“Triggering autonomy, not just automation”: Design implications for LLM-based home automation assistants for people with intellectual disability

Smart home technologies offer considerable potential to enhance autonomy, safety, and quality of life for individuals with intellectual disabilities (ID). Yet, configuring and personalizing such systems remain challenging due to complex interfaces and the technical expertise required. Traditionally, End-User Development (EUD) is often proposed as a way for non-experts to autonomously configure home devices. This work investigates the extent to which Large Language Models (LLMs) can provide effective support to people with ID in applying EUD. It explores whether an LLM-powered conversational interface can support individuals with ID in shaping smart home behaviors through natural conversations. Insights from exploratory sessions involving users with ID, caregivers, and stakeholders demonstrate a positive attitude towards this approach and suggest that conversational interaction can foster accessibility of automation programming through iterative clarifications. We conclude by outlining design opportunities and broader implications that can inform future research towards LLM-based EUD interfaces for people with ID.