Transcript Slide 1

Engineering Psychology and
Human Performance
Human Information Processing
• Knowledge of physical performance
(i.e. speed and accuracy) is important
in understanding a task (or the effects
upon the task from an ergonomics
intervention program).
• But such knowledge is not sufficient to
understand WHY performance changed
from a human operator perspective.
A MODEL for Human
Information Processing
1.
2.
3.
4.
5.
6.
Sensory Processing
Perception
Memory and Cognition
Response Selection and Execution
Feedback
Attention
Sensory Processing
• Information and events in the environment
must gain access to the brain.
• Visual, Auditory, Tactile, Smell
• All sensory systems have an associated
short-term sensory store
– Temporary mechanism for prolonging the
representation of the raw stimulus (length
depends on the type of stimulus)
– Can be confounded if the operator is distracted
Perception
• Raw sensory data relayed to the brain
must be interpreted (given meaning)
through a stage of perception.
– Decode the meaning of the signal
• Generally proceeds rapidly and
automatically
• Driven by sensory input (rapid; bottom-up
processing) or long-term memory (topdown processing)
Memory and Cognition
• Cognitive operations generally require more
“processing” time.
• Rehearsal, reasoning or image transformation is
carried out by working memory.
• “Plans”; “Diagnose” ~ these are vulnerable to
disruption when attentional resources are
diverted to other mental activities.
• Material that is rehearsed in working memory
can get access to long-term memory – learning
takes place.
Response Selection and Execution
• The understanding of a situation, achieved
through perception and augmented by
cognitive transformations will trigger the
selection of a response.
• The selection of a response must precede
the execution (requiring motor effort –
which has it’s own qualities for success).
Feedback
• Actions are directly sensed by the
operator OR influence the system
within which the operator functions.
– Flow of information can be initiated at any
point and is continuous.
Attention
1. Selective Attention
2. Focused Attention
3. Divided Attention
A model of human information
processing stages
Signal Detection
Overview:
• The process of detection may involve
many states of categorization:
– Operator to decide whether a “signal” is
present.
– Operator to attend to several signals
Signal Detection Paradigm ~
The Four Outcomes of Signal Detection
Theory
Setting the Response Criterion
Hypothetical distributions underlying signal detection theory
Consider the circumstances!
• Radiologist examining X-Rays of referred cancer
patient.
– More likely to say “yes” but increases the chances of a
“false alarm” - CONSERVATIVE
Consider the circumstances!
Plant operator warned about unnecessary shutdowns
and the financial implications
– More likely to say “no” but increases the chances of a
“missed event” - RISKY
• For what ever reasons:
– When signal and noise have similar
characteristics
– Operator deficiencies (training, fatigue,
stress)
• Chances (probability) of false alarms or
misses increases!
MISSES
FALSE ALARMS
The Probability of Signal Detection Theory
• Signal detection theory is able to
prescribe where the optimum beta (β)
should fall:
– The likelihood of observing a signal
– The COSTS/BENEFITS (payoffs) of the four
possible outcomes
• In almost all circumstances, signal and
noise will overlap.
Attention in Perception and Display Space
• The limitations of human attention represent
one of the most formidable bottlenecks in
human information processing.
– I failed to notice the words of a speaker because I
was looking out the window
– We had so many tasks to perform that some were
omitted
SELECTIVE ATTENTION
• Begins with the EYE
and VISUAL
SAMPLING.
– Seek for
information/search
for a target
– We learn a lot by
assessing visual
scanning behaviour
SELECTIVE ATTENTION
• The FOVEA has about
a 2 degrees of visual
angle.
• To keep objects in the
foveal vision ~
– Pursuit Movement ~
eye follows a target
moving across the
visual field – “smooth
motion”
– Saccadic Movement –
jump from one
stationary point to
another – “jerky
movements”
SELECTIVE ATTENTION
• Supervisory Control Sampling
– In the aircraft cockpit or the process
control console, many information sources
have to be sampled periodically.
– Often divide the environmental stimulus
into CHANNELS along which critical
EVENTS may periodically occur.
SELECTIVE ATTENTION
• Environmental Sampling is guided by
the expected cost that results when an
event is missed.
• The probability of missing an event is
directly related to the event frequency
and uncertainty.
SELECTIVE ATTENTION
Visual Scanning
1. Mental Model Guides
2. Adjustment to Event Rate
3. Sampling Affected by Arrangement
4. Memory Imperfect; Sampling Imperfect
5. Preview Helps
6. Processing Strategies – Cognitive Tunneling
VISUAL SCANNING
Mental Model Guides
•
The mental model consists of a set of expectancies
of how frequently and when events will occur.
•
Correlation of events amongst channels.
•
As expertise develops, mental models become
refined as do the sampling strategies.
VISUAL SCANNING
Adjustment to Event Rate
•
People learn to sample channels with higher event
rates more frequently (and lower event rates less
frequently)
VISUAL SCANNING
Sampling Affected by Arrangement
•
Operators like to scan horizontally rather than
diagonally.
•
It is proposed that operators use simplifying rules
and HEURISTICS based on channel arrangement to
decrease attentional demands – this leads to
systematic biases in performance.
•
Arrangements matter
* Heuristics – creating a model as a working
hypothesis to a goal or solution
VISUAL SCANNING
Memory Imperfect; Sampling Imperfect
•
•
People tend to sample information sources more
often than they would need to if their short-term
memory was better. Two consequences:
•
“Oversampling” occurs of channels with low event rates
•
People forget to sample channels with low event rates
Suggests the use of “sampling reminders”
VISUAL SCANNING
Preview Helps
•
When people are given a preview of scheduled events that are
likely to occur in the future, sampling a channel switching
become somewhat more optimal.
•
Now an “external model” is featured – somewhat else guiding
attention.
•
If the number of channels increases too much, people fail to
take advantage of the preview – because of the heavy
memory load.
VISUAL SCANNING
Processing Strategies – Cognitive Tunneling
•
Scanning behaviours may reflect the operator’s
mental model of the environment. This includes
biases in the operator’s strategies.
•
Fixation on feedback on an event can be a
