Modeling in HCI

1
Modeling in HCI

• Stuart Card
• Palo Alto Research Center
• (PARC)
• Stanford University, CS376
• November 19, 2009

2
Why Model?
3
EXAMPLE POINTING DEVICES
Mouse. Engelbart and English
4
TRADITIONAL METHOD EVALUATION
Sun Labs
5
Engelbart
6
EXPERIMENT MICE ARE FASTEST
7
WHY? (ENGINIEERING ANALYSIS)
3
Why these results? Time to position mouse
proportional to Fitts Index of Difficulty
ID. Proportionality constant 10 bits/sec, same
as hand. Therefore speed limit is in the
eye-hand system, not the mouse. Therefore, mouse
is a near optimal device.
Mouse
2
Movement Time (sec)
1
T 1.03 .096 log2 (D/S .5) sec
0
2
3
1
4
5
6
IDlog (Dist/Size .5)
2
8
ENGINEERING ANALYSIS (Modeling)

• Insightful
• Accumulate into a discipline
• Generative

9
CUMULATING INTO A DISCIPLINEChapanis Report on HF

• (National Research Council)
• Experimental methods alone are inadequate.
• Of 40 non-experimental techniques in human
  factors, only 2 were validated and taught.

10
TO BE GENERATIVE

• Task analysis
• Approximation
• Calculation
• Zero-parameter predictions

11
EXAMPLE ALTERNATIVE DEVICES
Headmouse No chance to win
12
ATTACHING POINTING DEVICE
Use transducer on high bandwidth muscles
13
EXAMPLE STRUCTURING THE TASK SPACE BY PROJECTING
THE MODEL
Word
TIME (msec)
Period
Paragraph
Char
500
2000
0
1000
1500
Mouse (Arm)
Hard
Easy
Head- mouse (Head)
Hard
Easy
Fingers
Hard
14
EXAMPLE BEATING THE MOUSE
Use transducer on high bandwidth muscles
15
DESIGNS FROM RESTRUCTURED TASK SPACE
Work with Bill Moggridge, IDEO
16
EXAMPLE DESIGN SPACE
17
MORPHOLOGICAL DESIGN GENERATINGALL
INPUTDEVICES
18
POINTING DEVICES
19
MODEL HUMAN PROCESSOR

• Processors and Memories applied to human
• Used for routine cognitive skill

20
(No Transcript)
21
EXAMPLE ZERO-PARAMETER CALC

• ProblemInventor claims he invented 600 wpm
  typewriter. License and develop?
• Solution 1Half stroke tM 70
  ms/charWhole stroke tM tM 140
  ms/charbut if between hands, overlap tM
  70 ms 131 words/min

22
EXAMPLE ZERO-PARAMETER CALC

• Solution 2 (range calculation)Half stroke
  tM70 30100 ms/char 131 30892
  words/min
• Conclusion
• Bogus claim. Throw himout!

23
TASK ANALYSIS GOMS(GOALS, OPERATORS, METHODS,
SELECTION RULES)
task analysis

• GOAL EDIT-MANUSCRIPT repeat until done
• GOAL EDIT-UNIT-TASK
• GOAL ACQUIRE-UNIT-TASK if not remembered
• GET-NEXT-PAGE if at end of page
• GET-NEXT-TASK if an edit task found
• GOAL EXECUTE-UNIT-TASK
• GOAL LOCATE-LINE if task not on line
• select USE-QS-METHOD
• USE-LF-METHOD
• GOAL MODIFY-TEXT
• select USE-S-COMMAND
• USE-M-COMMAND

24
PREDICTS TIME WITHIN ABOUT 20
25
SAE RECOMMENDED PRACTICE J2365

• Predict task times for car navigation systems
• Check compliance with SAE J2364 (15-Second Rule)
• Note To estimate times while driving, multiply
  by 1.3 to 1.5.
• Based on GOMS and work by Paul Green at Univ. of
  Michigan Transportation Research Institute.

