Input Techniques and Models
Transcript of Input Techniques and Models
Input Techniques and Models
Properties of an input device
Evaluation and analysis of input devices
Input device states
Interaction modalities
Hinckley, Ken., Input Technologies and Techniques, Chapter 7. Handbook of Human-Computer Interaction, ed. by Andrew
Sears and Julie A. Jacko. Lawrence Erlbaum & Associates.
10/5/2010 2
What‘s an input device
• Input devices sense physical properties of people, places or things
• However, they do not operate in isolation, i.e. need visual feedback
– otherwise similar to a pen without paper
• Must include:
– the physical sensor (positioning wheels)
– the feedback presented to the user (cursor)
– the ergonomic of the device (fits in hand)
– interplay between all the interaction techniques supported by a system (clicking, moving, selecting, etc.)
• Need an understanding of input technologies to design interaction techniques that match a user‘s natural workflow
10/5/2010 3
Input Device Properties
• Several properties characterize most
devices:
– Property sensed
– Number of dimensions
– Indirect vs. direct
– Device acquisition time
– Gain
10/5/2010 4
Property Sensed
• Most devices sense– Linear position (tablets sense position of pen)
– Motion (mice sense change in position)
– Force (Isometric joysticks, IBM Trackpoint)
– Angle or change in angle (rotary input)
• Absolute input device– position sensing
• Relative input device– motion sensing
• Relative device requires visual feedback but also can be inefficient due to clutching
10/5/2010 5
Number of Dimensions
• Devices sense one or more input dimensions
–Two linear dimensions (mouse, x/y)
– Angular dimension (knob)
– 6 degree-of-freedom (magnetic tracker, senses 3-position
dimensions and 3 orientation dimensions)
• A pair of knobs is a 1D+1D device, mouse with scroll
wheel is a 2D+1D multi-channel device
• Multiple degree-of-freedom devices sense three or
more simultaneous dimensions of spatial position or
orientation
10/5/2010 6
Degrees of Freedom vs. Dimensions
the bat
Ware, C. and Jessome, D. (1988), Using the Bat: A Six Dimensional Mouse for Object
Placement. IEEE Computer Graphics and Applications. November 8-6, 65-70.
Usually confuse the idea of dimensions and degrees-of-freedom
10/5/2010 7
Mapping Degrees of Freedom to Dimensions
A Two-Ball Mouse Affords Three Degrees of Freedom
I. Scott MacKenzie, R. William Soukoreff, & Chris Pal
CHI 97 Electronic Publications: Late-Breaking/Short Talks
3DOF Mouse
10/5/2010 8
Mapping Degrees of Freedom to Dimensions
Ravin Balakrishnan, Thomas Baudel, Gordon Kurtenbach, George W.
Fitzmaurice. (1997). The Rockin'Mouse: Integral 3D manipulation on a plane.
Proceedings of ACM CHI 1997 Conference on Human Factors in Computing
Systems, p. 311-318.
4DOF Mouse
10/5/2010 9
Degrees of Freedom vs. Dimensions
Hinckley, K., Sinclair, M., Hanson, E., Szeliski, R., Conway, M., The
VideoMouse: A Camera-Based Multi-Degree-of-Freedom Input Device, ACM
UIST'99 Symposium on User Interface Software & Technology, pp. 103-112.
5DOF Mouse
10/5/2010 10
Indirect vs. Direct
• Mouse is indirect, i.e. must move mouse to
move pointer on screen
• Touch-screens or tablets are direct input
devices, i.e. unified input and display
surface
– Occlusion is typically a problem
Diversity of input ‗devices‘
Positioning
Absolute positioning
Absolute positioning
Absolute positioning
Absolute positioning
Absolute positioning
Relative positioning
clutching
Input mode
Direct input
motor & visual space
Indirect input
motor space
visual space
Direct Indirect
Relative
Absolute
Central question
How to simultaneously leverage properties
of each input or positioning mode for
interaction?
Dual-Surface Input:Augmenting One-handed Interactions with
Coordinated Front and Behind-the-screen
Input
Xing-Dong Yang, Edward Mak,
Pourang Irani, Walter Bischof
ACM MobileHCI 2009
One-handedness is common
Over 80% of mobile users employ or prefer
one-handed operation [Karlson & Bederson, 2006]
Input device: Thumb
Reach
Karlson at al, 2006
Input ―mismatch‖
High occlusion + low accuracy
Input ―mismatch‖
user view hardware view
Shift
Vogel & Baudisch, ACM CHI 2007
Dual-surface input
Dual-surface model
Direct input
Move cursor with absolute input on front
Indirect input
‗Adjust‘ with relative on back
Key point: coordinated
Nanotouch
Baudisch & Chu, ACM CHI 2009
Insights #1
Coordinated front (absolute) + back (relative) superior than either alone
For back, indirect works better
Hardware solution for fluid move between
modes works
ARC-Pad:Absolute + Relative Cursor Positioning for
Large Displays with a Mobile Touchscreen
David McCallum, Pourang Irani
ACM UIST 2009
Large Displays
Reach is problematic
Custom hardware
Not always available or expensive
Coupling a mobile device
Relative
large distance to travel
lots of clutching!
