Guest lecture on guerilla usability testing and paper prototyping
Usability with Project Lecture 10 – 10/10/08
description
Transcript of Usability with Project Lecture 10 – 10/10/08
© Simeon Keates 2008
Usability with ProjectLecture 10 – 10/10/08Dr. Simeon Keates
© Simeon Keates 2008
Exercise – part 1
Last week you were asked to bring in 4 items• Landline telephone• Mobile telephone• TV remote control• 1 other item
This week…
Perform exclusion calculations on each product using the data on:• http://www.eng.cam.ac.uk/inclusivedesign/
Page 2
© Simeon Keates 2008
Exercise – part 2
Identify the common methods of interacting with the product
Identify which of the 7 DFS capability scales are involved in the interaction
Based on the DFS scales, estimate the limiting capability demand for each scale
Page 3
© Simeon Keates 2008
Exercise – part 3
Report the number and %age of people excluded by each capability demand• For 16+ and 75+
Report the total number and %age of people excluded by the product• For 16+ and 75+
Prepare a 5 minute presentation to discuss:• Your exclusion calculation assumptions• Your exclusion calculation results• What were the principal causes of exclusion?• What do you think should be done to reduce the exclusion for each product?
Page 4
© Simeon Keates 2008
What is “reasonable accommodation”?
Page 5
© Simeon Keates 2008
Defining “reasonable accommodation”
Must offer “reasonable accommodation”• BUT what is reasonable?
Not defined explicitly• Companies left guessing
Will be defined in courts• Major risk/headache for companies
Page 6
© Simeon Keates 2008
Attitudes to “reasonable accommodation”
EQUITABLE ACCESS
MINIMUM(compliance)
Access to functionality
IDEALAccess to
functionalityin same time
EQUITABLE ACCESS
MINIMUM(compliance)
Access to functionality
IDEALAccess to
functionalityin same timeIDEOLOGICAL
DIVIDE
Prag
mat
ists
Idea
lists
Page 7
© Simeon Keates 2008
Interesting questions for companies
Is the equitable access ideal possible?• Is the equitable access minimum possible?• “ Equal, but different ” problem
Users with functional impairments => longer times
Can technology always make up the difference in user capabilities?
3 case studies…
Page 8
© Simeon Keates 2008
Case study 1: The personal information point
Page 9
© Simeon Keates 2008
The information point accessibility assessment
Sensory assessment: Screen too high and not adjustable Audio output not duplicated Visual output not duplicated
Motor assessment: Need to stand Reaching and dexterity demands
45% of target users excluded
Is this “reasonable”?Page 10
© Simeon Keates 2008
Case study 2 – Cursor assistance for motor-impaired users
Symptoms that can affect cursor control:
Tremor Spasm Restricted motion Reduced strength Poor co-ordination
Page 11
© Simeon Keates 2008
User group behaviours
0
0.5
1
1.5
2
2.5
3
3.5
Group OA Group P Group Y Group A
0
20
40
60
80
100
120
Group OA Group P Group Y Group A
Target activation times
Peak velocities
No. of incorrect clicks
0
1
2
3
4
5
Group OA Group P Group Y Group A
Page 12
© Simeon Keates 2008
Summarising the differences
Younger adults (IBM interns)• Shortest (1), fastest (1), more errors (3) - slapdash
• “I can fix it”• Games culture?
Adults (IBM regulars)• Shorter (2), faster (2), fewest errors (1)
• Best compromise between speed and accuracy? Parkinson’s users• Longer (3), slowest (4), fewer errors (2)
• Slow, but sure Older adults• Longest (4), slower (3), most errors (4)
• Vision difficulties?• Lack of experience
Page 13
© Simeon Keates 2008
A method of cursor assistance
Haptic gravity wells:
Target
Gravity well
Attractive force
Page 14
© Simeon Keates 2008
Experimental set-up
Page 15
© Simeon Keates 2008
The effect of gravity wells
Target
Page 16
© Simeon Keates 2008
Results - Throughput
0
2
4
6
8
10
MI AB
Page 17
© Simeon Keates 2008
Case study 2 summary
Haptic gravity wells are clearly very helpful MI users “with” on similar level to AB users “without”
BUT: AB users also improve “with” Is this “equal” time? Is this “reasonable”???
Page 18
© Simeon Keates 2008
Case study 3 – Paperless office
AN Other wants to move to a paperless office• Currently receives 3.5 million pages per day
Paper documents are stored as TIFFs
Section 508 accessibility requirements• Sight-impaired• Low vision
Current solution – employ readers• “ Equal, but different. ”• Is this reasonable?
