Publications

A mixed-methods user study with autistic and neurotypical adults examining how LLMs like ChatGPT exhibit social and non-social biases toward neurodiverse users. Findings reveal that LLMs rely on medicalized generalizations, place communication burden on autistic users, and default to sensory-overloading text formats. Proposes actionable design implications including "Tone Toggles" and inquiry-first architectures for cognitive accessibility.