substantial waste of visual attention.
PARALLEL PROCESSING AND
DIVIDED ATTENTION
• Divided attention and parallel
processing is often a good thing (air
traffic controllers).
• However, it is sometimes very difficult
to narrow the focus of attention when
needed and shut out unwanted inputs.
• This failure occurs when divided
attention becomes mandatory rather
than optional.
Preattentive Processing and Perceptual
Organization
• Preattentive phase is carried out
automatically and organizes the visual
world into objects and groups of
objects
Preattentive Processing and Perceptual
Organization
• Grouping together of
similar items on a
display.
• Knowledge of where
one is on the display
(proximity, similarity,
common fate, good
continuation, closure)
will allow an accurate
guess of where others
are located.
Preattentive Processing and Perceptual
Organization
• Because all items of an organized
display must be processed together to
reveal organization, such parallel
processing is sometimes called
GLOBAL or HOLISTIC processing, in
contrast to LOCAL processing.
Global vs Local
• Global Display
Organization (top)
allows for faster
identification of
Local abnormality.
• Top display likely
has reduced:
– Attentional demands
– Risk for missed
event
The Proximity Compatibility Principle
Display Proximity – how close together
two display components are in spatial
terms
Processing Proximity – the extent to
which two information sources are
used within the same task (compare or
integrate).
The Proximity Compatibility Principle
• Close proximity will increase the possibility of
parallel processing by moving both dimensions into
foveal vision. It is even more likely when objects are
integrated a dimensions of a single object.
• Emergent features (colour coding,symmetry) can
help in this respect.
Visualization of Complex Dynamic Information
•Visualization
Perception
Decision /
response
selection
Working
memory
Long-term
memory
Memory
•i.e. “the use of computer-supported, interactive, visual
representations of abstract data to amplify cognition“
(Card et al., 1999)
•supports cognition by
– Relief of resources for memory and information
processing
– Facilitation of information retrieval
– Recognition of hidden patterns, tendencies etc.
– Reasoning at a perceptual level („direct
perception“)
– Improving visual attention during monitoring /
supervisory tasks
Goal: Reduction of effort for information access and processing
Proximity Compatibility Principle (Wickens & Carswell, 1995):
Object Integration and Emergent Features
•
Idea: Utilisation of human skills for pattern-recognition
by means of graphical transformation of complex data
•“Normal States” are visualized by
graphical attributes like
– Symmetry,
– Regular Alignment or
– Parallelism
•
(Images: Wickens, 1992)
•Deviations from Normal States generate
– Asymmetry,
– Irregular Alignment,
– Skewness.
Comparison of Mental Transformations:
Conventional vs. “Direct Perception“ Polar Displays
Actual value
Rule
ID-Crits (@ memory)
50 > ISR
DST
…
ALT
164 ALT
ESM Down Beat EM
ISR
 ALTciv
 EMciv/FF
= 30 NM
ALTciv = [5000;FL 250]
EMciv/FF = {…}
n * Integration: actual value vs. setpoint
Conventional Display
Polar Display
Air track
uncritical
Reference area
= situation knowledge
Air track
critical
Emergent
features
Result
true
false
false
SU
Integration
of attributes
PCP: Display Proximity ~ Task Proximity & Integration of
Knowledge
Position and flight path have to be extracted by
means of several flight control displays.
Mental transformation to outside view is required.
Physical proximity of
altitude and vertical speed
Proximity of frequently used displays
Position and flight path have
to be extracted by means of
several flight control displays.
Basic-T
conventional HUD
Direct perception of
position and flight path
Integrated Display
Air Picture of a Combined Air Operations
Centre (CAOC)
Visualization:
 Tracks & related events
and actions
 A-priori information:
e.g. airways, corridors,
(restricted) areas etc.
 Interactive ENC-chart
S-57 / VMAP L1
 Event- and Threat-triggered
Polar Displays
Usage by toolbox using
well-known methaphers:
 Scrolling
 Cartographic zooming
 Track search
 Formation editing
 Identification / Classification
 Display settings
predefined filters, styles, etc.
Graphical Representation of Information
and Integration of A-priori Knowledge for AAW Situation Analysis
Track Detail Display
Tactical Situation
Display
Polar Displays
Toolbox
User Support
Display
Secondary Display for Detailed Information
Relief of working memory:
 Parameter ranges /
tendencies
 Predefined tolerances for
non-critical values
 State classification (IFF/ESM)
 History plots: Values / states
along time / distance
 Information on demand
Utilize pattern recognition:
 Heterogeneous values cause
emergent line characteristics
 Parameter constellations
(Alt/Spd)
 Graphical representation of
user support system “explains”
results
Ecological Interface Design (EID)
Vicente & Rasmussen (1992)
•Approach:
•
System errors result from violations of
system operation constraints…
 Unexpected errors can be managed,
– if the operator possesses a valid mental model of the system AND
– if the system constraints are visualized referring to environmental factors influencing
system behavior.
•Constraints of system operations
–
–
–
–
–
Physical attributes
temperature, radiation, illumination etc.
System functionalities
different operational phases
Structural aspects of the work system flexible vs. determinated work system
Organizational structure
hierarchies, safety culture
Working climate
restrictions through rules, competitors
•Fundamental concepts
– Abstraction hierarchies
– Visualization of constraints: Invariants, system structure and interrelations
– Taxonomy of human information processing (Rasmussen, 1983)
EID: Abstraction Hierarchies
•Abstraction hierarchies offer different views which differ regarding level
of detail and resulting information elements.
•Hierarchy levels:
– Functional purpose:
Purpose which should be accomplished by the system
– Abstract function:
Defined process / system structure
regarding mass, energy, information or information flow.
– Generalized function:
Basis functions of system components.
– Physical function:
Functional characteristics and connections of system components.
– Physical form:
Structure and position of system components.
•Each level represent another class of constraints for system operation.
Influencing Factors in Adaptive Systems
•
Situation
–
–
dynamic:
static:
Objects / prioritized tasks, system condition
Knowledge (Scenario, Situation, Competences)
 needs & possibilities of actions
(“What should be done?”)
•

