Computer Science Honours 2011
Predictive evaluation technique requires a model for how a user interacts with an
f
interface
Model is
abstract
quantitative
approximate
estimated from user experiments
D ’ h b ild UI
Don’t have to build UI prototype
can compare design alternatives with no implementation whatsoever
implementation whatsoever
Don’t have to test real live users
Theory provides explanations of Theory provides explanations of UI problems UI problems
so it points to the areas where design can be improved
user testing may only reveal problems, not explain them
B d t f b h i
Based on expert error‐free behavior
Si th bli ti f “Th P h l f H
Since the publication of “The Psychology of Human‐
Computer Interaction”, the GOMS model has been one of the most widely known theoretical concepts in HCI y p
Card, Moran, & Newell in 1980
Used by UI professionals to model user behaviour
User's behaviour is modelled in terms of:
User s behaviour is modelled in terms of:
Goals
Operatorsp
Methods
Selection rules
Goals
Goals
are what the user is trying to accomplish.
these can be defined at various levels of abstraction, from very high level goals (e g WRITE RESEARCH PAPER) to low level high‐level goals (e.g. WRITE‐RESEARCH‐PAPER) to low‐level goals (e.g. DELETEWORD).
higher‐level goals are decomposable into subgoals, and are arranged hierarchically
a a ged e a c ca y
Operators
are the elementary perceptual, motor or cognitive actions that are used to accomplish the goals (e.g. DOUBLE‐CLICK‐MOUSE,
d bl
PRESS‐INSERT‐KEY). Operators are not decomposable
it is generally assumed that each operator requires a fixed amount of time for the user to execute, and that this time
interval is independent of context (e g CLICK MOUSE button interval is independent of context (e.g. CLICK‐MOUSE button takes 0.20 seconds to execute)
M th d
Methods
are the procedures that describe how to accomplish goals.
a method is essentially an algorithma method is essentially an algorithm that the user has that the user has
internalized that determines the sequence of subgoals and operators necessary to achieve the desired goal
Selection rules
Selection rules
specify which method should be used to satisfy a given goal, based on the context
since there may be several different ways of achieving the same goal, selection rules represent the user's knowledge of which method must be applied to achieve the desired goal
f f
Four popular variants of the GOMS family:
CMN‐GOMS (original)
Keystroke‐Level Model (KLM)
NGOMSL
CPM‐GOMS
All produce quantitative and qualitative
predictions of how people will use a proposed
system
Functionality coverage
Functionality coverage
if the designer has a list of likely user goals, GOMS models can be used to verify that a method exists to achieve each of these goals
of these goals
Execution time
GOMS models can predict the time it will take for the user to carr o t a goal
to carry out a goal
▪ locate bottlenecks
▪ compare different UI designs to determine which one allows users to execute tasks quicker
to execute tasks quicker
Help systems
since GOMS models are an explicit representation of user activity, they can assist in designing help systems and
activity, they can assist in designing help systems and tutorials to assist users in achieving goals
I th i l t GOMS t h i OM l
Is the simplest GOMS technique: OM only
Estimate execution time for a task, the usability analyst lists the sequence of operators and then analyst lists the sequence of operators and then totals the execution times for the individual
operators
Breaks down the user’s behaviour into a sequence of the primitive operators
h i l (
i l f k)
physical (moving muscles to perform task)
mental
(breaks task into chunks and represent the
time needed for the user to recall the next step from
time needed for the user to recall the next step from
long‐term memory)
Classes of operators (Kieras, 1993): Classes of operators (Kieras, 1993):
K to press a key
P to point with a mouse to a target on a display
H to home hands on the keyboard or other device
H to home hands on the keyboard or other device
D to draw a line segment on a grid
B to mouse down or up
BB to mouse click
BBBB to mouse double click
MM to mentally prepare to do an action or a closely related to mentally prepare to do an action or a closely related series of primitive actions
R to represent the system response time during which the user has to wait for the systemy
Kieras, D. (1993). Using the keystroke-level model to estimate execution times. University of Michigan
Step 1 Lay out methods for doing the task
Step 1: Lay out methods for doing the task
Step 2: Write out the basic action sequence (keystroke‐
level operators)
Step 3: Select the operators and durations that will be
Step 3: Select the operators and durations that will be used (KPBHD)
Step 4: List the times next to physical operators for the task
task
Step 4a: Include system response time (R) in necessary
typically system response time appears instantaneous, so can be ignoredg
Step 5 : Next add the mental operators (M) and their times
Step 6 : Sum the times of the operators
Keystroke determined by typing speed
▪ 0.28 s average typist (40 wpm)
▪ 0.08 s best typist (155wpm)b
▪ 1.20 s worst typist
Pointing determined by Fitts’s Law
Pointing determined by Fitts s Law
▪ ~ 1.1 s for all pointing tasks
Drawing determined by steering law
Drawing determined by steering law
B mouse up or down (as in dragging) ~ 0.1 s
BB mouse click (single) ~ 0.2 s d bl l k
BBBB mouse double click ~ 0.4 s
Homing estimated by measurement 0.36 s
b k b d d
(between keyboard and mouse)
Mental preparation (for next step) estimated b
by measurement 1.35 s
Basic psychological principle: physical operations in methods are chunked into
b h d
submethods.
