Guideline Aggregation: Web Accessibility

19
The problem Barrier walkthrough Experimental plan Results Discussion and conclusions Guideline Aggregation: Web Accessibility Evaluation for Older Users Giorgio Brajnik (1), Yeliz Yesilada (2), Simon Harper (2) (1) Dip. di Matematica e Informatica University of Udine, Italy www.dimi.uniud.it/giorgio (2)School of Computer Science University of Manchester Manchester, UK W4A 2009 c Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

description

Web site evaluation methodologies and validation engines take theview that all accessibility guidelines must be met to gaincompliance. Problems exist in this regard as contradictions withinthe rule set may arise, and the type of impairment or its severityis not isolated. The Barrier Walkthrough (BW) method goes someway toaddressing these issues by enabling barrier types derived fromguidelines to be applied to different user categories such as motoror hearing impairment, etc. In this paper, we use set theory tocreate a validation scheme for older users by combining barriertypes specific to motor impaired and low vision users,thereby creating a new ``older users'' category from the results ofthis set addition. To evaluate this approach, we have conducted a BWstudy with four pages, 19 expert and 48 non-expert judges. Thisstudy shows that the BW generates reliable data for the proposedaggregated user category and shows how experts and non-expertsevaluate pages differently. The study also highlights a limitationof the BW by showing that a better aggregated user category wouldhave been created by having a severity level of disability fordifferent impairment types. By extending the BW with theseimpairment levels, we argue that the BW would become more useful forvalidating Web pages when dealing with users which multipledisabilities and thus we would be able to create a ``PersonalisedValidation and Repair'' method.

Transcript of Guideline Aggregation: Web Accessibility

Page 1: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Guideline Aggregation: Web AccessibilityEvaluation for Older Users

Giorgio Brajnik (1), Yeliz Yesilada (2), Simon Harper (2)

(1) Dip. di Matematica e InformaticaUniversity of Udine, Italy

www.dimi.uniud.it/giorgio(2)School of Computer Science

University of ManchesterManchester, UK

W4A 2009

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 2: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

The problem with analytic evaluation methods

I conformance reviews (eg. wrt WCAG20) arenon-contextualized, not specific

I evaluators are not guided into assessing consequences ofviolations

I there’s no reliable way to rate severity of violations

Our approach

1. Provide context to evaluators: focus on specific barriersand user categories (eg. blind, motor impaired, cognitivelyimpaired, low vision, ...)

2. Provide more formalized ways to rate severity

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 3: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Multiple impairments

How to cope with multiple impairments and combinatorialexplosion?

I eg. older people

I Dynamic Aggregation:

1. do the evaluation for primitive categories2. and then aggregate3. eg. barriers for older people = barriers for low vision∪ those for motor impaired ∪ ...

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 4: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Barrier Walkthrough

1. Analytic method; similar to "heuristic walkthrough"2. Based on barriers (ako "vulnerability points")3. Failure modes are contextualized within usage scenarios4. This helps evaluators in rating severity = F(impact,

persistence) in {1,2,3}5. See http://www.dimi.uniud.it/giorgio/

projects/bw/bw.html

(Brajnik, ICCHP 2006; ASSETS 2007)

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 5: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Example of a barrier

Rich images lacking equivalent text

I Users: Blind persons using a screen readerI Cause: The page contains some image that provides

information (e.g. a diagram, histogram, picture, drawing,graph) but only in a graphical format; no equivalent textualdescription appears in the page.

I Failure mode: The user, even if s/he perceives that thereis an important image, has no way to get the information itcontains. In addition s/he spends time and effort trying tofind out where in the page or site that information is buried.

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 6: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Experiment

GoalTo explore which conclusions are invariant wrt aggregation.

I Do certain differences among sites disappear?I How does reliability change?I How does correctness of evaluations change?I How does the difference b/w expert/non-expert change?

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 7: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Plan

Mixed design experiment

I 19 experts + 51 non-experts applying BW; 61 barrier types(within-subj)

I 2 primitive user categories: low vision, motor impaired(within-subj)

I 1 aggregated category: older adults = union of individualbarriers found for primitive categories

I 4 pages (1 page/subject, between-subj): IMDB.com,Facebook.com, novascotiaquilts.com, Sam’s Chop House

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 8: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Spreadsheet

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 9: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

True Barriers Types

Correct ratingsthose where the majority of experts agreed on their severity

Results:I Experts: 27 out of 61 barrier types ("ambiguous links",

"functional images w/o text", "inflexible layout", "missinginternal links", ...)

I Non-experts: 24 out of those 27 (missed: "forms w/olabels", "moving content", "no css support")

I Certain barriers are specific for specific user categories

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 10: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Reliability

Reproducibilitygiven (barrier type, user group, page)rep = 1− sd

M if positive; 1 if M = 0; 0 otherwisewhere M, sd are mean/std.dev of weighted severity

Agreementgiven (user group, page)on all barrier types compute the ICC (Intraclass CorrelationCoefficient – relative and absolute consistency)

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 11: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Reproducibility

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 12: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Reproducibility

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 13: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Mean weighted severities

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 14: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Correctness

I Error rate E = IC+I

I Accuracy = % ofreported barriers thatare correct

I Sensitivity = % ofcorrect barriers that arereported

I F.measure = 2A·SA+S

Ratings:

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 15: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Error rates

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 16: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

F-measure

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 17: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Invariant properties

1. Aggregation does not worsen the problem of missedbarriers

2. Reliability: experts are consistently more reliable; samepattern across pages

3. Severities: experts are more judgmental; ranks of pagesdo not change

4. Quality: error rates maintain a similar difference (expert vsnon-experts)

5. Quality: F-measure conf. intervals shrink; they keep samerelationship

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 18: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Conclusions

1. Aggregation seems to work: it enables contextualizedevaluations and leads to results that are potentially valid

2. It could be extended to cope with degrees of impairment

Limitations

1. We did not validate our conclusions against anindependent assessment

2. We don’t know if the same conclusions would hold for anyset of primitive user categories

Questions?

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users

Page 19: Guideline Aggregation: Web Accessibility

The problem Barrier walkthrough Experimental plan Results Discussion and conclusions

Evaluation framework

I based on reliability (reproducibility + agreement),correctness (error rate, accuracy, sensitivity andF-measure)

I is viableI is discriminatory

It can be used to assess pros and cons of an evaluationmethod.

c© Brajnik, Yesilada, Harper Guideline Aggregation: Web Accessibility Evaluation for Older Users