AI and Culture in AAC: Opportunities and Challenges for Culturally Responsive Symbol Design

Abstract

Augmentative and Alternative Communication (AAC) provides support for individuals with complex communication needs through a range of multimodal approaches [1].  In some cases, pictographic symbol sets can offer a foundation for communication, language and literacy development.  However, concerns have been raised about the cultural relevance of some symbols when they are used across diverse communities and contexts.

Whilst symbol libraries continue to expand, advances in generative artificial intelligence (AI) presents new opportunities for rapid creation and adaptation of symbols working alongside manually designed collections.  AI models can generate personalised visual representations from text prompts, with the potential to reflect local environments, cultural identities and individual preferences.

However, although AI-generated imagery offers promise, it also raises concerns [2]. Current AI models are trained on datasets that may under-represent some communities whilst over-representing others. Visual cultural stereotypes, inappropriate depictions and inaccurate portrayals of disability [3] can all occur. Furthermore, symbols generated through AI may appear visually acceptable without being meaningful or acceptable to AAC users. Moreover, successful image production often requires careful prompting, review and editing, particularly when representing people, social settings, clothing, food and culturally significant occasions.

This presentation explores the opportunities and challenges of using AI to support culturally responsive symbol design. Drawing on experiences from the development and evaluation of AI-supported symbol creation, examples will illustrate how cultural descriptors can alter aspects of dress, community settings and general visual appearance. These examples will also demonstrate situations where AI-generated symbols require human review due to bias, ambiguity or reduced semantic accuracy.

To underpin future developments, a set of guidelines for ethical and inclusive AI-supported symbol generation is proposed. The guidance consists of five interconnected principles: representation, participation, transparency, adaptability and evaluation.

Representation focuses on ensuring that symbol generation processes aim to include diverse cultures and languages rather than reinforcing dominant perspectives. Participation recognises the importance of involving AAC users, families, communication partners and professionals in the review and development process. Transparency requires openness regarding how symbols are generated, selected and edited, together with acknowledgement of the limitations of AI models. Adaptability supports personalisation and local cultural variation whilst maintaining symbol clarity and usability. Evaluation emphasises the ongoing review of symbols for visual clarity, semantic alignment, cultural appropriateness and user acceptance.

It is important to position AI as a tool to support, rather than replace human-centred AAC practice. Whilst AI has the potential to extend the cultural range, personal relevance and availability of pictographic symbols worldwide, meaningful involvement of AAC users and their communities remains essential.

References

1. American Speech-Language-Hearing Association (ASHA) Augmentative and Alternative Communication (AAC). Available at: https://www.asha.org/Practice-Portal/Professional-Issues/Augmentative-and-Alternative-Communication/ Last Accessed: 24 Jan 2026.
2. Light, J., McNaughton, D., Beukelman, D., Fager, S.K., Fried-Oken, M., Jakobs, T. and Jakobs, E., Challenges and opportunities in augmentative and alternative communication: Research and technology development to enhance communication and participation for individuals with complex communication needs. Augmentative and alternative communication, 35(1), pp.1-12. (2019)
3. Tevissen, Y., Disability representations: Finding biases in automatic image generation. arXiv preprint arXiv:2406.14993. (2024)