HCI Test 3 Flashcards
Arrange elements in one layer
(1) sequence if there is one
(2) functional groups (if there are)
(3) frequency then alphabetic order
Layout analysis
Arrangement of methods in one layer:
(1) group elements by their functions
(2) arrange functional groups according to their importance/sequence/frequency of usage
Pos and neg of layout analysis
+: easy to use, low cost
-: low reliability
Link analysis
Goal is to reduce the eye or motor movement distance on the interface
Pos and neg of link analysis
+: easy to use
-: ?
Types of UIs (3)
Command, graphic ui, multimodal (wearable)
Command interface advantages/disadvantages
+: lower demand on hardware
-: higher memory load on user, non-intuitive, poorer human performance
Graphic UI advantages/disadvantages
+: lower memory load on users, intuitive, better human performance
-: higher demand on hardware
Multimodal UIs (3)
- Pen gesture recognition
- Speech recognition
- Multimedia (movies, animations…)
Multimodal UIs advantages/disadvantages
+: utilize human’s capacity
-: higher requirement on hardware
Multiple Resource Theory
Wickens…
(1) Responses: Verbal, Spacial, Manual, Vocal
(2) Modalities: Visual, Auditory
(3) Codes: Spacial Verbal
(4) Stages: Encoding, Central Processing, Responding
Input devices (4)
(1) hands
(2) voice
(3) eyes
(4) other (foot, brain?)
Hand input devices (2)
(1) keyboards (qwerty, dvorak)
(2) handwriting
qwerty keyboard
sacrifice human performance because of usage habits
problems with qwerty keyboards
(1) workload: lh>rh
(2) some frequently used letters (eg. e) are not on the same row, but some non-freq used ones are on home row
dvorak keyboard
(1) infrequent keys leave the home row
(2) workload: rh>lh
handwriting and voice recognition - how it works
match characteristics of the input stream with stored patterns, many for each possible word
Handwriting and voice recognition - technical difficulties (4)
(1) segmentations - separate into letters, recognize
(2) individual differences -> program training
(3) voice - noise
(4) voice - privacy
Handwriting and and voice recognition - spatial and temporal segmentation issues
temporal - optimal waiting time
- spatial - optimal number and size of -windows
- recognition accuracy
Recognition accuracy vs. task completion time
downward slope, horizonal asymptote
Eye-direct control usage
(1) people with disabilities
(2) hands are busy
Dr. Hawking
- pneumonia
- tracheotomy
- machine that synthesized speech based on vibrations in trachea
- Siemens recently made new eye-direct control UI
2 types of brain-computer interfaces
(1) non-intrusive - outside of scalp
(2) intrusive - implanted
3 types of output devices
(1) visual
(2) auditory
(3) tactile
2 types of visual output devices
(1) traditional (CRT, LCD)
2) non-traditional (VR - immersive, augmented - semi-immersive
Traditional visual display
wide screen to fit pictures
non-traditional visual display - adv/disadv
+: 3D depth perception, tracking head motion
-: motion sickness, ?
auditory display type
3D sound system - applicaiton - truck driver warning system
3D sound system - truck driver warning system
(1) modality
(2) beep - voice might take time to process
Tactile display examples
frozen wind, lane departure warning system
Benefits of auditory and tactile display
(1) when visual modality is occupied or very busy, auditory and tactile information will utilize the other modalities to convey information
(2) you may neglect to see it, but it is hard to neglect to hear it
(3) in some circumstances, it is more natural than visual display (e.g. the departure warning example)
3 types of I/O types
(1) traditional
(2) VR
(3) Auditory and tactile
Motivations to model
(1) predict and generate human behavior
(2) evaluate and improve interface design efficiently (save time and expense of experiment)
(3) unify many experimental studies
(4) model can be integrated into intelligent system design
KLM
Keystroke level model by Card, prediction of user performance time by adding each step’s time up
Elements of KLM
K - key P - point mouse H - home on keyboard M - mentally prepare R - system response time
assumption of KLM
single task and there is no overlap
KLM’s pos/neg
+: easy to learn, quick to use
-: no practice effect, no overlap among steps, no hierarchical structure, no fatigue effect
GOMS
Goals, Operators/Methods, and Selection Rules by Card, Moran Newell, a hierarchical analysis of task steps and estimation of performance time
founders of AI and cognitive science
Newell and Simon
Variation of GOMS
NGOMSL - natural GOMS language - Kieras
NGOMSL
Method for goal - followed by steps (procedure)
Selection rules for goal - followed by if/then statements (if text is word, then accomplish goal: Highlight arbitrary text.)
Major improvement - add if…then, and return
GOMS/NGOMSL pos/neg
+: hierarchical analysis, more flexible (e.g. if-then rule)
-: single task, affected by different user strategies
CPM-GOMS
Critical Path Method-GOMS
GOMS setup
Goal:...
[select: Goal:...
- ....
- ...]
