Modeling Annotator Disagreement with Demographic-Aware Experts and Synthetic Perspectives
This work addresses the problem of representing diverse perspectives in NLP for researchers and practitioners, but it is incremental as it builds on prior models and data augmentation techniques.
The paper tackled modeling annotator disagreement in subjective NLP tasks by introducing DEM-MoE, a demographic-aware model that routes inputs based on annotator demographics, and by using LLM-generated synthetic annotations for data imputation, showing competitive performance across demographic groups and moderate alignment of synthetic data with human annotations.
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.