a common way people view hate speech, there are also smaller differences tied to group identity. This new method allows researchers to test and understand how meaning changes across different groups in a clear, statistical way.
Supervised Semantic Differentialsemantic gradientinteraction gradienthate speech annotationmoderator variablessemantic meaningstatistical moderationgroup differencesannotator bias
Authors
Felix Ostrowicki, Hubert Plisiecki
Abstract
We introduce interaction SSD, an extension of Supervised Semantic Differential that models how semantic meaning varies across moderators such as groups, traits, or conditions making this variation testable and interpretable. The method estimates a main semantic gradient, an interaction gradient, and conditional gradients, all interpretable through standard SSD tools. We illustrate it on the UC Berkeley Measuring Hate Speech corpus, testing whether annotator racial identity moderates hate-speech judgments of comments targeting people of color. The interaction model detects a significant moderation effect: the shared gradient contrasts dehumanizing hostility with counter-speech, while the interaction gradient reveals smaller group-linked differences in which semantic cues predict hate-speech ratings. Interaction SSD makes moderated meaning-outcome relationships statistically testable and interpretable.