Absolute
imprecise
a b c
d e
Hybrid Pointing
Forlines et al, ACM UIST 2006
ARC-Pad
ARC-Pad
ARC-Pad
Video
Insights #2
Seamless transition between absolute &
relative positioning is crucial
no explicit mode switch
via interpretation of input gesture
Software based solution requires some training
Guidelines
Avoid explicit mode switching, should be
implicit
software based – proper interpretation of input
hardware based – novel augmentations
Seamlessly coordinate input modes
Proper ergonomics to accommodate novelty
10/5/2010 48
Device Acquisition Time
• Acquisition time: average time to move
hand to device
• Homing time: average time to return to a
‗home‘ position, i.e. mouse to keyboard
• For common desktop workflows, pointing
and selecting dominate acquisition/homing
time
– integration of pointing with keyboard may not
improve overall performance
10/5/2010 49
Gain
• Control-to-display gain or C:D ratio
– distance moved by an input device/distance
moved on the display
• Composite measurement taking into
account device size and display size
• Gain has very little effect on the time to
perform pointing movements
– not a commonly used metric
10/5/2010 50
Pointing Devices
• Mouse:– Invented in ‘67
– Used for pointing
– Picks up changes in x, y
– As good as pointing with finger
– Integrated with buttons/wheels etc
• Trackball:– Senses relative motion of partially exposed ball in
2DOF
– Engage different muscle groups than the mouse, but an alternative for those who experience discomfort
10/5/2010 51
Pointing Devices
• Isometric joystick:
– Force sensing
– Rate of cursor is proportional to the force exerted
– Returns to center when released
– Good when space is at a premium
• Isotonic joystick:
– Sense angle of deflection
– Different than isometric joystick
J. Lipscomb and M. Pique (1993). Analog Input Device
Physical Characteristics. SIGCHI Bulletin 25 (3): 40-45.
10/5/2010 52
Pointing Devices
• Touchpads:
– Small touch-sensitive devices found on laptops
– Use relative mode for cursor control
– Can operate in absolute mode by dragging
finger and leaving it on edge of the pad
– Necessitates multiple clutchings
• Touchscreens/pen-operated devices:
– Fingers, or electromagnetic digitizers
– Parallax error, mismatch between sensed input
position and apparent position
10/5/2010 53
Input Device States
• Disaccord between states of a GUI and
states and events sensed by devices
• Difficult to support interface primitives such
as click, drag, double-click, and right-click
• Useful to diagram device states
– Identifies relationship between events sensed by
input device and interaction technique
demands
10/5/2010 54
Three-State Model of Graphical Input
• Buxton's 3-state model for graphical input devices
• Expression of the operation of computer pointing
devices in terms of state transitions
• Expressive vocabulary for exploring the relationship
between pointing devices and the interaction
techniques they afford
10/5/2010 55
Input Device States
10/5/2010 Comp 4020 - HCI 2 (PPI) 56
Three-State Model of Graphical Input
• Three-states:– The states are Out of range (State 0, for clutching or repositioning a mouse on a mouse pad;
– Tracking (State 1) for moving a tracking symbol such as a cursor about a display
– Dragging (State 2) for moving an icon on the desktop or for grouping a set of objects or a range of text
• Seems simple and obvious but can add insight to the existing body of pointing device research
– can be extended to multi-button interaction, stylus input, and direct vs. indirect input
10/5/2010 57
Input Device States
• Based on Buxton‘s model, the mouse &
touch sensitive devices are a two-state
device
10/5/2010 58
Activity: Input Device States
• Based on the previous state diagram, can you
describe a limitation of touch-sensing input (PDA‘s)
• Do you know of a device that supports all 3 states
10/5/2010 Comp 4020 - HCI 2 (PPI) 59
Input Device States
• Can fully capture core interaction by
extending the 3-state model
10/5/2010 60
Evaluation and Analysis of Input Devices
• Keystroke-Level Model
• Fitts‘ Law
• Hick-Hyman Law
• Accot‘s Steering Law
10/5/2010 61
Models and Modeling
• A model is a simplification of reality, but useful only if
it helps in understanding some phenomenon or
behavior:
– Design, evaluate or help understand complex behavior
• Models sit on a continuum:
Analogy &
Metaphor
Mathematical
Equations
Descriptive
Models
Predictive
Models
Fitt’s LawGuiard’s Model
10/5/2010 62
Keystroke Level Model
• Only address one aspect of task performance: time
• Predicts expert error-free task-completion time with the following inputs:
– a task or series of subtasks
– method used
– command language of the system
– motor-skill parameters of the user
– response-time parameters of the system
• Predict time an expert would take to execute the tasks– Assuming no errors
10/5/2010 63
KLM operators
• Six operators
– Keystroke• Avg time determined by std typing tests
– Pointing• Pointing with a mouse or other device on a display to select an
object.
