Jeffrey P. Bigham Anna C. Cavender, Jeremy T. Brudvik, Jacob O. Wobbrock * and Richard E. Ladner

1. WebinSitu: A Comparison of Blind and Sighted Browsing Behavior Jeffrey P. Bigham Anna C. Cavender, Jeremy T. Brudvik, Jacob O. Wobbrock* and Richard E. Ladner Computer Science & Engineering The Information School* University of Washington

2. Introduction Study Overview • Proxy-based observation for one week • 10 Blind and 10 Sighted (Ages 18-63) • Either Internet Explorer or Firefox • Blind participants used JAWS • 21,442 Pages • 4,204,904 Events Geographic Diversity of Users

3. Introduction in situ Study • Valuable Qualities • Participants use their own tools • Familiar, preferred web pages • Observe longer time periods • Usage Patterns in Usual Browsing • Effects of web accessibility • Coping strategies employed • Differences in content chosen to view

4. Introduction Important Complement to Prior Work [1] Takagi et al.Analysis of navigability of Web applications for improving blind usability. 2007. [2] Watanabe et al.Experimental evaluation of usability and accessibility of heading elements. 2007. [4] Coyne et al. Beyond alt text: Making the web easy to use for users with disabilities. 2001. [3] Lazar et al. Determining the impact of computer frustration on the mood of blind users. 2006.

5. Outline • Introduction • Experimental Setup and Study Design • Browsing Differences • Effects of Content

6. Setup and Study Design A Proxy-Based System • Used UsaProxy1 [1] Richard Atterer et al. Knowing the User's Every Move - User Activity Tracking for Website Usability Evaluation and Implicit Interaction. WWW 2007

7. Setup and Study Design More than a regular proxy Requests GET http://webinsight.cs.washington.edu/, 10 Oct 2007 04:12:30 GET http://webinsight.cs.washington.edu/pics/web-eye.gif, 10 Oct 2007 04:12:30 GET http://webinsight.cs.washington.edu/css/style.css, 10 Oct 2007 04:12:31 … Actions Keypress, ctrl f, 10 Oct 2007 04:12:35 Mouse, 540x232, 10 Oct 2007 04:12:36 Focus, Text Box (name), 10 Oct 2007 04:12:36 AJAX, url=“http://www.cs.washington.edu/.../foo.php, 10 Oct 2007 04:12:36 Page Changed, 10 Oct 2007 04:12:39 … Content Image, alt=“Contact Us”, src=“http://www.washington.edu/pics/contact.gif Link, name=“University of Washington”, url=“http://www.washington.edu” …

8. Setup and Study Design Easy Setup and Deployment • No New Software to Install • Works with Existing Tools

9. Outline • Background • Experimental Setup and Study Design • Browsing Behavior • Effects of Content

10. Browsing Differences Using the Mouse • Blind Users Don’t Use a Mouse • but, sometimes they have to % of Pages with Mouse Movements per Participant

11. Browsing Differences Using the Mouse (why) “…if there's a command in a form or shopping cart that says, ‘click here,’ with no labeled button, I must route my cursor to that position…”

12. Browsing Differences Probing: Following a link and returning in less than 30 seconds Call for Papers technical papers Technical Program technical program

13. Browsing Differences Web Pages with Probes (p < 0.01)

14. Browsing Differences Browsing Efficiency • Blind Users Less Efficient • Overall, ~2x longer per page • Contrast to 10x on completing tasks1 • Why not more? • Web pages, not tasks • Accustomed to Web Pages • “errors” (including probing) (p < 0.1) [1] Takagi et al.Analysis of navigability of Web applications for improving blind usability. 2007.

15. Browsing Differences Using Google

16. Outline • Background • Experimental Setup and Study Design • Browsing Differences • Effects of Content

17. Effects of Content Images and Alternative Text (empty) Blind Users are Smart http://www.domain.com/proceed.gif http://www.domain.com/pubbank-button.gif http://www.domain.com/239080s.gif

18. Effects of Content Images and Appropriate Alt. Text • % of Images with App. Alt. Text • Did not influence browsing behavior • Influenced Clicking Behavior: ClickedImages with App. Alt. Text (p < 0.01) % of Images Assigned Appropriate Alternative Text on Visited Pages

19. Effects of Content Skip Links “Skip top navigation and go to home page content” 822 Skip Links Blind users clicked 5.6% “Skip links are almost always broken.”

20. Effects of Content • Dynamic Content • 15.0x fewer pages viewed (p < 0.07) • 19.3x fewer interactions with dynamic content (p < 0.01) • AJAX • 7.5x fewer (p < 0.05) • Flash • 44.1% were ads • Blind participants used for sound content • Only 5.6% were main content

21. Effects of Content Summary and Future Work • Main Points • Facilitated new type of study • Confirmed anecdotal observations • Interesting new directions • Many Remaining Questions • Efficiency and experience • Content requires using the mouse • Annotation of dynamic content (ARIA) • Extent of Flash accessibility • MANY OTHERS

22. The End WebInSight webinsight.cs.washington.edu Thanks to: National Science Foundation Max Aller, Richard Atterer, Darren Gergle, Steve Gribble, Sangyun Hahn, Scott Rose, Lindsay Yazzolino.

23. Background and Motivation Important Complement to Prior Work [1] Takagi et al.Analysis of navigability of Web applications for improving blind usability. 2007. [2] Watanabe et al.Experimental evaluation of usability and accessibility of heading elements. 2007. [3] Lazar et al. Determining the impact of computer frustration on the mood of blind users. 2006. [4] Coyne et al. Beyond alt text: Making the web easy to use for users with disabilities. 2001.