Large language models for autistic and neurodivergent individuals: Concerns, benefits and the path forward
Chris Papadopoulos
Abstract
The emergence of large language models (LLMs) like ChatGPT has transformed access to information, offering immediate responses to diverse queries. While these technologies present significant opportunities, they also pose unique challenges for the autistic and wider neurodivergent community. This commentary critically examines the implications of LLMs within this community, focusing on potential misinformation, reinforcement of stigmatising medical models, misrepresentation of non-speaking neurodivergent individuals, and issues of equity and privacy. Through personal experimentation with LLMs and discussions with community members, concerns are raised about the perpetuation of stereotypes, the risk of over-reliance on potentially flawed AI outputs, and the inadequate support for diverse communication needs. Conversely, LLMs offer transformative benefits such as accessible information, personalised learning, and supportive tools for communication and professional development. To maximise these benefits while mitigating risks, a multifaceted approach is proposed, emphasising awareness, collaborative development with neurodivergent individuals, ongoing feedback mechanisms, policy advocacy, and positive representation.