The Anatomy of Socio-Demographics in Pluralistic Offensiveness Modeling: When and Where It Helps, and the Role of Reasoning

A study posted to arXiv (2610.37896v1) finds that socio-demographic information yields small and inconsistent gains for individual-level offensiveness predictions from prompted and trained LLMs, but clearer gains at the group and population levels because aggregation attenuates idiosyncratic variation. Across three multi-annotator datasets, the researchers report that a reasoning-and-prediction format decomposing each judgment into an overall baseline plus a demographic-specific deviation improves both components, and that explicit reasoning about how text interacts with annotator demographics produces more accurate baseline, shift and final estimates than direct demographic conditioning, which rarely yields informative shifts. The authors suggest the mixed evidence in prior work may partly stem from individual-level evaluation.

arXiv:2610.37896v1 Announce Type: new Abstract: Subjective judgments such as offensiveness often vary substantially across individuals, motivating pluralistic modeling approaches that incorporate annotators' socio-demographic information to capture this variation. However, prior findings on the benefits of socio-demographic information for pluralistic modeling are mixed, calling for a systematic analysis of both its effectiveness and the mechanisms through which it improves prediction. In this work, we study $\textbf{when}$ and $\textbf{where}$ socio-demographics benefit pluralistic offensiveness modeling, as well as $\textbf{the role of explicit reasoning}$ in leveraging this information. Across three multi-annotator datasets with prompted and trained LLMs, we first show that demographic gains are small and inconsistent for individual predictions but become clearer at the group and population levels, because aggregation attenuates the idiosyncratic variation that masks small but systematic demographic signals. Second, using a reasoning and prediction format that explicitly decomposes individual judgment into an overall baseline and a demographic-specific deviation, we show that the gains arise from both components: demographic information improves baseline estimation and enables meaningful demographic-specific shifts, and interventions show that better component estimates improve final predictions. Finally, comparing direct demographic conditioning with explicit reasoning about how the text interacts with annotator demographics, we find that reasoning yields more accurate baseline and shift estimates as well as final predictions, whereas conditioning alone rarely produces informative shifts. These results suggest that the mixed evidence in prior work may partly stem from individual-level evaluation, and that the value of socio-demographic information depends on both how it is evaluated and how models are guided to use it.