Enriching Web Information Scent for Blind Users
Markel Vigo, Barbara Leporini, Fabio Paternò · 2009 · Proceedings of the 11th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '09) · doi:10.1145/1639642.1639665
Summary
This paper investigates whether annotating web links with accessibility scores of their target pages can improve navigation efficiency for blind screen reader users. The core idea draws on information foraging theory: just as sighted users follow visual cues ("information scent") to predict whether a link will lead to useful content, blind users could benefit from knowing in advance how accessible a target page is. The system works by automatically evaluating web pages using two complementary tools — Magenta (which checks WCAG accessibility) and ACB (Accessibility Checker for the Blind, which evaluates usability guidelines specific to blind users). The outputs are combined through a novel Logistic Scoring Preferences (LSP) framework that aggregates multiple checkpoint scores into a single 0-100 accessibility score using logical relationships between evaluation criteria (conjunction, disjunction, simultaneity, and replaceability). This score is then displayed as an annotation next to each link. A key technical contribution is the Dependencies Solver component that handles overlapping evaluation issues between the two tools, ensuring checkpoints are not double-counted. The authors also introduce the Usability Guidelines for the Blind (UGB), a set of guidelines organised around four principles: structure and arrangement, content appropriateness, navigation, and consistency.
Key findings
A remote user test with 16 blind JAWS users (mean age 43, 43% browsing more than 2 hours daily) evaluated two tasks — browsing by navigating and searching by navigating — each with annotated and non-annotated link versions. With annotated links, the navigation paradigm fundamentally changed: users shifted from sequential link-by-link browsing to selectively targeting links with high accessibility scores. In the browsing task, the maximum score chosen was 75 out of 100 (median 52, mean 56) for annotated pages, while the searching task showed similar selective behaviour. Users who navigated with annotated links performed the browsing task with mean efficiency of 54 (vs 34 for non-annotated) and mean searching efficiency of 41 (vs 34). However, there was not general agreement between the accessibility scores and users' actual perception of page accessibility — users found scores "not always accurate" and the correlation between computed scores and perceived accessibility was present but imperfect. Despite this, 14 of 16 users found annotations helpful, rating them 3 out of 5 on average. Users were particularly positive when links were homogeneous regarding topic (e.g., a directory of travel guides), where the scores helped differentiate between otherwise similar-looking links.
Relevance
This research addresses a fundamental navigation challenge for blind web users: the inability to quickly scan and assess pages before committing to visiting them. Sighted users can glance at a page's visual layout to judge its quality and relevance, but screen reader users must laboriously traverse content to make the same assessment. Link annotation with accessibility scores offers a form of "preview" that can save significant navigation effort. The finding that annotations shifted behaviour from sequential to selective browsing is significant — it suggests that providing any form of predictive quality signal fundamentally changes how blind users approach link lists. The disconnect between computed scores and perceived accessibility highlights the well-known limitation of automated evaluation tools: they cannot capture the full user experience. The LSP scoring framework's approach to combining multiple evaluation criteria using logical relationships (rather than simple averaging) offers a more sophisticated model for aggregating accessibility metrics. For practitioners, this work suggests that even imperfect accessibility indicators can improve navigation when users are choosing among many similar links.
Tags: web accessibility · blind users · screen readers · information scent · information foraging · adaptive navigation · automated evaluation · web usability
Standards referenced: WCAG 1.0