Dario Salvucci
26
SAE J2365 OPERATOR TIMES
Paul Green UMITRI
27
LHX HELICOPTER SIMULATION(Corker, Davis,
Papazian, Pew, 1986)

• POP-UP-AND-SCAN
• POP-UP-FOR-SCAN
• in parallel-do
• LOOK-FOR
• POP-UP
• STABILIZE-CRAFT
• HOVER-AND-SCAN
• in-parallel-do
• HOVER
• SCAN

GOMS used as task analysis to code doctrine
28
IMMEDIATE BEHAVIOR
29
HUMAN INFORMATION INTERACTION
30
GOMS

• Routine cognitive skill
• Well-known path

31
Information Search

• Problem solving
• Heuristic search
• Exponential if dont know what to do

32
OPTIMALITY THEORY
Optimal Foraging Theory
Information Foraging Theory
Information
Energy




Useful info Time
Energy Time
Max
Max
33
Information Foraging Theory
People are information rate maximizers of
benefits/costs Information has a cost structure
34
INFORMATION PATCHES
e.g. desk piles, Alta vista search list unlike
animals foraging for food, humans can do patch
construction
35
CHARNOVS MARGINAL VALUE THEOREM
max gain when slope of within-path gain g
average gain R (tangent in diagram)
Gain
R
g(tW)
Within-patch time
Between-patch time
tB
t
36
BETWEEN-PATCH ENRICHMENT
Gain
R2
R1
g(tW)
Within-patch time
Between-patch time
tB1
t1
tB2
t2
enrichment
Example arrange physical office efficiently
37
WITHIN-PATCH ENRICHMENT
Example Better filtering of search hits
Behavior adapts to cost structure of environment.
g1(tW)
Within-patch time
Between-patch time
38
WITHIN-PATCH ENRICHMENTINFORMATION SCENT
perception of value and cost of a path to a
source based on proximal cues
Tokyo
New York
San Francisco
39
RELEVANCE-ENHANCED THUMBNAILS

• Emphasize text that is relevant to query
• Text callouts
• Enlarge text that might be helpful in assessing
  page
• Enlarge headers

Allison Woodruff
40
PHASE TRANSITION IN NAVIGATION COSTS AS FUNCTION
OF INFORMATION SCENT
150
150
Probability of choosing wrong link (f)
.150
.150
100
100
Number of pages visited
.125
50
50
.100
.100
0
0
0
2
4
6
8
10
0
2
4
6
8
10
Depth
Notes Average branching factor 10
Depth 10
41
IMPORTANCE FOR WEB DESIGN
Jarad Spool, UIE
42
MACHINE MODELING OF INFORMATION SCENT
new
cell
Information Goal
medical
patient
Link Text
treatments
dose
procedures
beam
43
PREDICTION OF LINK CHOICE
35
50
(b) Yahoo
(a)
ParcWeb
30
40
25
Predicted frequency
30
20
R2 0.72
Predicted frequency
15
20
10
R2 0.90
10
5
0
0
0
10
20
30
40
50
0
5
10
15
20
25
30
35
Observed frequency
Observed frequency
44
USER FLOW MODEL
User need (vector of goal concepts)
45
BLOODHOUND PROJECT
INPUT
Starting Point www.xerox.com Task look for
high end copiers
OUTPUT usability metrics
Chi, et al
46
Information Cost Landscapes Exercise
47
Moving to a Patch
48
How long to get to any one itemin a patch?
total items in patches
Gain
n items accessible
?
?
Time
t2
ave. time for patches
time to get one item
49
Example Rectangular patch of patches
D3
D4
D5
D6
D7
D1
D2
C2
C3
C4
C5
D8
D24
C1
B1
C6
D23
B2
B3
D9
C16
A
B8
B4
C7
D10
D22
C15
B7
B6
B5
C8
D11
D21
C14
C12
C11
C10
C9
D12
D20
C13
D17
D16
D15
D14
D13
D19
D18
10 items/patch
50
Task Names (Patch Names)
51
Distances From Patch A
52
Patch Distances
53
COKCF Calculation Table
54
COKCF Calculation TableSorted by Cost
55
Cost-of-Knowledge Characteristic Function (COKCF)
56
COCKF Calculation Table Sorted by Cost
0 ? 0.5 (10 items)
1 ? 0.5 (80 items)
2 ? 0.5
57
COCKF Calculation Table Grouped by Class Interval
58
Smoothed COKCF
59
Cost of Knowledge Characteristic Function
Gain in Knowledge
Cost Time
60
Spiral Calendar
61
COCKF Calculation Table
62
Design
Test (5 Blocks x 11 Trials)Order
counterbalanced

