Tailored presentation of dynamic content
Andy Brown, Caroline Jay, Simon Harper · 2010 · Proceedings of the 2010 International Cross Disciplinary Conference on Web Accessibility (W4A) · doi:10.1145/1805986.1806018
Summary
This paper presents the SASWAT browser, a research prototype built as a self-voicing extension to Firefox (based on Fire Vox), designed to address a fundamental challenge of the Web 2.0 era: how should screen readers notify users about dynamic page updates? At the time of writing, most screen readers either ignored dynamic content changes entirely or presented them in a simplistic way that did not account for the nature of the update or the user's current activity. Many screen reader users had never knowingly encountered common types of updating content such as auto-suggest lists, news tickers, or dynamically loaded page sections. While WAI-ARIA's live regions specification offered a standards-based solution, it was implemented by very few websites. The SASWAT browser therefore aimed to make dynamic updates accessible on non-ARIA-enabled sites. The system's approach was grounded in eye-tracking studies of sighted users, which revealed distinct patterns in how people attend to different types of dynamic content. The researchers used these behavioral findings to develop rules for tailoring how updates are detected, classified, and presented to screen reader users.
Key findings
Eye-tracking studies revealed a clear behavioral pattern: sighted users nearly always looked at updates they had requested or initiated (such as following a link or typing in a form), but largely ignored automatic updates like news tickers. The SASWAT browser translates these patterns into a three-stage processing pipeline for dynamic content. First, low-level DOM mutation events are clustered into meaningful user-level updates (additions, removals, replacements, or rearrangements) using a chunking algorithm. Second, updates are classified by how they were initiated — user-requested, user-initiated, or automatic. Third, presentation rules are applied based on the classification: automatic updates trigger only a brief non-speech audio cue with a reviewable list, while user-initiated updates receive a distinct audio notification plus a spoken announcement that differentiates the type of change. Focus management moves the user to new content except for removals, and an automatic bookmarking feature lets users quickly return to controls after previewing updated content — mirroring the sighted behavior of glancing at new content then returning to the original focus point.
Relevance
This research directly influenced how we think about accessible dynamic content and the practical interpretation of WAI-ARIA live regions. The insight that update presentation should be tailored based on initiation type (user-requested vs. automatic) and content type remains a foundational principle for accessible web application design. For developers, the key takeaway is that not all dynamic updates deserve equal attention — bombarding screen reader users with every DOM change is as problematic as ignoring changes entirely. The chunking approach for grouping low-level DOM events into meaningful updates addresses a real implementation challenge that developers of assistive technologies still face. While the SASWAT browser itself was a research prototype, its findings informed both screen reader development and best practices for ARIA implementation, making this an important piece of the historical record on dynamic content accessibility.
Tags: screen readers · dynamic content · AJAX · Web 2.0 · eye tracking · ARIA live regions · assistive technology
Standards referenced: WAI-ARIA