AccessAI: Bridging the gap between inaccessibility and Digital Inclusion
Rajeshwori Malla (Washburn University)
Digital accessibility in higher education is often split across separate auditing tools, manual fixes, and ad hoc assistive work, widening the gap between finding barriers and resolving them, especially as institutions move toward WCAG 2.1 Level AA expectations under policies such as ADA Title II. AccessAI is a unified, browser-based system that combines WCAG-aligned issue detection, automated remediation (e.g., alt text and structural fixes), real-time user-side controls (contrast, typography, focus, motion), and alternative accessible formats for documents from the web and LMS, with selective use of AI and OCR routed through a privacy-aware design. This project evaluates whether unifying these functions improves outcomes relative to fragmented scan-then-fix workflows. On controlled and production pages, AccessAI matched or exceeded sampled commercial auditors on category coverage in a structured benchmark; accessibility scores rose (e.g., +23 points on a difficult W3C-style page and +15 points on average across 50 sites). At scale, the system applied 720 automated repairs, skipped 210 for safety, and left 640 items for manual follow-up (e.g., tables, landmarks, ambiguous controls), with repairs concentrated in AI-generated image descriptions and link/name fixes. Exploratory interviews with accessibility leadership, screen reader users, and toolkit users indicate clearer reading experiences for complex content and appreciation for a single interface, alongside requests for faster access to controls. AccessAI does not replace full compliance auditing or human judgment on complex material; AI remains imperfect on math and diagrams, and cloud use affects latency and privacy. Overall, the work shows that a unified remediation pipeline can materially improve scores and throughput while making remaining human work explicit, supporting more scalable, user-centered inclusion efforts in educational settings.
