Exploring the Usability of Gaze-based Mobile Communication in Ghana
Victoria Austin, Gifty Ayoka, Giulia Barbareschi, Richard Cave, Catherine Holloway · 2025 · ASSETS 2025: 27th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/3663547.3746395
Summary
This study evaluates the usability of Look to Speak, a free Android-based eye-gaze AAC application developed by Google Creative Lab, for people with severe communication disabilities in Ghana. The research involved training 10 local speech and language therapists (SLTs) and conducting onboarding and evaluation sessions with 15 clients who had limited or no functional speech due to conditions including cerebral palsy (12 participants), stroke (2), and Parkinson's disease (1). Look to Speak enables users to select pre-written phrases or images by directing their gaze left, right, or up while the smartphone's front camera tracks eye movements. The study was grounded in feminist HCI, postcolonialism, and disability justice frameworks, with the researchers explicitly positioning themselves as seeking to understand what works and what does not rather than assuming the technology would be beneficial. Data collection included semi-structured interviews with clients, caregivers, and SLTs, video-recorded onboarding sessions, and one month of SLT monitoring. The research team included members based in Ghana, the UK, and Japan, with some members experiencing communication impairments themselves. The study represents one of the first evaluations of gaze-based AAC technology in sub-Saharan Africa, addressing a critical gap given that 28-49% of people globally with communication disabilities lack access to needed assistive technology, with access rates as low as 3% in some Global South countries.
Key findings
Despite initial enthusiasm from all participants, Look to Speak largely failed to deliver substantial communication benefits for most users. The findings reveal a complex web of interacting barriers: (1) The application's interaction design was counter-intuitive — users must look away from an item to select it, and the multi-step binary selection process (up to 3 steps to choose from 16 items) created high cognitive load. (2) Language was a major barrier — Look to Speak only speaks English and does not support Ghanaian languages; SLTs experimented with phonemic spelling of Ghanaian words with mixed results. (3) Default content was culturally irrelevant, requiring extensive customization before the app could be useful. (4) Five participants preferred using the touchscreen rather than eye gaze, yet the app did not optimize its interface for touch interaction (e.g., still detecting and responding to unintended eye movements). (5) Lack of adequate postural support and wheelchair access prevented several participants from maintaining the stable positioning required for eye tracking. (6) The application provided poor feedback when things went wrong — no guidance on improving positioning, no indication of why selections failed. (7) SLTs had limited knowledge of gaze-based AAC before the study, making it difficult to assess which clients would benefit. Six of 15 clients demonstrated significant cognitive communication difficulties that made the application unsuitable. Some powerful moments did occur — one participant (P1) used the app to say "Mum, I love you" for the first time, creating an emotionally powerful experience for her family.
Relevance
This study provides a critical reality check on the deployment of high-tech assistive technologies in the Global South. For accessibility practitioners, the key takeaway is that technology alone is insufficient — a functioning ecosystem including trained professionals, appropriate seating/positioning equipment, culturally and linguistically relevant content, and ongoing support is essential for AAC to succeed. The finding that a free smartphone app still failed most users highlights that cost reduction alone does not solve accessibility barriers. The paper's design recommendations are broadly applicable: AAC apps need local language support (or at minimum, user-recorded audio), culturally appropriate default content, consistent interaction modes, meaningful error feedback, and integration with broader care pathways. The SMART matrix framework (micro/meso/macro levels) provides a useful lens for understanding why assistive technology deployments fail in low-resource settings. Limitations include the small sample size, short monitoring period, and focus on a single application.
Tags: augmentative communication · AAC · eye gaze · Global South · Ghana · smartphone accessibility · cerebral palsy · communication disability · assistive technology ecosystem · low-resource setting