RULE 0: Place M operators
in front of all K, B, P
R M’ di t l l t d t
Remove M’s according to rules related to chunking of actions
Rule 1: Anticipated by prior operation p y p p
PMBB
PBB (point and then click is a chunk)
Rule 2: If string of MK’s or MB’s is a single cognitive unit delete all but first
cognitive unit, delete all but first
MKMKMK ‐> MKKK (same as MK3)
Rule 3: If K terminates a constant string, such as d E h d l M
command Enter, then delete M
MK3(dir)MK(enter)
MK3(dir)K(rtn) (typing “dir”
command followed by enter y is a chunk) )
OFFICE 2003
H (Home on mouse)
OFFICE 2007
H (Home on mouse)
P (Edit)
BB (click on mouse button
‐ press/release)
P (Find)
BB (click on mouse button ‐ press/release)
P (Find)
BB (click on mouse button)
H (Home on keyboard)
button press/release)
H (Home on keyboard)
K6 (Type six characters
i t Fi d di l b )
H (Home on keyboard)
K6 (Type six characters into Find dialogue box)
K (Enter key on keyboard)
into Find dialogue box)
K (Enter key on keyboard)
K (Enter key on keyboard)
Use Find Command to locate a six character word (Office 2003 vs. Office 2007):
H (Home on mouse)
Rule 0
( )
MP (Edit)
MBB (click on mo se b tton)
Rule 0
Rule 0
MBB (click on mouse button) MP (Find)
Rule 0
MBB (click on mouse button) H (Home on keyboard)
Rule 0
( o e o eyboa d)
MKMK... (Type 6 characters)
MK (R k di l b
Rule 0 Rule 0
MK (Return key on dialogue box starts the find)
Rule 0
H (Home on mouse) H (Home on mouse) MP (Edit)
( l k b )
Rule 1: delete M
MBB (click on mouse button) MP (Find)
u d
P anticipates B
MBB (click on mouse button) H (Home on keyboard)
Rule 1: delete M P anticipates B
H (Home on keyboard)
M6K (Type six characters)
Rule 2: single cognitive unit
MK (Return key on dialogue box starts the find)
Rule 3: delete M terminates a constant string
Office 2003 Office 2007 H
MP BB MP
Plug in real numbers from experiments K: 0.28 secs
P: 1.1 secs H 0 36
H MP BB BB
H M6K K
H: 0.36 secs BB: 0.2 secs M: 1.35 secs
BB H M6K K
2H = 0.72 3M = 4.05
2P = 2 2 9.33 sec
2H = 0.72 2M = 2.7
6 68 sec 2P = 2.2
2BB = 0.4 7K = 1.96
1P = 1.1 1BB = 0.2 7K = 1.96
6.68 sec
Now it’s your turn:
Shift-click selection vs. Del key 6 times – assume hand is on mouse to start
Shift‐click selection
Del key N times
M
P [start of word]
BB [click]
Del key N times
M
P [start of word]
BB [click] l k M
P [end of word]
K [shift]
BB [click]
H K [shift] M
BB [click]
H [to keyboard]
M
6K [Del]
Total: 2M + P + H + 6K +
M
BB
K [Del]
Total: 3M + 2P + 2K + 2BB
4 05 + 2 2 + 0 56 + 0 4 = 7 21s
= 6.04s 4.05 + 2.2 + 0.56 + 0.4 = 7.21s
• Tested against 11 different
• Tested against 11 different interfaces (3 text editors, 3 graphical editors, and 5
•command-line interfaces like FTP and chat)
• 28 expert users
• Each task done 10 times
• Within 20% of actual time
Comparing designs & methods
Parametric analysis
as we vary the parameter n (the length of the
word to be deleted)
O l t d i ti ( ll l d)
Only expert users doing routine (well‐learned) tasks
Only measures efficiency y y
not learnability, memorability, etc.
Ignores
errors (methods must be error free)
errors (methods must be error‐free)
parallel action (shift‐click)
mental workload (e.g. attention & g WM limits)
planning & problem solving (how does user select the method?)
fatigue
fatigue
N t h i ti l i id li
Not as easy as heuristic analysis, guidelines, or cognitive walkthrough
Only works for goal‐directed tasks
Only works for goal directed tasks
Assumes tasks are performed by expert users
Evaluator must pick users’ tasks/goals
Does not address several important UI issues, such as
d bili f
readability of text
memorability of icons, commands
Does not address social or organizational impact
Does not address social or organizational impact
f
Can KLM work for other devices?
Mobile phones
Cameras
…
Holleis, P., Otto, F., Hubmann, H., Schmidt, A.
(2007) Keystroke‐Level Model for Advanced
b l h f
Mobile Phone Interaction, SIGCHI Conference on Human Factors in Computing Systems,
l
San Jose, CA, April/May 2007