Temp access: Reflecting on multimodal GAI as an accessibility technology for temporary disability
Kate S. Glazko · 2025 · ASSETS 2025: 27th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/3663547.3749830
Summary
This paper presents an autoethnographic account of using multimodal generative AI (GAI) tools as accessibility technology during a period of temporary disability. The author, an accessibility researcher, experienced an illness that simultaneously impacted verbal communication, visual processing, and manual dexterity, creating overlapping access needs that would traditionally require multiple specialized assistive technologies. Instead, the author turned to commercially available GAI tools including ChatGPT, Claude, Gemini, and NotebookLM to address these barriers through a single class of technology. The paper traces the author's experience through three phases: a crisis phase where survival-oriented use took priority over concerns about output quality; a stabilization phase where the author began evaluating trade-offs between accessibility benefits and concerns about academic integrity; and a transition phase where the author navigated continued use as the disability partially resolved. The methodology draws on critical autoethnography within a disability studies framework, positioning the author's lived experience as a valid site of knowledge production. The research context is significant because most assistive technology research focuses on permanent or long-term disability, leaving a gap in understanding how people navigate access needs during temporary impairments where the timeline for recovery is uncertain. The paper contributes to emerging discourse about GAI as an accessibility tool by examining a case where the technology's low barriers to adoption and multimodal flexibility made it uniquely suited to the user's situation.
Key findings
The paper identifies several key tensions in using GAI for temporary disability access. First, GAI tools required no fitting, specialist training, or financial investment, making them immediately adoptable during a health crisis — a stark contrast to traditional AT that requires assessment, procurement, and learning curves that may not align with uncertain temporary disability timelines. Second, multimodal flexibility allowed a single class of tools to address communication, visual processing, and motor access needs simultaneously, avoiding the complexity of assembling multiple specialized AT solutions. Third, the author experienced significant conflict between using GAI as legitimate accommodation and perceiving it as compromising academic integrity, reflecting broader tensions around AI use in professional and academic settings that become especially fraught when intersecting with disability accommodation. Fourth, the paper documents an unintended pattern of continued GAI use after partial recovery, raising questions about where accommodation ends and convenience begins. The author found that once GAI tools were integrated into workflows during the disability period, disentangling legitimate access use from habitual use became difficult. Finally, the paper highlights how temporary disability creates unique identity tensions — the author struggled with disability disclosure and felt caught between identifying as disabled (which would justify accommodation) and expecting recovery (which made claiming disability identity feel illegitimate).
Relevance
This paper opens an important conversation about GAI as informal assistive technology, particularly for the growing population of people with temporary or episodic disabilities who may not qualify for or seek formal AT provision. For accessibility practitioners, the findings highlight that temporary disability represents an underserved use case — most AT ecosystems assume stable, long-term needs, while GAI tools may serve as a bridge technology during periods of fluctuating access needs. The tension between accommodation and academic integrity has direct implications for organizational accommodation policies, which may need updating to address AI-assisted work as a legitimate form of disability accommodation. The paper also raises critical questions about technology dependency and the boundary between access and convenience that are relevant as GAI tools become more prevalent. A limitation is that this is a single-author autoethnography, representing one person's experience with specific impairments and professional context, so generalizability is limited.
Tags: generative AI · temporary disability · assistive technology · autoethnography · multimodal AI · communication access · disability identity · technology adoption · AI ethics