• Varies from 0.8 – 1.5 seconds
– Homing• Bring ‗home‘ hands on the keyboard or other device
• 0.4 seconds based on various studies
– Mental• 1.35 seconds, experimentally determined
– Response
• Texecute = TK + TP + TH + TM + TR
10/5/2010 64
Encoding Methods
• E.g., replace 5 letter word with another in a text
editor
• Reach for mouse Hmouse
• Point to word Pword
• Select word K
• Home on keyboard Hkeyboard
• Call replace cmd M,Kreplace
• Type new 5 letter word 5Kword
• Texecute = TM + TP + 2TH + 7Tk
10/5/2010 65
Where does KLM apply?
• Where users perform tasks at expert level
– users have mastered a skill
– users are not problem solving
– users know what to do, just act on the steps
• Qualitatively
– Used for designing training programs, help systems
– Focus design on problem areas
• Quantitatively
– Good predictions of performance time
– Maybe some predictions on learning
10/5/2010 66
KLM weaknesses
• The model applied to skilled users, not to beginners or intermediates
• The model doesn't account for either learning of the system or its recall after a period of disuse
• Even skilled users occasionally make errors; however, the model doesn't account for errors
• Mental workload is not addressed in the model
10/5/2010 67
Hick-Hyman Law
• Law for choice reaction-time, given in the form of a prediction equation
• Given a set of n stimuli (flashing objects), associated with n responses (selecting object), reaction time (RT) can be given as follows:
RT = a + b log 2(n) – a, b empirically determined constants
• Examples include:– measuring and predicting time to select items in a
hierarchical menu (Landauer & Nachbar, 1985)
– predicting text-entry rates on soft keyboards with non-qwerty layouts, since users have to visually scan the layout (MacKenzie et al., 1995, 1999)
– mode selection on tabletPCs (Ruiz & Lank, 2006)
10/5/2010 68
A Quiz Designed to Give You Fitts
•http://www.asktog.com/columns/022DesignedToGive
Fitts.html
• Microsoft Toolbars offer the user the
option of displaying a label below each
tool. Name at least one reason why
labeled tools can be accessed faster.
(Assume, for this, that the user knows
the tool and does not need the label just
simply to identify the tool.)
10/5/2010 69
A Quiz Designed to Give You Fitts
1. The label becomes part of the target. The
target is therefore bigger. Bigger targets, all
else being equal, can always be
acccessed faster. Fitt's Law.
2. When labels are not used, the tool icons
crowd together.
10/5/2010 70
A Quiz Designed to Give You Fitts
• You have a palette of tools in a graphics
application that consists of a matrix of 16x16-
pixel icons laid out as a 2x8 array that lies
along the left-hand edge of the screen.
Without moving the array from the left-hand
side of the screen or changing the size of the
icons, what steps can you take to decrease
the time necessary to access the average
tool?
10/5/2010 71
A Quiz Designed to Give You Fitts
1. Change the array to 1X16, so all
the tools lie along the edge of the
screen.
2. Ensure that the user can click on
the very first row of pixels along
the edge of the screen to select a
tool. There should be no buffer
zone.
10/5/2010 72
A Quiz Designed to Give You Fitts
• Microsoft offers a Taskbar which can be oriented
along the top, side or bottom of the screen,
enabling users to get to hidden windows and
applications. This Taskbar may either be hidden or
constantly displayed. Describe at least two reasons
why the method of triggering an auto-hidden
Microsoft Taskbar is grossly inefficient.
10/5/2010 73
A Quiz Designed to Give You Fitts
• Screen edges are prime real estate. You don't waste an entire edge that could be housing a couple of dozen different fast-access icons just for one object, the Taskbar
• The auto-hidden Taskbar is entirely too easy to display by accident. Users are constantly triggering it when trying to access something that is close to, but not at, the edge
• The Taskbar would not have any of these problems, yet be even quicker to get to if it were located at any one of four corners of the display. Throw the mouse up and to the left, for example, and you'll have a taskbar displayed. Fast access without the false triggering
10/5/2010 74
Fitts‘ Law
• Robust and highly adopted model of human movement
• Originated as interest of applying information theory to the
analysis and understanding of difficulty of movement tasks &
human rate of information processing
• Used Shannon‘s law for information capacity
C = B log2(S / N + 1)
» S is the signal power and N is the noise power
• Based on the following analogies:
– Amplitude of aimed movement == electronic signal
– Spatial accuracy of movement == electronic noise
10/5/2010 75
Fitts‘ Law
• Described the analogy in two
papers:
– a serial, or reciprocal, target
acquisition task wherein subjects
alternately tapped on targets of
width W separated by amplitude A
– experiment using a discrete task,
wherein subjects selected one of two targets in response to a stimulus
light
Fitts, P. M. (1954). The information capacity of the human motor
system in controlling the amplitude of movement. Journal of
Experimental Psychology, 47, 381-391.
Fitts, P. M., & Peterson, J. R. (1964). Information capacity of
discrete motor responses. Journal of Experimental Psychology, 67,
103-112.
