[Paper Review] "I Use ChatGPT to Humanize My Words": Affordances and Risks of ChatGPT to Autistic Users

Renkai Ma, Ben Z. Zhang+5 more|arXiv (Cornell University)|Jan 25, 2026

Digital Mental Health Interventions0 citations

TL;DR

The paper analyzes autistic users' interactions with ChatGPT via social media posts to identify both cognitive support benefits and associated risks, using a Technology Affordance lens.

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 a technology affordance lens 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 risks to their well-being: reinforcing delusional thinking, erasing authentic identity through automated masking, and triggering conflicts with the autistic sense of justice. As part of our preliminary work, this poster identifies 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, bidirectional translation, and the delineation of emotional validation from reality.

Motivation & Objective

• Understand how autistic users perceive and use ChatGPT as a cognitive scaffold and communication aid.
• Identify the benefits (executive function offloading, emotional regulation, translation, identity validation) and risks (delusion, automated masking, ethical conflicts).
• Apply the Technology Affordance framework to unify benefits and risks in autistic interactions with LLMs.

Proposed method

• Collected 3,984 posts from Reddit, X, and Tumblr (Jan 1, 2023 – Sep 30, 2025) discussing ChatGPT and autism.
• Used an LLM-assisted pipeline (GPT-4o-mini) with Chain-of-Thought prompting to filter posts for relevance to ChatGPT and autism.
• Conducted inductive thematic analysis to derive affordances and risks, with 239 initial affordance codes and 50 risk codes; intercoder reliability established.
• Cited and integrated prior literature to frame affordances and analyze dualities in autistic users’ interactions with LLMs.

Experimental results

Research questions

• RQ1 What are the perceived affordances of LLM chatbots like ChatGPT for autistic users?
• RQ2 What risks accompany these affordances, and how do they interact with autistic experiences and ethics?
• RQ3 How can a technology affordance lens inform neuro-inclusive design for LLMs?

Key findings

• Autistic users leverage ChatGPT to offload executive dysfunction and generate momentum for tasks.
• ChatGPT serves as an emotional regulation tool, providing a neutral space free from judgment.
• It functions as a translator between neurodivergent and neurotypical communication, aiding professional and social interactions.
• Users report Autistic identity validation through algorithmic mirroring of cognitive styles as a source of belonging.
• Risks include reinforcement of delusional thinking, erosion of authentic identity via automated masking, and ethical conflicts with the autistic sense of justice.