Bots, Brains and Barriers: AI and Autism
Ayhan Alman · Zenodo (CERN European Organization for Nuclear Research) · 2025
This reflective practice article examines the potential of artificial intelligence (AI) to support autistic people in navigating everyday communication and workplace challenges. The author, writing from a position of personal investment in assistive technology, facilitated a community group discussion that drew on two recent empirical studies. The first, by Jang and colleagues, found that autistic workers generally preferred responses from a large language model (LLM) over those from a human when seeking workplace communication advice, though a specialist raised concerns about the quality of some AI guidance. The second, by Carik and colleagues, surveyed 200 autistic participants and found LLMs useful for identifying and reframing negative self-talk, whilst participants noted neurotypical bias and occasionally unhelpful response styles. The author supplemented these findings with a live Turing Test demonstration, in which group members attempted to distinguish AI generated responses from human written ones, revealing that prompt engineering can make AI output appear more natural and empathetic. The article concludes by acknowledging the broader ethical dimensions of AI development, including labour exploitation, environmental costs, and the risk of neurotypical bias, whilst arguing that meaningful and responsible use of these technologies holds genuine promise for autistic communities.