10/5/2010 76
Fitts‘ Law
• Quantify a movement task's difficulty — ID, the
index of difficulty
ID = log2(A / W + 1) (bits)
A = amplitude, W = width of object
• Movement time to complete a task is predicted
using a linear equation
MT = a + b * ID (secs)a & b are empirically determined using linear regression
• Throughput (TP) or Index of Performance (IP) is
TP = ID / MT (bits/sec)
10/5/2010 77
Fitts‘ Law
• To determine a and b design a set of tasks with
varying values for A and W (conditions)
• For each task condition
– multiple trials conducted and the time to execute each is recorded and stored electronically for statistical analysis
• Accuracy is also recorded
– either through the x-y coordinates of selection or
– through the error rate — the percentage of trials selected
with the cursor outside the target
10/5/2010 78
Fitts‘ Law
Same ID → Same Difficulty
Target 1 Target 2
10/5/2010 79
Fitts‘ Law
Smaller ID → Easier
Target 2Target 1
10/5/2010 80
Fitts‘ Law
Larger ID → Harder
Target 2Target 1
10/5/2010 81
Fitt’s Law
A (pixels) W (pixels) ID (bits) Device 'A' Device 'B'
ER (%) MT (ms) ER (%) MT (ms)
40 10 2.32 2.08 665 1.25 1587
40 20 1.58 3.33 501 2.08 1293
40 40 1.00 1.25 361 0.42 1001
80 10 3.17 2.92 762 2.08 1874
80 20 2.32 1.67 604 2.08 1442
80 40 1.58 1.67 481 0.83 1175
160 10 4.09 3.75 979 2.08 2353
160 20 3.17 5.42 823 1.67 1788
160 40 2.32 4.17 615 0.83 1480
Mean: 2.40 2.92 644 1.48 1555
Example data sets for two devices from a Fitts' law experiment
10/5/2010 82
Fitt‘s Law
10/5/2010 83
Fitt‘s Law
• If primary goal in Fitts‘ law experiment is to determine performance between devices/interaction techniques, then throughput (TP) is best criterion
– TP = ID/MT
• If for a given device ID = 4.09 bits and task is executed in MT = 979 ms– human rate of information processing for that task is 4.09 / 0.979 =
4.18 bits/s or TP = 4.18 bits/s
• Mean throughput across all the A-W conditions for Device 'A' is TP = 2.40 / 0.644 = 3.73 bits/s
• For Device 'B', TP = 2.40 / 1.555 = 1.57 bits/s
• Using throughput we conclude users' performance with Device 'A' was about 3.73 / 1.57 = 2.4 times better than performance with Device 'B'
10/5/2010 84
Setting it up
– MacKenzie, I. S. (1995). Movement time prediction in human-computer interfaces. In R. M. Baecker, W. A. S. Buxton, J. Grudin, & S. Greenberg (Eds.), Readings in human-computer interaction (2nd ed.) (pp. 483-493). Los Altos, CA: Kaufmann. [reprint of MacKenzie, 1992]
– http://www.yorku.ca/mack/GI92.html
• Vary A,W values for at least 4 ID conditions– Small A, small W– Small A, large W
– Large A, small W
– Large A, large W
– 2-4 variations in between
• Clicking start position presents object to click on– Record whether user missed
– Record time to click on stimulus
• Design with several repetitions and several blocks
10/5/2010 85
Case Study #1: Text Entry Rates on Mobile Phones
• Can we predict text entry rate on mobiles using Fitts‘ Law?
10/5/2010 86
Case Study #1: Text Entry Rates on Mobile Phones
• Two main approaches:– Multi-tap:
• presses each key one or more times to specify the input character
• large overhead: 33 key presses 15 characters of text
• "on average" multi-tap method requires 2.034 keystrokes per character
77 88 444 222 55 0 22 777 666 9 66 0 333 666 99
q u i c k _ b r o w n _ f o x
– One-key disambiguation:• Add linguistic knowledge to make best guess
• Can be ambiguous in some cases, have to correct
7 8 4 2 5 0 2 7 6 9 6 0 3 6 9
q u i c k _ b r o w n _ f o x
10/5/2010 87
Case Study #1: Text Entry Rates on Mobile Phones
• Text entry on a mobile phone, for example, consists
of aiming for and acquiring (pressing) a series of
keys "as quickly and as accurately as possible―
• Time to press any key, given any previous key, can
be readily predicted using Fitts' law
• For index finger input = MT = 165 + 52 ID
• and for thumb input = MT = 176 + 64 ID
10/5/2010 88
Case Study #1: Text Entry Rates on Mobile Phones
• Elements to build a text-entry prediction model are:
– Information on position and size of keys (ruler)
– Letter assignment to keys (any phone)
– Relative probabilities of digrams (probabilities of letter pairs) in target language (sources)
A B C D Z Space
A 0.00002 0.00130 0.00290 0.00360 0.00011 0.00047
B 0.00130 0.00013 0.00000 0.00000 0.00000 0.00034
C 0.00340 0.00000 0.00012 0.00000 0.00000 0.00044
D 0.00099 0.00001 0.00000 0.00035 0.00000 0.02500
Z 0.00003 0.00000 0.00000 0.00000 0.00008 0.00002
Space 0.01800 0.00960 0.00810 0.00480 0.00002 0.00000
t-h or e-space have high P
g-k or f-v have low P
10/5/2010 89
Case Study #1: Text Entry Rates on Mobile Phones
Method Predicted Expert Entry Rate (wpm)
Index Finger Thumb
Multi-tap
- wait for timeout
- timeout kill
22.5
27.2
20.8
24.5
One-key with disambiguation 45.7 40.6
• Time to enter each i-j sequence is predicted using
Fitts‘ law giving MTij, weighted by the probability
of the digram in the target language Pij
MTL = ∑∑ (Pij × MTij )
WPM = MTL × (60 / 5) (avg 5 chars/word)
2 assumptions:
- all words are in dictionary
- when ambiguity arises the intended word is the most probable
10/5/2010 90
Case Study: Using Fitts‘ to redesign text entry
http://www.exideas.com/ME/
http://www.exideas.com/ME/Pressfolder/PressReleaseDec2-2003.html
Nesbat, S. “A System for Fast, Full-Text Entry for Small Electronic
Devices“, Proceedings of the Fifth International Conference on Multimodal
Interfaces, ICMI 2003 (ACM-sponsored), Vancouver, November 5-7, 2003.