• Warm-up (1 Block x 11 Trials)

63
COCKF Calculation Table
64
Raw Results
30
25
20
15
ACCESS TIME (s)
10
5
0
100
101
102
103
104
105
106
DAYS BACK
65
Many Interfaces for Foraging areDirect Walk
Display2
Display3
Display1
Etc
Click,Gesture, Etc
Click,Gesture, Etc
Click,Gesture, Etc
Examples WWW, Mac Finder, HyperCard
66
GOMS ANALYSIS

• Century-Method GOAL DO-TASK
• GOAL ACCESS-DAY-CALENDAR
• GET-YEAR . . . if necessary
• GOAL SELECT-CENTURY (1700s)
• POINT-TO (Century1700-1790s) gt
  ltCentury-Displaygt
• GET-YEAR . . . if necessary
• GOAL SELECT-DECADE (1710s)
• POINT-TO (Decade1710-1719) gt ltDecade-Displaygt
• GET-YEAR . . . if necessary
• GOAL SELECT-YEAR (1719)
• POINT-TO (Year1700-1790s) gt ltYear-Displaygt
• GET-MONTH . . . if necessary
• GOAL SELECT-MONTH (November)
• POINT-TO (Month1700-1790s) gt ltMonth-Displaygt
• GET-DAY . . . if necessary
• GOAL SELECT-WEEK ()
• POINT-TO (Weekcontains 23) gt ltWeek-Displaygt
• GET-DAY . . . if necessary
• GOAL SELECT-DAY (23)

Cycle 1
Cycle 2
Cycle 3
Cycle 4
Cycle 5
Cycle 6
67
COCKF Calculation Table
68
Arranged by Interaction Cycles
30
25
20
Century
ACESS TIME (s)
Decade
15
Year
10
Month
Week
5
Day
0
100
101
102
103
104
105
106
DAYS BACK
69
COCKF Calculation Table (Grouped)
70
Access Time 3.3 3.5 NCycles
30
25
20
15
ACESS TIME (s)
10
5
0
0
2
4
6
NUMBER OF SELECT-DISPLAY CYCLES
71
COCKF Calculation Table (Grouped)
72
107
Spiral Calendar
106
105
104
INFORMATION ACCESSIBLE (Days)
103
102
101
100
0
20
40
60
80
100
120
ACCESS TIME COST (s)
73
Sun CM

• Same task
• Same Procedure
• Restricted to Direct Walk methods

Results of GOMS analysis Access Time 1.3
3.9m 1.4 P 0.36 B sec wherem number of
point, menu pull-down, selectP number of point
selectB number of button presses
74
107
Spiral Calendar
106
CM
105
104
INFORMATION ACCESSIBLE (Days)
103
102
101
100
0
20
40
60
80
100
120
COST (s)
75
No Week Calendar (calculated)
107
Spiral Calendar
106
105
CM
104
INFORMATION ACCESSIBLE (Days)
103
102
101
100
0
20
40
60
80
100
120
COST (s)
76
2 sec User-Action Cycle (calculated)
107
Spiral Calendar
106
105
CM
104
INFORMATION ACCESSIBLE (Days)
103
102
101
100
0
20
40
60
80
100
120
COST (s)
77
2 sec User-Action Cycle No Week(calculated)
107
106
105
104
INFORMATION ACCESSIBLE (Days)
103
102
101
100
0
20
40
60
80
100
120
COST (s)
78
Summary

• Cost of Knowledge Characteristic Function(COKCF)
  is metric for cost landscape
• Can use to measure
• Can use for task analysis

79
Another COKCF calculation
80
SUMMARYENGINEERING ANALYSIS / MODLING

• Insightful
• Accumulate into a discipline
• Generative