Designing a LLM‑Enhanced Storytelling AAC Tool for Children with Complex Communication Needs

Abstract

Traditional Electronic Graphic Symbol-Based (GS-based) Augmentative and Alternative Communication (AAC) systems for individuals with Complex Communication Needs (CCN) historically focus on functional communication, such as basic requests and urgent needs. While essential, these static pictogram grids often fail to foster creativity, emotional expression, or active social participation. To address these limitations, this paper introduces an innovative, AI-enhanced, narrative-focused tablet application designed to empower children with CCN to co-construct dynamic and personalised stories. The application utilises an agentic generative AI framework powered by Large Language Models (LLMs) to interpret a child’s selected pictogram tokens and transform them into multi-page, natural-language narratives. To reduce cognitive overload, caregivers can curate tailored card subsets utilising ARASAAC pictograms and personal photos. The system then generates a multimodal story that matches the child’s specific visual and linguistic processing profile, incorporating AI-generated illustrations, text, and text-to-speech audio. We present the core user requirements and design implications gathered from qualitative interviews with clinical experts and therapists. Additionally, we provide a comprehensive usage scenario and preliminary specialist feedback evaluating the tool’s accessibility, therapeutic value, and the balance between user autonomy and content control. By synthesising clinical insights into reusable software specifications, this work offers a foundational design framework for future researchers and developers engineering assistive, generative technologies for vulnerable populations.