advice

directed assistance

dynamic regulation
of input load
•
•
Operator activity
–
–
Input sequences ( supervisory control)
Focus of attention
•
 actual operator activity
(“Which task is in focus?”)
•
•
Operator Functional State (OFS)
–
•
Estimation of resource capacity
and operator capability (normative / individual)
 need for user support
(“Overload State?”)
Eye-based Assessment of Operator Functional State in
Adaptive Systems
250
• Point of interest
• Mental workload, vigilance
60
A1
eye blink
frequency
Faktor-bezogener Mittelwert
– Point of gaze, scan paths
– Oculomotoric parameters
Faktor-bezogener Mittelwert
•Remote / Head-based Eye Tracking
A2
A1  s A .1 40
A2  s A .2
fixation
frequency
A1
200
A2
A1  s A .1
A2  s A .2
150
20
•Assessment techniques
100
WPT
– Corneal-retinal potential (EOG)
– Cornea reflex infrared ocularography
– Marker-based tracking
WPT
60
30
FA
FA
20
FA
sk A A
40
FA
sk A A
2

2

20
10
s kA  1
s kA  1
0
5
10
15
WPT
 Indicators for
visual input load
0
5
10
WPT
15
ATTENTION IN THE AUDITORY MODALITY
• Auditory Sense can take input from any
direction and thus there is no analog to
visual scanning as an index of selective
attention.
• Most auditory input is transient
– Thus preattentive characteristics (organize
mentally) is more difficult after the event
has passed (sound stops)
Auditory Divided Attention
• A general theory states that unattended
channel of auditory input remains in
preattentive short-term auditory store
for about 3-6 seconds.
• The contents of this store can be
examined if a conscious switch of
attention is made.
Auditory Divided Attention
• Even unattended auditory channels
may make content with long-term
memory.
– Car horn honking
– Your name being spoken
– These are not meaningless blobs of noise!
• It is possible to think of an “auditory object’
as a sound with several dimensions – thus
parallel processing can occur.
• Auditory warning alerts have been designed
to capitalize on our parallel processing
ability using redundant dimensions:
– Pitch
– Timber
– Interruption Rate
– All in various combinations