Page 19
© Simeon Keates 2008
The study documents
Almost fully unconstrained Content:• Unconstrained vocabulary
Text:• Typed• Handwritten• Annotated• Stamps
Graphical content:• Diagrams• Charts • Graphs
Page 20
© Simeon Keates 2008
Examples of the study documents
Page 21
© Simeon Keates 2008
Examples of the study documents (cont.)
Page 22
© Simeon Keates 2008
Examples of the study documents (cont.)
Page 23
© Simeon Keates 2008
Readability metrics (text)
Translation rates:• Character-by-character• Word-by-word
Number and %ages of errors:• Level 1 - Minor• Level 2 - Moderate• Level 3 - Serious
Page 24
© Simeon Keates 2008
TIFF file
OCR – The scanning process..................................11111.........11...11.......11.....11......11.....11.............11..........11111........111..11.......11....11......11.....11......11.....11......11....111......111..1.11.1.....1111..111..................
111
1
1
1
11 11
11
1
11
1111
11
1
111 1111
11 1
11
1
1 111
11
111 1
11 1
1
111111 1
1 1
1
1
Page 25
© Simeon Keates 2008
OCR – Possible sources of scanning errors
Data LOSS NOISE
Page 26
© Simeon Keates 2008
Comparing three OCR engines
“…also develop the skills to invert containers to get objects *inside. He should begin to find small details in a favorite picture book (a bird in a *tree, a small fish in the *ocean). His understanding of familiar objects should…”
FineReader:
“…*also *develop *the *skills *to *irxvert *containers *to *get *ob^ects *inside. *?e *should *begin *to *Znd *small *details *i? *a favorite *picture *baa? *?a *bird *in *a *true, *a *small *ash *in *the *ocean}. *his *understanding *of *familiar *ob^ects *should…”
OmniPage:
“…*also *de???op *the *s?il?s *ta *ivart?an#ainer?to *e?ob??cts?n?id?. *?e *shau?ti *b?ta *Znd *srnali *details *i?a *favarita *picture *baa??bi?rd *in *a *tra?,a *srr?a????in *tk?e *o?ean}. *?is *und?rt?a?af *fa.?i?iar *ob?ects *hau?d *co??i?u?ta *de?eiap *d?i?houi d…”
Recognita:
Page 27
© Simeon Keates 2008
OCR results – Calculating the error rates
Record the document properties• # of words, characters• Font types (e.g. typed, handwritten) and sizes
Count instances of error types• Redaction errors• Spaces +ed, -ed• Format errors (e.g. wrong case, incorrect text positioning)• Extraction errors (i.e. incorrect translation)
• By character• By word
Classify severity• Level 1 – minor• Level 2 – moderate• Level 3 – severe
Calculate %age error rates
Note: classification for sighted users
Page 28
© Simeon Keates 2008
OCR results – An example extracted document – 1
Extracted text:
*evaluators, shQWfag’an interest in imitating words *and sp *eech.^j^kd real words along^vith j argon to exjgpss. himself . *dflffVily indicated that they understand most of what tie *says.^H^^owedhisuse of two+ word phrases
Original text:[Typed page document]
Page 29
© Simeon Keates 2008
OCR results – An example extracted document – 2
Extracted text:*IBISES6?? *fc?day *?P *a *yearly *SJn *exam’ *She *is *a *40 *^ear *old *white *feraale status post ^aginal hysterectomy five years ago. She has continued to have some difficulty with loss || of urine upon coughing or sneezing. I had given her some samples of Ditropan last year but || *SShZ *^ *t0 *^ *theSe’ *ShS *feelS *that *her *wei^ contributes a ^reatleal *Z *££ problems *with *mcontmence She has had some continuing problems with depressive *sympW *S^e cries very easily and it is getting a little bit worse. She also feels very *withdrawn *She tells roe that her sister in Florida had a similar history and was on *Paxil and did.
Original text:[Typed page with notes document]
Page 30
© Simeon Keates 2008
OCR results – An example extracted document – 3
Extracted text:
*2j*»rlfar Cardiology || *^^m Chart: *3£4U3& *Dr *-^ || *0 _. *, Medications: *Adenosinc *Dose: || Dose: *jjj&f»- *f-^- *\ *Dobutaimne
Original text:[Pictures and Graphs document]
Page 31
© Simeon Keates 2008
OCR results – Overall word error % rates
Typed page 6.50 %(1 word in 15)
Typed page with notes
8.12 %(1 word in 12)
Faxes 14.45 %(1 word in 7)
Pictures and graphs
23.45 %(1 word in 4)
Handwritten reports
36.35 %(1 word in 3)
EKGs 49.72 %(1 word in 2)
A “typical sentence” contains 7 words.