Goal:...CPM-GOMS setup
Visual Perception
Cognitive Operators
Eye Movement
CPM-GOMS pos/neg
+: Multiple Tasks
-: Only at the time domain, time consuming
Classifications of Modeling
(1) KLM/GOMS - procedure modeling
(2) Simulation - production systems modeling
(3) Math Modeling - Deterministic, Stochastic
Production Systems Theory guy
Herbert Simon
HAM/ACT-R founder
John Anderson
SOAR founder
Allen Newell
EPIC and GOMS founder
David Kieras
Discrete x Serial Stages models
Subtractive, Additive, General Gamma
Discrete x Network Configurations models
Critical Path Network
Continuous x Serial Stages models
Cascade, Queue series
Continuous x Network Configurations models
Queuing network
Discrete x Procedure Models and Methods
CPM-GOMS
Continuous x Procedure Models and Methods
QN-MHP
Discrete x Production Systems
SOAR
Continuous x Production Systems
CAPS
4 types of simulation models
(1) EPIC
(2) ACT-R
(3) CAPS
(4) QN-MHP
EPIC
Executive Process-Interactive Control
- Simulation Model
- Core Assumption: No processing limit in the cognition part, limit is in the motor part
- Certain parameters come from MHP
ACT-R
Adaptive Control of Thought - Rational
- Simulation Model
- Core Assumption: Cognitive System works in serial manner
- Perceptual/Motor from EPIC
ACT-R progression
HAM -> ACT-R 1.0 (no motor/perceptual) -> ACT-R/PM (Motor Perceptual from EPIC)
SOAR
An architecture for general intelligence
- Simulation model
- Core Assumption: AI model - no processing limit
- Used in many systems as AI rather than cognitive model of human (e.g. missiles system)
QN-MHP
Queuing network Model Human Processor
- Human behavior emerges naturally as entities are presented in the different routes in the network
AI versus cognitive modeling
AI: Realize the function as human, but don’t care how human actually does it
Cog model: focus on how human actual performs
Ex: dish washer
2 types of math models
SEEV, model verification
SEEV
- Salience (bottom-up processing)
- Effort
- Expectancy (top-down factor calibrated to bandwidth of events that occur at location)
- Value (importance and relevance)
SEEV formula
P(A) = S + Ex + V - Ef
How to verify model’s prediction
trajectory of eye movement
2 indices to judge a model
R squared, RMS
R squared
square of correlation coefficients (trend/pattern)
RMS
Root Mean Square
sqrt(sum of xi^2/n) where xi is Model value - Data value and n = number of conditions
SATO
T = a + blog2(D/W)
Why evaluate and test?
to get a mental model of the user
Where to evaluate and test?
- usability testing room (easy to control variables, subjects may change their real behavior)
- natural task setting
How to evaluate and test?
usability evaluation methods, usability testing methods
usability evaluation methods
(1) cognitive walkthrough
(2) think aloud method
(3) cooperative evaluation
(4) checklist
4 questions for cognitive walkthrough
(1) do users have a goal in mind
(2) do users notice that there are cues to complete their goals?
(3) can users link the correct cue with their goal?
(4) if users perform a correct/wrong action, do they get feedback?
4 notes for cognitive walkthrough
(1) take the perspective from users
(2) consistent with design principles
(3) very easy to use and no subjects needed
(4) can only find 40% of usability problems
3 notes for think aloud method
(1) one of the most effective ways to explore users’ real-time thoughts
(2) only use small amount of subjects
(3) subjects “say while doing”
pos/neg for think aloud
+: discover what’s going on in users’ minds
-: intrusive, difficult to conduct for large number of subjects
3 notes for cooperative evaluation
(1) need a small number of subjects/users
(2) an experimenter and subject/user cooperatively explore the UI and complete several tasks
(3) subject/user think aloud during the process
cooperative evaluation pos/neg
+: very natural way to find usability problems
-: users’ verbal response and behavior might be affected by experimenter, only a few subjects
QUIS
- questionnaire for user interface satisfaction
- developed by Chin et al.
- can be used either by designers or users
- a detailed implementation of design principles
- categorized evaluation
- not only usability, but some perception of hardware issues
QUIS pos/neg
+: quick and easy
-: sometimes not able to provide specific design suggestions
SUMI
- Software Usability Measurement Inventory
- developed by Kirakowski
- mainly for software
SUMI pos/neg
+ easy and relatively quick, free
-: for software only, has only 3 choices, no room for open-ended comments
Usability testing - two major ways
(1) test UI prototypes without formal experimental design (identify usability problems, improve the interface design quickly
(2) Test UI prototypes with formal experiment design (find optimal design, benchmark/products, comparisons, etc
Steps to testing UI prototypes without formal experimental design
(1) build prototypes
(2) design tasks
(3) invite a few potential or target users
(4) ask users to carry out these tasks (either with or without thinking aloud)
(5) recording all actions and verbal activities
(6) simple analysis of the results
Without experimental design - pos/neg
+: quick and natural
-: do not know the exact causal relationship among variables with enough data support, cannot formally use results in publications, test reports, etc.
With experimental design - pos/neg
+: relatively clean causal relationship, formal usage of the data
-: designing and running experiment is time consuming
Independent variables
variables manipulated by the researcher or usability experts
Dependent variables
variables observed by the researcher or usability experts
One of the keys of a successful design
Controlling confounding variables
Within subjects design
levels of a certain variable experienced by each subject
between-subjects design
levels of certain variables only experienced by some of subjects/users
Why number of subjects is an important issue
determines validity and how strong your conclusion is
Equation for number of subjects
NX where N is a natural number and X is combinations of all between subjects variables
OR can be calculated by estimated variability of data
OC Curves
Operating Characteristic Curve
x axis of OC curve
d = abs(mu - mu0)/sigma where mu is the average of the data, mu0 is the standard
Statistical power
beta; should be greater than or equal to .6
standard deviation
sqrt(sum of (xi-avg(x))^2/(n-1))
Power
1- beta; should be greater than or equal to 7
