Chapter 1. Introduction

1
Chapter 1. Introduction

• ENGINEERING PSYCHOLOGY AND HUMAN FACTORS
• Human factors designing machines to accommodate
  the limits of the human user
• fundamental goals -- reduce error, increase
  productivity, and enhance safety and comfort when
  the human interacts with a computer
• Engineering Psychology a psychological
  perspective (brains information-processing
  capabilities) to the problems of system design
• not simply to compare two possible designs for a
  piece of equipment, but to specify the capacities
  and limitations of the human data from which the
  choice of a better design should be directly
  deducible
• The process of human factors engineering
• Task Analysis proceeded hierarchically
• Physical actions and goals
• Cognitive mental operations
• Choice of research method (table 1.1 in pp.7)
• Field studies, Accident and incident report,
  Surveys
• Task simulations
• Laboratory experiments
• Literature and handbooks
• Models

2

• Performance measurement
• Measures of speed or time
• Measures of accuracy or error
• Measures of workload or capacity demands
• Measures of preference
• A MODEL OF HUMAN INFORMATION PROCESSING
• Sensory Processing
• STSS (short term sensory store) iconic (0.5
  sec) to echoic memory (2 4 sec)
• Perception
• Perceptual processing decode the meaning from
  the raw sensory data
• proceeds automatically and rapidly
• driven both by sensory input (bottom-up
  processing) and by inputs from LTM about what
  events are expected (top-down processing)
• Cognitive process infer
• Cognition and Memory
• generally require greater time, mental effort, or
  attention
• cognitive operations such as rehearsal,
  reasoning, or image transformation are carried
  out by using what is called working memory, a
  vulnerable, temporary store of activated info.
• WM plans, diagnose, resource limited
• LTM less vulnerable, more permanent

3

• Response Selection and Execution
• Feedback
• the flow of information can be initiated at any
  point
• the flow of information is continuous action
  causes perception
• Attention
• supply of mental resources

Attention Resources
Long-term Memory
Selection
Working Memory Cognition
Sensory Processing STSS
Perception
Response Selection
Response Execution
System Environment (Feedback)
4
Chapter 2. SDT, IT

• SIGNAL DETECTION THEORY
• The signal detection paradigm
• Signal Detection Theory (SDT) two stages of
  information processing
• sensory evidence is aggregated concerning the
  presence or absence of the signal
• a decision is made about whether this evidence
  indicates a signal or not
• P(H) P(M) 1 and P(FA) P(CR) 1
• Setting the response criterion Optimality in SDT
• response bias -- liberal or risky (conservative)
• liberal prone to saying yes -- detecting most
  of the signals but making many false alarms
• conservative saying no most times and making
  few false alarms but missing many signals
• The position of the decision criterion (Xc)
• beta (?)
• Positively related with Xc
• The ratio of neural activity produced by signal
  and noise at Xc

5

• shifting Xc to the right ? ? gt 1 (conservative) ?
  fewer yes responses fewer hits and fewer false
  alarms
• shifting Xc to the left ? ? lt 1 (liberal) ? more
  yes responses more hits and more false alarms
• beta and Xc define the response bias or response
  criterion
• truly neutral criterion setting ? P(H) P(CR)
  and P(M) P(FA)
• Signal Probability
• Equal probabilities, no costs, no benefits --
  optimal performance (minimize errors) at ?1
• If a signal is more likely, the criterion should
  be lowered
• If signal probability is reduced, beta should be
  adjusted conservatively
• ?opt -- defines where beta should be set,
  depending on the ratio of the prob which noise
  and signals occur in the environment
• ? -- determined by the observer and must be
  derived from empirical data
• Payoffs
• optimal is no longer defined in terms of
  minimizing errors but is now maximizing (or
  minimizing) the total expected financial gains
  (losses)
• Signals never to be missed reward for hits and
  penalties for misses a low set of beta

6

• Human Performance in Setting Beta
• Humans do adjust beta as dictated by changes in
  payoffs and probabilities.
• Sluggish beta
• Less risky when the ideal beta is high
• Less conservative when the ideal beta is low
• More pronounced when beta is manipulated by
  probabilities than by payoffs
• reflection of the operators need to respond
  creatively by introducing the rare response
  more often than is optimal
• operator misperceive probabilistic data
  overestimate rare events and underestimate
  frequent events
• Bisseret (1981) compared the performance of
  experts and trainees
• experts used lower beta settings, being more
  willing to call for a correction training
  should be devoted to the issue of criterion
  placement
• Sensitivity
• Separation of noise and signal distributions
  along the X axis
• Measure d units of their standard deviations
  (0.5 to 2.0)
• If the separation is large high sensitivity
• If the separation is small low sensitivity

7

• THE ROC CURVE
• Theoretical Representation
• receiver operating characteristic (ROC) curve is
  useful for obtaining an understanding of the
  joint effects of sensitivity and response bias on
  the data from a signal detection analysis
• P(H) and P(FA) are critical P(M) 1- P(H) and
  P(CR) 1 P(FA)
• the value of beta along the ROC curve is equal to
  the slope of a tangent drawn to the curve at that
  point this slope is equal to 1 along the
  negative diagonal
• at the negative diagonal line, P(H) 1- P(FA)
• points on the positive diagonal chance
  performance P(H) P(FA) ? the signal cannot be
  discriminated from the noise the signal and
  noise distributions are perfectly superimposed
• points in the lower right region worse than
  chance performance ? the subject says signal
  when no signal is perceived the subject is
  misinterpreting the task
• Empirical Data
• a set of discrete points instead continuous and
  smooth curves
• not fall precisely along a line of constant
  bowedness or a 45 degree slope slightly
  shallower
• the distributions are not precisely normal and of
  equal variance
• the tilting of the ROC curve -- difficulties for
  the use of d as a measure of sensitivity ? use
  da
• only one or two points in the ROC curve
• alternative measure for sensitivity the measure
  P(A) or the area under the ROC curve

