Supplementary Material for Identifying, Explaining, and Correcting Ableist Language with AI

This repository contains supplemental information for the paper "Identifying, Explaining, and Correcting Ableist Language with AI" by Smith et al., published in the proceedings of the CHI Conference on Human Factors in Computing Systems 2026 (CHI ‘26). Contributions include a novel dataset of narrative-based nuanced ableism annotations crowdsourced from members of the Disability Community on Prolific. Please view the README for details on the contents and creators of the data. The annotations were based on seven AI-generated stories (1 for each of the 7 disability communities studied in this work). The full text for the stories are provided in the file “Nuanced Ableism Story Probes”. Utilization of this dataset is intended for training and evaluating AI models in the detection and correction of nuanced bias, specifically ableism. It can also serve as a resource for developing inclusive writing tools about broader forms of bias and used for educational materials about disability-centered communication. For more information about the motivation of this work, methods of inquiry, and study results, please refer to the full paper.