Disability-First Dataset

An approach to AI dataset creation, articulated by Theodorou et al. and others, that treats serving a disability community as the primary objective rather than collecting disability data as a minority slice of a general-purpose dataset. Examples include VizWiz (blind photographers), Orbit (few-shot object recognition for BLV users), ASL Citizen (Deaf signers), and AS-70 (Mandarin stuttered speech). Disability-first datasets are typically sourced directly from disabled contributors, designed around their communication practices and access needs, and evaluated against tasks the community itself prioritises. Tang et al. (CHI 2026) extend the principle to dataset annotation.