MessagEase Onscreen Keyboard
Example of an interface design
which can be adapted to multiple
devices
10/5/2010 91
Diverging Text Entry Technologies
Device Text Entry Technology
Cell phones Multi-tap
PDAs and Tablet computers Graffiti, Jot, QWERTY
Email devices QWERTY
TV remote controllers, watches Scroll and Pick
Car navigation systems Several proprietary
Point of Sale devices ABCD, multi-tap
10/5/2010 92
Using Letter Frequency
10/5/2010 93
Nine Most Frequent Letters: Double Click
E
T
N
R
O
I
A
S
H
10/5/2010 94
Eight Less Frequent Letters: Two Taps
D
C
U
P
G
B
Q
J
10/5/2010 95
Remaining Nine Letters: Two Taps
F
M
Y
W
V
X
K
Z
10/5/2010 96
Adding Space, Shift, and Mode
10/5/2010 97
Special Characters
38 special characters entered by two taps;
6000+ characters can be entered with combine.
10/5/2010 98
Soft Keyboard Design
Hard Key Soft Key
Most Frequent Letters Double Click Single Tap
Less Frequent Letters Two Clicks Single Drag
The same
mapping
used for
letters
10/5/2010 99
Special Characters
Entered with a single drag
10/5/2010 100
Optimization and Evaluation
• Exhaustively simulated all permutations of letters within each group
• The configuration with the max speed was selected
10/5/2010 101
Fitts‘ Law
Movement Time from one
key to another:
MT = a + b*log2(A/W+1)
WA
10/5/2010 102
Digraph Probability
The probability Pij that letter j will
follow letter i in a body of text:
Pij = 1
A B C D Z Space
A 0.00002 0.00130 0.00290 0.00360 0.00011 0.00047
B 0.00130 0.00013 0.00000 0.00000 0.00000 0.00034
C 0.00340 0.00000 0.00012 0.00000 0.00000 0.00044
D 0.00099 0.00001 0.00000 0.00035 0.00000 0.02500
Z 0.00003 0.00000 0.00000 0.00000 0.00008 0.00002
Space 0.01800 0.00960 0.00810 0.00480 0.00002 0.00000
10/5/2010 103
Performance Measure – Hard Key
Calculation of max theoretical entry speed:
• Movement Time
–MT = a + b log2(A/W+1)
• Total time (2 Clicks)
–CT = MT1 + MT2
• Total time (Dble Click – no
movement)
–CTDC = 2a + b log2(A/W+1)
• Average Time
–CTav = (Pij × CTij)
• Speed
–WPM = (1/ CTav) ×(60/5)
10/5/2010 104
Performance Measurement – Soft Key
– Most frequent characters – Single tap: • TLi = (1/4.9) log2 [(D0-i/W) + 1]; if D0-i> 0,
• TLi = a; if D0-i = 0
– Less frequent characters – Drag:• TLjk = t0-j + tdown + tj-k + tup
• TLjk = (t0-j + tdown+ tup) + (tj-k + tup+ tdown) –(tdown+ tup)
• TLjk = TLj + TLk– a t0-j: time to move to key j
tj-k: time to move from key j to key k
tdown: time to move stylus down
tup: time to move stylus up
10/5/2010 105
Hard Key (Cell Phone) Comparison
User Study
0
2
4
6
8
10
12
Multi-tap MessagEase
WP
M
Theoretical
0
10
20
30
Multi-tap MessagEase
WP
M
130%
209%
10/5/2010 106
10
30
3638
43
50
0
10
20
30
40
50
60
Graffiti
Comparison with Other Soft Keyboards
Includes all special
characters
Graffiti QWERTY Fitaly Opti Metropolis MessagEase
(all based on Fitts’ law except for Graffiti)
10/5/2010 107
Comparison with Other Soft Keyboards
• Given the same area MessagEase has:
– Only 11 keys, but enters more characters
– Its Keys are 3-5 times bigger; therefore faster!
10/5/2010 108
Advantages and Disadvantages
Disadvantages Advantages
New letter assignment, requires
learning a new pattern
Applicable to both hard-key and Soft-
key
Some keys may become clutteredFull text entry:
letters, numbers, special Chars.
