Accessibility by Demonstration: Enabling End Users to Guide Developers to Web Accessibility Solutions
Jeffrey P. Bigham, Jeremy T. Brudvik, Bernie Zhang · 2010 · Proceedings of the 12th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2010) · doi:10.1145/1878803.1878812
Summary
This paper introduces Accessibility by Demonstration (ABD), a novel approach to accessibility evaluation that lets assistive technology users record the accessibility problems they encounter as shareable, replayable demonstrations and send them directly to web developers. ABD is implemented as an extension to WebAnywhere, a web-based screen reader that requires no installation. When a blind user encounters an accessibility problem, they press a single keyboard shortcut (CTRL+R) to retroactively capture a recording of their recent interactions. The recording is packaged as a parameterized URL that encapsulates both the user's actions and the WebAnywhere screen reader itself, so developers can experience the problem simply by clicking a link — no software installation or screen reader expertise required. The URL loads WebAnywhere in the developer's browser and replays the recorded interactions, showing what the screen reader reads aloud with visual highlighting, displaying keyboard shortcuts used, and providing semantic descriptions of each action. Developers can then interrupt the replay at any point, navigate the page themselves, make code changes, and reload the ABD URL to verify their fixes iteratively. The retroactive recording feature is key: users do not need to anticipate problems — they simply browse normally and capture problems after they occur, using a buffer of recent actions.
Key findings
In an evaluation with 15 web developers (self-rated skilled at web design, mean 5.07/7, but inexperienced with accessibility, mean 3.93/7), participants using WebAnywhere-ABD recognized and fixed 41.9% of accessibility problems compared to only 25.0% using WCAG guidelines alone — a statistically significant difference (t(24)=2.32, p<.05). The WAVE evaluation tool achieved 44.1% (not significantly different from ABD). However, ABD significantly outperformed WAVE on reading order problems (50.0% vs 36.4%, t(6)=2.32, p<.05), which are particularly subtle usability issues that automated tools struggle to convey effectively. The study tested four common accessibility problem types: tables causing incorrect reading order, improper use of heading tags (using span instead of h1-h6), missing/inappropriate alternative text, and incorrect tab order from tabindex values. ABD was most effective for problems that are difficult to detect through automated scanning but immediately apparent when experienced through assistive technology, such as reading order issues caused by CSS absolute positioning.
Relevance
ABD addresses a fundamental gap in accessibility practice: the disconnect between people who experience accessibility problems and the developers who can fix them. Traditional approaches either rely on automated tools that produce overwhelming lists of warnings (many false positives) or require developers to install and learn complex screen reader software. ABD bridges this by letting users demonstrate real problems in context and letting developers experience them directly. The approach embodies a key accessibility principle: the most effective way to understand an accessibility barrier is to experience it, not just read about it. For accessibility practitioners, ABD provides a model for user-developer communication that goes beyond bug reports — it creates empathy by putting developers in the screen reader user's position. The concept of retroactive recording (capturing problems after they occur without advance planning) is particularly valuable because users cannot always predict when they will encounter barriers. While WebAnywhere has limitations (no ARIA support at the time), the ABD concept remains relevant and has influenced modern accessibility testing approaches that emphasize experiencing content through assistive technology rather than solely relying on automated checkers.
Tags: web accessibility · accessibility testing · screen readers · blind and low vision · accessibility evaluation · developer tools · user research
Standards referenced: WCAG