Learning Moral Diversity: Modelling Individual Perspectives in Moral Classification of Texts

2026-06-21

AI Summary

Traditional approaches to moral text classification typically aggregate annotator labels, thereby overlooking the inherent subjectivity and individual variation in moral judgments and failing to capture the true diversity of human moral perspectives. This work proposes a personalized modeling paradigm based on pretrained language models, incorporating annotator-specific embedding layers to explicitly represent individual moral viewpoints. The proposed method substantially improves prediction accuracy for individual annotators’ moral judgments and yields interpretable representations of their moral perspectives. To the best of our knowledge, this study is the first to systematically demonstrate both the necessity and effectiveness of accounting for individual differences in moral classification tasks.

Abstract

Understanding moral values in social media text offers insight into moral judgement formation, and supervised NLP models trained on crowdsourced data have achieved strong classification performance. However, most approaches simplify the problem by aggregating multiple annotators' labels into a single "ground truth", overlooking the inherent subjectivity of the task. In practice, there are disagreements between annotators caused by personal viewpoint or inherent ambiguities, particularly for short tweets. Here, we extend a pretrained language model with a layer that learns annotator-specific features. Our model improves predictions of individual annotations and yields representations that reveal meaningful insights into annotators' moral perspectives. We show that models trained on aggregated labels may hide variation and give a misleading impression of performance. Overall, we demonstrate that disagreement reflects the inherent subjectivity of the task and that modelling individual perspectives creates benefits for moral classification of texts.

Problem

• moral classification
• subjectivity
• annotator disagreement
• crowdsourced labels
• moral diversity

Innovation

• moral classification
• annotator-specific modeling
• subjectivity
• pretrained language model
• crowdsourced data