Size and language agnostic
Deterministic and unambiguous
One handed (hard-key) operation
10/5/2010 109
Applications:
Any small mobile device that cannot sport a full QWERTY keyboard
10/5/2010 110
Watches
•Scroll-and-Pick is too slow
•QWERY is too small
•MessagEase will have buttons
large enough for a stylus
10/5/2010 111
Car Navigation Systems
•One handed touch typing
•Unifying an otherwise
fragmented text entry systems
used
10/5/2010 112
Example: Expanding targets
• Acquisition of Expanding Targets. Proceedings of
ACM Conference on Human Factors in Computing
Systems (CHI) 2002, pages 57-64
10/5/2010 113
Example: Expanding targets
Furnas
Generalized fisheye views
CHI 1986 Bederson
Fisheye Menus
UIST 2000
Mackinlay, Robertson, Card
The Perspective Wall
CHI 1991
10/5/2010 114
Mac OS X ―dock‖
Does this make acquisition easier ?
• Size of the interface widget (viewing region)
changes dynamically
– Provide the user with a magnified target area at their focus
of attention (area around the cursor)
– Expanding toolbar implemented in latest Apple OS X
operating system
10/5/2010 115
Advantages and Disadvantages
• Advantages
– Icons are displayed in reduced size to solve the
increasing number of commands and icons
– Larger amount of screen real estate devoted to
the display of the underlying data
• Disadvantages
– Can reduce the user‘s ability to select the
desired icon efficiently
10/5/2010 116
Fitts‘ Law
• 3 different scenarios describing what Fitts‘ Law is modeling
– Iterative corrections model
– Impulse variability model
– Optimized initial impulse model
10/5/2010 117
Iterative Corrections Model
• States that the movement consists of many discrete
sub-movements
• Each sub-movement takes the user closer to the
target
• Sub-movements are triggered by feedback
indicating the target has not been reached yet
10/5/2010 118
Impulse Variability Model
• Movement consists of initial impulse delivered by
the muscles towards the target, flinging the limb
towards the target
• Last part of movement time consists of limb
coasting towards target
• Either type of explanation cannot explain the Fitt‘s
Law completely given a range of tasks
10/5/2010 119
Optimized Initial Impulse Model (1)
• Most complete and successful explanation to
the Fitts‘ Law
• Combination of the iterative corrections and
the impulse variability models
– movement is initiated towards the target
– task is completed if the movement lands at the
target
– another movement is required if it lands outside
the target
– same processes will be carried out until the
target is reached
10/5/2010 120
Distance
Speed
What does Fitts‘ Law really model?
W
UndershootOvershoot
Open-loop
Closed-loop
10/5/2010 121
Expanding Targets
• Basic Idea:– Big targets can be acquired faster, but take up
more screen space
– So: keep targets small until user heads toward them
– Can this be used for devices with small viewing space?
Cancel
Okay
Click Me !
10/5/2010 122
Experiment Goals
The experiment was designed to answer the following
questions for a typical expanding target selection
task:
1. Can such a task be modeled by Fitts‘ law?
2. If it can be modeled by Fitts‘ law, is it possible to predict
performance in such tasks from a base set of data where
no expansion takes place?
3. Is movement time dependent on the final target width and not the initial one at onset of movement?
4. At what point should the target begin expanding?
5. Do different target expansion strategies affect
performance?
10/5/2010 123
Experimental Setup
Target
Start Position
W
A
10/5/2010 124
Experimental Setup
Expansion:
How ?
Animated
Expansion
10/5/2010 125
Experimental Setup
Expansion:
How ?
Fade-in
Expansion
10/5/2010 126
Experimental Setup
Expansion:
How ?
When ?P = 0.25
10/5/2010 127
Experimental Setup
Expansion:
How ?
When ?P = 0.5
10/5/2010 128
Experimental Setup
Expansion:
How ?
When ?P = 0.75
10/5/2010 129
Pilot Study
• 7 conditions:
– No expansion (to establish a, b values)
– Expanding targets
• Either animated growth or fade-in
• P is one of 0.25, 0.5, 0.75
– (Expansion was always by a factor of 2)
10/5/2010 130
Pilot Study
7 conditions
x 16 (A,W) values
x 5 repetitions
x 2 blocks
x 3 participants
= 3360 trials
10/5/2010 131
Pilot Study: Results
Time
(seconds)
ID (index of difficulty)
10/5/2010 Comp 4020 - HCI 2 (PPI) 132
Pilot Study: Results
Time
(seconds)
ID (index of difficulty)
)1(log2 W
Aba
10/5/2010 Comp 4020 - HCI 2 (PPI) 133
Pilot Study: Results
Time
(seconds)
ID (index of difficulty)
)1(log2 W
Aba
)12
1(log2
W
Aba
10/5/2010 134
Pilot Study: Results
Time
(seconds)
ID (index of difficulty)
P = 0.25
10/5/2010 135
Pilot Study: Results
Time
(seconds)
ID (index of difficulty)
P = 0.5
10/5/2010 136
Pilot Study: Results
Time
(seconds)
ID (index of difficulty)
P = 0.75
10/5/2010 137
• Pilot Study suggests the advantage of
expansion doesn‘t depend on P
• So, set P = 0.9 and perform a more rigorous
study
Implications
10/5/2010 138
Full Study
• 2 conditions:
– No expansion (to establish a, b values)
– Expanding targets, with
• Animated growth
• P = 0.9
• Expansion factor of 2
10/5/2010 139
Full Study
2 conditions
x 13 (A,W) values
x 5 repetitions
x 5 blocks
x 12 participants
= 7800 trials
10/5/2010 140
Results
Time
(seconds)
A, W values
10/5/2010 141
Results
Time
(seconds)
10/5/2010 142
Results
Time
(seconds)
10/5/2010 143
Results
Time
(seconds)
10/5/2010 144
Results
Time
(seconds) P = 0.9
10/5/2010 145
Implications
• For single-target selection task,
– Expansion yields a significant advantage, even
when P=0.9
• What about multiple targets ?
