SMART-TBI: Design and Evaluation of the Social Media Accessibility and Rehabilitation Toolkit for Users with Traumatic Brain Injury
Yaxin Hu, Hajin Lim, Lisa Kakonge, Jade T. Mitchell, Hailey L. Johnson, Lyn S. Turkstra, Melissa C. Duff, Catalina L. Toma, Bilge Mutlu · 2024 · ASSETS '24: Proceedings of the 26th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/3663548.3675641
Summary
This paper introduces SMART-TBI (Social Media Accessibility and Rehabilitation Toolkit), a browser extension designed to make Facebook more accessible for people with traumatic brain injury (TBI). TBI can cause a range of cognitive and communication challenges — including difficulties with attention, memory, executive function, social cognition, reading comprehension, and written expression — that create significant barriers to social media participation. Yet social media is a critical channel for maintaining social connections, which is especially important for people with TBI who often experience social isolation after injury. SMART-TBI comprises five aids targeting specific TBI-related barriers. The Writing Aid uses GPT-4 to help users compose posts and comments by suggesting improvements to grammar, clarity, tone, and appropriateness — addressing difficulties with written expression and social pragmatics. The Interpretation Aid provides AI-generated explanations of potentially confusing content like sarcasm, idioms, and ambiguous posts — addressing challenges with social cognition and pragmatic language comprehension. Filter Mode allows users to hide potentially distressing content (political posts, negative news, triggering images) — addressing emotional dysregulation and sensitivity common after TBI. Focus Mode reduces visual clutter by hiding non-essential interface elements like sidebars, suggested content, and notifications — addressing attention and executive function difficulties. Facebook Customization provides interface modifications like larger text, simplified layouts, and reduced animations. The toolkit was evaluated with eight participants with moderate-to-severe TBI and five TBI rehabilitation experts (speech-language pathologists and occupational therapists) through demonstration sessions and semi-structured interviews.
Key findings
Participants and experts had mixed but generally positive reactions to the toolkit. The Focus Mode and Filter Mode were the most universally valued — participants described Facebook as "overwhelming" and appreciated the ability to strip it down to essential content. One participant noted that they had stopped using Facebook entirely due to sensory overload, and Focus Mode might enable them to return. The Writing Aid received the most complex reactions: some participants found it empowering to have AI-assisted writing that caught errors they might miss due to cognitive fatigue, while others felt it threatened their autonomy and authenticity ("that's not my voice"). Rehabilitation experts raised concerns about the Writing Aid potentially masking deficits that clinicians need to monitor for recovery tracking. The Interpretation Aid was valued for explaining sarcasm and idioms but raised privacy concerns — participants were uncomfortable with AI reading their social media conversations. A critical theme was psychological safety: participants with TBI were acutely aware of their cognitive changes and feared being judged, making them cautious about AI tools that might highlight their difficulties or share information about their impairments. Privacy control emerged as essential — participants wanted granular control over what the AI could access and who could know they were using assistive tools. The tension between business goals (engagement-maximising algorithms, notifications, suggested content) and accessibility needs was explicitly identified: many of Facebook's design patterns that drive engagement are precisely the features that create barriers for users with cognitive impairments.
Relevance
SMART-TBI addresses an underserved population in accessibility research — people with acquired cognitive disabilities from TBI, who number in the millions globally but receive far less attention than sensory or motor disabilities in digital accessibility work. The toolkit demonstrates that cognitive accessibility on social media requires fundamentally different interventions than physical or sensory accessibility: reducing complexity, supporting comprehension, scaffolding communication, and filtering emotionally triggering content. The tension between AI assistance and user autonomy is particularly nuanced for TBI — unlike correcting a typo, AI-rewriting someone's social media post touches on identity, voice, and the complex relationship between a person and their changed cognitive abilities. For accessibility practitioners, the finding that platform engagement patterns directly conflict with cognitive accessibility needs is a systemic critique: social media platforms designed to maximise attention capture are inherently hostile to users with attention and executive function impairments. The involvement of rehabilitation experts provides a clinical perspective often missing from HCI accessibility research, highlighting concerns about AI tools masking deficits and the importance of designing assistive tools that support rather than replace cognitive rehabilitation goals.
Tags: traumatic brain injury · social media accessibility · cognitive accessibility · Facebook · AI-assisted communication · rehabilitation · browser extension · social participation