Modeling Annotator Disagreement with Demographic-Aware Experts and Synthetic Perspectives

TL;DR

This paper introduces DEM-MoE, a demographic-aware model for capturing annotator disagreement in subjective NLP tasks, and explores synthetic data generation via LLMs to enhance training data diversity.

Contribution

It presents DEM-MoE, a novel architecture that incorporates demographic information for better modeling of annotator disagreement, and investigates synthetic annotations for data augmentation.

Findings

DEM-MoE outperforms prior models across demographic groups.

Synthetic annotations from LLMs align moderately with human judgments.

Blending real and synthetic data improves model performance depending on dataset structure.

Abstract

We present an approach to modeling annotator disagreement in subjective NLP tasks through both architectural and data-centric innovations. Our model, DEM-MoE (Demographic-Aware Mixture of Experts), routes inputs to expert subnetworks based on annotator demographics, enabling it to better represent structured, group-level variation compared to prior models. DEM-MoE consistently performs competitively across demographic groups, and shows especially strong results on datasets with high annotator disagreement. To address sparse demographic coverage, we test whether LLM-generated synthetic annotations via zero-shot persona prompting can be used for data imputation. We show these synthetic judgments align moderately well with human annotations on our data and offer a scalable way to potentially enrich training data. We then propose and evaluate approaches for blending real and synthetic data using strategies tailored to dataset structure. We find that the optimal strategies depend on dataset structure. Together, these contributions improve the representation of diverse perspectives.