10/5/2010 146
Implications for Design (1)
• Experimental results can influence the design of
buttons, menus, or other selectable widgets
• Interface with multiple expanding targets does not
need to predict cursor's trajectory to anticipate
which widgets to expand
– Instead, just expand widgets as the cursor approaches
them
10/5/2010 147
Implications for Design (2)
• Expansion Strategies for adjacent
widgets (e.g. toolbars)
• Expanding a widget around its
center will cause overlap &
occlusion with nearby widgets– Expanding a group of widgets around a
group’s center
– Expand nearest widgets and move
adjacent widgets away
– Expand nearest widgets, but allow some
overlap as well as expand adjacent widgets
so they are easier to see
10/5/2010 148
Summary
• Expanding targets acquisition can be accurately
modeled by Fitts‘ Law
• User performance is aided by target expansion
• Targets that are always expanded can be acquired
just as fast as targets that expand just as the user
reaches them
• Implications of results can be applied towards the
design of UI widgets for devices with limited viewing
space
10/5/2010 149
Improvement to Fitts‘: Bubble Cursor
TO READ:
• Tovi Grossman, Ravin Balakrishnan. The Bubble
Cursor: Enhancing target acquisition by dynamic
resizing of the cursor’s activation area, ACM
CHI, 2005, p. 281-290.
10/5/2010 150
Bubble Cursor
10/5/2010 151
Bubble Cursor
Improvements by
– Decreasing A
• Drag-and-pop [Baudisch et al.]
• Object pointing [Guiard et al.]
– Increasing W
• Area cursor [Kabbash & Buxton]
• Enhanced area cursor [Worden at al]
• Expanding targets [McGuffin & Balakrishnan]
– Decreasing A and Increasing W
• Semantic pointing [Blanch et al]
10/5/2010 152
Design of Bubble Cursor
Modification to area cursor
Problem! Circular cursor resolves problem
10/5/2010 153
Design of Bubble Cursor
Size problem
10/5/2010 154
Design of Bubble Cursor
10/5/2010 155
Design of Bubble Cursor
10/5/2010 156
Design of Bubble Cursor
10/5/2010 157
Design of Bubble Cursor
10/5/2010 158
Results
10/5/2010 159
Results
10/5/2010 160
Experiment 2
10/5/2010 161
Results
10/5/2010 162
Models for Trajectory-Based HCI Tasks
• Trajectory tasks are becoming more common
– Navigating through nested menus
– Drawing curves
– Moving in 3D worlds
• Cannot be successfully modeled using Fitts‘ law
• ―Steering through tunnel‖ as paradigm to represent
trajectory-based tasks
“Beyond Fitts’ Law: Models for trajectory based HCI tasks.”
Proceedings of ACM CHI 1997 Conference
10/5/2010 163
Beyond pointing: Trajectory based tasks
10/5/2010 164
Beyond pointing: Trajectory based tasks
• Experimental paradigm focused on is steering
between boundaries (constrained motion)
• It appears that the time to produce trajectories sets
the relative speed-accuracy ratio: the larger the
amplitude, the less precise the result is.
• Want to derive and validate quantitative
relationships between completion time and
movement constraints in trajectory-based tasks
10/5/2010 165
Beyond pointing: Trajectory based tasks
10/5/2010 166
Beyond pointing: Trajectory based tasks
10/5/2010 167
Beyond pointing: Trajectory based tasks
10/5/2010 168
Steering Law
10/5/2010 169
Crossing Based Interfaces: Motivation
• Pointing is most universal and best adapted
interaction paradigm in human computer
interfaces
• However some disadvantages exist:
– time-consuming if object pointed to is small
– pointing-driven widgets consume screen real estate
– double-clicking is not trivial for novice users, due to rapid
succession of clicks (temporal dependence)
“More than dotting the i's --- foundations for crossing-based interfaces.” Johnny Accot ,
Shumin Zhai. Proceedings of the SIGCHI conference on Human factors in computing systems.
