From Possibility to Practice: Implementing AI in AAC to Support Personalised, Culturally Responsive Communication

Abstract

Artificial intelligence is moving rapidly from a speculative topic in augmentative and alternative communication (AAC) to a practical set of tools that can support communication through enhanced access, personalisation, vocabulary development, symbol creation, language generation, voice output, and participation. The AAC field has already begun to explore AI-enabled approaches such as predictive text, natural language processing, personalised speech synthesis, multimodal input, image recognition, and generative symbol or vocabulary support. Sennott et al. argue that AI has the potential to transform AAC across both low-tech and high-tech systems, while more recent work on transformer-based prediction and foundation models illustrates how AI may help tailor AAC systems to individual users and contexts.

Yet the implementation of AI in AAC remains uneven. While research and prototypes increasingly demonstrate potential, evidence of everyday use by AAC users, families, practitioners, educators, and service providers is still emerging. The challenge is no longer simply whether AI can be used in AAC, but how it can be implemented ethically, accessibly, safely, and in ways that strengthen rather than replace the agency, identity, and communication rights of AAC users. This is especially important given wider global inequities in access to assistive technology and AAC.

This panel will explore practical examples of AI implementation in AAC, using the development and deployment of AI-supported symbol and language tools to shape wider discussion. Through a focus on how AI is being used to assist the creation of personalised AAC content to better reflect the user’s language, culture, age, environment, interests, and lived experience. The panel will consider how such tools can support AAC professionals, families, teachers, and community partners to respond more quickly to communication needs, particularly where existing symbol sets or vocabulary resources are limited, culturally inappropriate, unavailable in the user’s language, or difficult to adapt. Emerging systems combine large language models, contextual cues, personalised text-to-speech, or image recognition to support more expressive communication and reduce effort in message generation.

Many studies remain exploratory, but indicate a growing recognition that AI systems can introduce risks, including inaccurate outputs, over-prediction of what a person wants to say, reduced autonomy, cultural bias, privacy concerns, data governance challenges, and unequal access to high-quality tools. Concerns about authorship, control, and communication autonomy are particularly important in AAC. ISAAC’s position on facilitated communication highlights the importance of evidence-based approaches that do not cast doubt on authorship, a principle directly relevant to AI-supported message generation and prediction.

The panel will therefore move beyond demonstration to critical reflection. It will ask what practical implementation looks like when AI is introduced into communication across a range of settings. Panelists may include AAC users, researchers, clinicians, developers, educators, family members, and organisational leaders. The discussion will examine the opportunities and tensions involved in using AI to generate communication content, support vocabulary growth, develop personalised symbols, assist literacy, or create multilingual resources. Particular attention will be given to the principle that AI should support the voice, authorship, and intent of the AAC user, not speak on their behalf with practical consideration given to how AI can help address gaps in symbol vocabularies and culturally relevant AAC resources; human review and co-design, how AAC users can retain control, how to evaluate whether AI improves participation and communication outcomes, privacy and consent and how to make AI-supported AAC available in underserved languages and low- and middle-income countries. The panel will also consider the skills required by practitioners and families, including prompt design, critical review of outputs, safeguarding, bias awareness, and ethical decision-making.

The panel will invite participants to consider what responsible innovation in AI and AAC should look like over the next five years. It will argue that the most useful AI in AAC will not be the most technically impressive, but the most accountable, inclusive, adaptable, and grounded in the rights and preferences of AAC users. By sharing practical experience, research insights, and implementation questions, the panel will support the ISAAC community to shape AI in AAC in ways that promote communication autonomy, equity, and participation.