"I use ChatGPT to humanize my words": Affordances and Risks of ChatGPT to Autistic Users

TL;DR

The paper investigates how autistic users perceive the utility and risks of ChatGPT through the Technology Affordance framework, using an inductive thematic analysis of 3,984 social media posts. It identifies four main affordances—external cognitive scaffolding, translation between neurodivergent and neurotypical communication, emotional regulation, and validation of autistic identity—and three concurrent risks: reinforcement of delusional thinking, automated masking erasing authentic voice, and ethical conflicts with the autistic sense of justice. The authors propose neuro-inclusive design directions emphasizing beneficial friction, Cognitive Forcing Functions, and bidirectional translation to balance support with user agency. This work highlights the need for AI systems that scaffold cognition without eroding autonomy or authentic identity, informing the development of more inclusive LLM chatbots.

Abstract

Large Language Model (LLM) chatbots like ChatGPT have emerged as cognitive scaffolding for autistic users, yet the tension between their utility and risk remains under-articulated. Through an inductive thematic analysis of 3,984 social media posts by self-identified autistic users, we apply the Technology Affordance framework to examine this duality. We found that while users leveraged ChatGPT to offload executive dysfunction, regulate emotions, translate neurotypical communication, and validate their autistic identity, these affordances coexist with significant risks: reinforcing delusional thinking, erasing authentic identity through automated masking, and triggering conflicts with the autistic sense of justice. This poster identifies these trade-offs in autistic users' interactions with ChatGPT and concludes by outlining our future work on developing neuro-inclusive technologies that address these tensions through beneficial friction and bidirectional translation.