April 2002
10/5/2010 170
Crossing interfaces: example designs
10/5/2010 171
Crossing interfaces: example designs
10/5/2010 172
Descriptive Models
• Do not yield empirical or quantitative measure of user performance– Not predictive
• Provide a framework/context for thinking methodologically about a problem or situation
• Typically a verbal or graphical description of classes or identifiable features in an interface– Facilitates categorization
• Purpose is to give the designer a tool for studying and thinking about user-interaction experience
• Two examples:– KAM
– Mapping DOF to Dimensions
– Guiard‘s Theory of Bimanual Interaction
10/5/2010 173
Key-Action Model
• Keyboards are essential in interacting with a
computer
• 101-keys categorized by function keys, numeric
keys, characters, control keys, etc.
• Define a descriptive model referred to as Key-
Action Model (KAM) as follows:
– Symbol keys – deliver graphical symbols (a,z,1,?)
– Executive keys – invoke actions in the application or at the
system level (ENTER, F1, ESC)
– Modifier keys – set up a condition to modify effect of
pressing a key (ALT, SHIFT, CTRL)
10/5/2010 174
Key-Action Model
• Some questions to consider regarding this model:
– Is the model correct?
– Is it flawed?
– Do all keyboard keys fit the model?
– Are there additional categories or sub-categories?
– Can it be improved to become more comprehensive?
– Is the model really useful?
10/5/2010 175
Guiard‘s Model
• Originated from the area of motor behavior
referred to as bimanual control or laterality
• Both hands perform a different set of tasks
• Given this knowledge and handedness of people,
interesting to evaluate how interaction
accommodates best the division
• Results in descriptive model of bimanual skill, given
by Guiard in 1987 paper
Guiard, Y. (1987). Asymmetric division of labor in human skilled bimanual action:
The kinematic chain as a model. Journal of Motor Behavior, 19, 486-517.
10/5/2010 176
Guiard’s Model
Hand Role and Action
Non-preferred • leads the preferred hand
• sets the spatial frame of reference for the preferred hand
• performs coarse movements
Preferred
• follows the non-preferred hand
• works within established frame of reference set by the non-
preferred hand
• performs fine movements
10/5/2010 177
Guiard‘s Model
• Example:
– a right-handed artist sketches a design of car
– acquires a template with left hand (non-preferred hand
leads)
– template is manipulated over the workspace (coarse
movement, sets the frame of reference)
– Right hand picks stylus (preferred hand follows) and placed
close to the template (works within frame of reference set
by the non-preferred hand)
– Artist sketches (preferred hand makes precise movements)
10/5/2010 178
Guiard‘s Model
• People naturally gravitate to using two hands
• Performance times are reduced
• Can be used for interfaces that employ:
– Drawing designs
– Fabricating virtual objects
– Positioning
– Reshaping
10/5/2010 179
Bimanual Control and Desktop Computer Affordances
• How does the distribution of keys on a keyboard facilitate
task division between right/left hands?
– where does interaction with the mouse fit into the model?
• Right side bias toward power keys (executive keys +
modifier keys, marked in red dots)
10/5/2010 180
Bimanual Control and Desktop Computer Affordances
• Dominance on right hand side, good
for the 80s but how does this work with
GUIs and pointing devices that are now
commonplace?
• Right handed have to reach over with
left or leave the mouse
• Is there an advantage for left-handed
users?
10/5/2010 181
Bimanual Control and Desktop Computer Affordances
Task Leading Movement Trailing/Overlapping
Movement
Delete
Left hand — manipulate pointer
with mouse and select text/object
by double clicking or dragging
Right hand — press DELETE
(probably with little finger)
Select an option
in a window
Left hand — manipulate pointer
with mouse and click on an option
Right hand — press ENTER
(Note: OK button is the default)
Click on a link in a
browser
Right hand — navigate to link via
PAGE UP and/or PAGE DOWN
keys
Left hand — manipulate
pointer with mouse and select
link by clicking on it
Open a file, open
a folder, or launch
a program
Left hand — manipulate pointer
with mouse and single click on icon
Right hand — press ENTER
(Note: avoids error prone
double-click operation)
Common tasks performed by a left-handed user manipulating mouse in the left hand
10/5/2010 182
Bimanual Control and Desktop Computer Affordances
• Tasks described previously are faster for left-handed users than right-handed users
• When pointing is juxtaposed with power key activation (excluding SHIFT, ALT, & CONTROL), the desktop interface presents a left-hand bias
10/5/2010 183
Bimanual Control and Desktop Computer Affordances
• Scrolling typically accomplished by dragging ―elevator‖ of scrollbar along the right-hand side of an application’s window
•Takes up to 2 secs per trial and is obtrusive and non-transparent
• In perspective of Guiard’s model of bimanual control, we can delegate scrolling to non dominant hand
Task Characteristics
Scrolling• precedes/overlaps other tasks
• sets the frame of reference
• minimal precision needed (coarse)
Selecting, editing, reading,
drawing, etc.
• follows/overlaps scrolling
• works within frame of reference set by scrolling
• demands precision (fine)
10/5/2010 184
Redesigning the Scrolling Interface
10/5/2010 185
What did we cover
• Text entry rate prediction on mobile phones
• Bimanual control and analysis of task division
between right and left hands
– Application to guide design choices
• Application of Fitts, expanding targets
• MessagEase: a commercial application
making use of Fitts‘ law for improving text
entry rates