An extraction error rate of 6.5% equates to 1 word error every 2 sentences.
Page 32
© Simeon Keates 2008
OCR results – Context metrics Text location awareness – PARTLY SATISFIED – columns only
• Does the data extraction technology output provide an indication of where the text is on the page? Table search – VERY LIMITED – recognised individual columns, not tables
• Does the data extraction technology recognise tables and support searching within them? Diagram detection – VERY LIMITED – recognised as “not text”
• Does the data extraction technology recognise diagrams and support searching within them? Graph detection – VERY LIMITED – as for diagram detection
• Does the data extraction technology recognise graphs (charts) and support searching within them? Dealing with uncertainty – SATISFIED – all engines highlighted uncertain text
• Does the data extraction technology recognise entities on the page that it cannot translate and highlight this?
Text emphasis – PARTLY SATISFIED – could, but not always correct• Does the data extraction technology recognise when the author of the document has selected a
particular item of text for special emphasis? Multiple selection lists – VERY LIMITED – words and columns, but no “meta” info
• Does the data extraction technology recognise multiple selection lists and can it identify the item(s) selected?
Page 33
© Simeon Keates 2008
Conclusions of OCR investigation
“ Current OCR technology is not capable of providing an acceptable level of text extraction from medical evidence as it is now received. ”
“ Technology cannot provide equitable access in this case. Alternative methods are required. ”
“ Equal, but different. ”
Page 34
© Simeon Keates 2008
Overall summary
Some products clearly not “reasonable”• Case study 1
Technology cannot always make up for lack of user capability• Case study 2• Even when it does – the goalposts move!!!
Page 35
© Simeon Keates 2008
Conclusion
What is needed is a framework for evaluating “reasonableness”
Based on quantifiable metrics
Reliable, repeatable, consistent, robust
Page 36
© Simeon Keates 2008
A framework for assessing acceptability – 1
Stage 1 – Identify each target user group/persona• e.g. blind users, >65s, etc.
Stage 2 – Identify each component step in the interaction per group• e.g. press Enter, activate OK button, move cursor to icon, etc.
Stage 3 – Compare number of steps per group• e.g. 10 for able-bodied, 30 for blind using screen reader
DECISION GATEWAY 1 Are the numbers of steps roughly equal?
If not – differences need to be justified or remedied
Page 37
© Simeon Keates 2008
A framework for assessing acceptability – 2
Stage 4 – Perform user studies with baseline user group• Calculate times, error rates, etc.
Stage 5 – Perform user studies with target user groups• Calculate times, error rates, etc.
DECISION GATEWAY 2 Could all of the users complete the task?
If not – causes of difficulties need to be removed or remedied
Page 38
© Simeon Keates 2008
A framework for assessing acceptability – 3
Stage 6 – Compare error rates for each group• e.g. 2 per trial able-bodied, 5 per trial blind using screen-reader
DECISION GATEWAY 3 Are the error rates the same or similar across user groups?
If not – significant differences have to be justified or remedied
Page 39
© Simeon Keates 2008
A framework for assessing acceptability – 4
Stage 7 – Compare times to complete tasks for each group + modifiers• e.g. number of component steps per group +• proportion of component steps affected by group disabilities +• relative importance of each step (3 = critical, 1 = peripheral) +• relative severity of the level of disability +• additional latencies from AT used
DECISION GATEWAY 4 Are the modified times the same or similar across user groups?
If not – significant differences have to be justified or remedied
Page 40
© Simeon Keates 2008
When we come back…
User trials• How to plan the trials• How to select users• How to conduct the sessions• How to analyse the data gathered• How to make design recommendations
Designing and evaluating for unusual circumstances• Airports• Mobile phones
Making the business case for usability• How to calculate the “bottom line” impact
Project• Finishing your design and then testing with “real” people!
Page 41
© Simeon Keates 2008
Exercise
Page 42
© Simeon Keates 2008
Exercise – part 1
Perform an exclusion analysis on your web-site • (As you did on Wednesday)
Prepare a summary of your calculation• Assumptions• Levels of capability required• Exclusion (total and %age) for 16+ and 75+
Make any changes necessary to your site• + any outstanding ones from last couple of weeks
Page 43
© Simeon Keates 2008
And finally…
Turn to the back page of today’s handout…
Page 44