[Paper Review] Aging Up AAC: An Introspection on Augmentative and Alternative Communication Applications for Autistic Adults

Lara J. Martin, Malathy Nagalakshmi|arXiv (Cornell University)|Apr 26, 2024

Assistive Technology in Communication and MobilityHealth Professions3 citations

TL;DR

This study investigates the needs of autistic adults using high-tech Augmentative and Alternative Communication (AAC) tools through in-depth interviews with 12 participants. It identifies nine key themes—ranging from input/output preferences to social stigma and lack of control—and proposes targeted guidelines for NLP researchers, AAC developers, and policymakers to create more inclusive, effective, and user-centered AAC applications tailored to autistic adults.

ABSTRACT

High-tech Augmentative and Alternative Communication (AAC) has been rapidly advancing in recent years due to the increased use of large language models (LLMs) like ChatGPT, but many of these techniques are integrated without the inclusion of the users' perspectives. Autistic adults have been particularly neglected in the design of AAC tools. We conducted in-depth interviews with 12 autistic adults to find the pain points of current AAC and determine what technological advances they might find helpful. We found 8 different categories of themes from our interviews: input flexibility, output flexibility, selecting or adapting AAC, contexts for AAC use, benefits, access as an adult, stumbling blocks for continued use, and control of communication. In this paper, we go through these categories in depth -- comparing each to prior work -- and then highlight novel findings to suggest possible research directions.

Motivation & Objective

• To understand the challenges autistic adults face when using existing AAC applications, especially those designed for children.
• To identify unmet technological and social needs of autistic adults using high-tech AAC tools.
• To explore how autistic identity influences AAC use and preferences.
• To evaluate autistic adults' attitudes toward integrating NLP techniques like large language models into AAC applications.
• To develop actionable, user-informed guidelines for NLP researchers, AAC developers, and policy makers to improve AAC accessibility and effectiveness for autistic adults.

Proposed method

• Conducted 12 in-depth, semi-structured interviews with autistic adults recruited online, focusing on personal AAC experiences and preferences.
• Used thematic analysis to identify nine recurring themes from interview data: input options, output options, AAC fit and selection, timing of AAC use, benefits, access, barriers to continued use, social concerns, and lack of control.
• Collected insights on user preferences for NLP features such as word prediction, predictive text, and LLM integration, while assessing privacy and control concerns.
• Developed design and policy guidelines based on user-identified pain points and desired features, targeting three stakeholder groups: NLP researchers, AAC application developers, and policy makers.
• Emphasized participatory design by centering autistic adult voices throughout the research process, ensuring user perspectives directly inform recommendations.
• Proposed a prototype development path to test and refine suggested features in real-world contexts, with iterative feedback loops to validate and refine guidelines.

Figure 1. One participant showing their on-screen keyboard embedded on the homepage in Proloquo2Go .

Experimental results

Research questions

• RQ1 What are the key pitfalls and challenges autistic AAC users encounter with existing AAC applications?
• RQ2 What types of features do autistic adults require or prefer in AAC applications?
• RQ3 How does an autistic person’s identity influence their use and perception of AAC tools?
• RQ4 How do autistic adults feel about the integration of NLP techniques—such as word prediction and large language models—into their AAC applications?

Key findings

• Autistic adults frequently experience a lack of control over their AAC tools, including limited customization and opaque algorithmic behavior, which undermines their autonomy.
• Many participants reported social stigma and anxiety related to using AAC in public, especially when others misinterpret their communication method as a sign of intellectual disability.
• Participants expressed strong preference for AAC tools that support natural language input and output, including word prediction and LLM-based assistance, but with clear privacy safeguards and no data storage without consent.
• A significant number of autistic adults were unaware of built-in accessibility features on smartphones, such as Text Relay, indicating a critical gap in information dissemination.
• Participants emphasized the importance of being able to switch or adapt AAC tools over time, particularly as their communication needs and identities evolve, yet few tools support this flexibility.
• The current design of most AAC applications is child-centric, leading to poor fit for autistic adults who are often already literate and have used oral speech for years, resulting in frustration and disengagement.

Figure 2. A box plot depicting the difference between interview times for participants who primarily used AAC during the interview (right, orange) and those who did not (left, blue). This data does not include the one participant who interviewed completely asynchronously.