8

• VIGILANCE
• To detect signals (intermittent, unpredictable
  and infrequent) over a relatively long period of
  time
• show lower vigilance level than desirable
• Vigilance decrement steep decline during the
  first half hour or so of the watch
• Vigilance Paradigms
• Free response paradigm
• a target event occurs at any time and nonevents
  are not defined power plant monitor
• Inspection paradigm
• events occur at fairly regular intervals (defects
  and normal items)
• Successive paradigm (target stimulus vs.
  presented stimulus)
• Simultaneous paradigm -- all the info needed to
  make the discrimination is present for each event
• Sensory paradigm signals represent changes in
  auditory or visual intensity
• Cognitive paradigm (symbolic or alphabetic
  stimuli)
• Measuring Vigilance Performance
• vigilance decrement can arise either as a result
  of a decrease in sensitivity or as a shift to a
  more conservative criterion
• Theories of vigilance
• Sensitivity Loss Fatigue and sustained Demand
  Theory
• Sustained attention task
• sustained demand theory sustained demand on
  mental resources causes sensitivity decrement

9

• Criterion Shifts Expectancy Theory
• Expectancy theory (Baker, 1961) vigilance
  decrement to an upward adjustment of the response
  criterion in response to a reduction in the
  perceived frequency of target events
• Arousal theory (Welford, 1968) decreasing
  arousal (increasing fatigue) decreases
  sensitivity, increases beta, and also lowers the
  subjective probability of a signal, again
  increasing beta.
• Techniques to Combat the Loss of Vigilance
• Increasing Sensitivity
• Show target examples (reduce memory load)
• Increase target salience
• Vary event rate
• Train observers
• Shift in Response Criterion
• Instructions
• Knowledge of results
• False signals
• Confidence level
• Other techniques

10

• INFORMATION THEORY (IT)
• The Quantification of Information
• human operator as a transmitter of info
• IT provides metrics to compare human performance
  across a wide number of different tasks
• Information reduction of uncertainty (bits)
• IT quantifies the amount of info conveyed by a
  statement, stimulus, or event
• The number of possible events that could occur, N
• The probabilities of those events
• The events sequential constraints, or the
  context in which they occur
• Number of Events
• minimum number of yes-no questions
• Hs log2 N where N is the number of equally
  likely alternatives
• Probability
• Hs log2 (1/Pi) where Pi is the prob. of
  occurrence of event i
• Have S Pilog2 (1/Pi)
• if the events are not equally likely, Have will
  be less than its value with equally probable
  events
• Sequential Constraints and Context
• info can be reduced by the context in which it
  appears
• sequential constraints reduce the info.

11

• Redundancy
• redundancy potential loss in info.
• redundancy (1 Have/Hmax) x 100
• Information Transmission of Discrete Signals
• not only how much info is presented to an
  operator but also how much is transmitted from
  stimulus to response (channel capacity), and how
  rapidly it is transmitted (bandwidth)
• ideal information transmitter HS HT HR
• stimulus-response matrix
• HT HS HR HSR (measure of dispersion or lack
  of organization within the matrix)
• Information transmission rate HT / RT ? measure
  of bandwidth (processing efficiency)
• ABSOLUTE JUDGMENT
• an observer assigns a stimulus into one of
  multiple categories along a sensory dimension
• Single Dimensions
• Experimental Results
• channel capacity ? 7 2 stimulus categories (2
  3 bits) fig. 2.12
• Affected by memory, learning and experience
• As the number of stimuli increases, sensitivity
  decreases.
• Edge Effect (Shiffrin Nosofsky, 1994) stimuli
  located in the middle of range of presented
  stimuli are generally identified with poorer
  accuracy than those at extremes lowered
  sensitivity

12

• Multidimensional Judgment
• Orthogonal Dimensions fig 2.13 (a)
• a continuous gain in total info. transmitted but
  a loss of info. transmitted per dimension
• Correlated Dimensions fig 2.13 (b)
• correlation 1 (as with the traffic light) than
  total Hs is just the Hs on any single dimension
• Information loss (HS HT) correlated lt
  orthogonal
• Information transmitted (HT) correlated gt any
  single dimension
• Orthogonal dimensions maximize HT, the efficiency
  of the channel
• Correlated dimensions minimize Hloss maximize
  the security of the channel
• A Theoretical Understanding
• General Recognition Theory (GRT) by Ashby and
  coworkers (1991, 1994, 1996)
• is based on SDT and generalize SDT to stimuli
  varying on more than one physical dimension (fig.
  2.15)

13
Chapter 3. Attention

• OVERVIEW
• Selective attention (cognitive tunneling)
• Pay attention to multiple things in sequence
  intentional but unwise choice
• Focused attention
• Pay attention to 1 thing tendency to be
  distracted by external environmental info.
• Divided attention
• Pay attention to multiple things at the same time
  simultaneously- limited ability to time sharing
  performance
• SELECTIVE ATTENTION
• Visual Sampling
• eye and visual sampling ? seek information and
  searches for targets
• Visual scanning behavior (attentional
  searchlight)
• Eye fixation system
• Fovea area perceives detail about 2of visual
  angle

14

• Pursuit movement the eye follows the target
• Saccadic movement
• discrete, jerky from one stationary in the visual
  field to next
• Sometimes superimposed on pursuit movement
• Saccade suppresses visual input
• Fixation display info. processed during
  fixation
• Location center of the fixation
• Useful field of view diameter around the
  central location which info. is extracted
• Dwell time how long the eye remains at that
  location
• Supervisory control context
• scans the display of a complex system under
  supervision
• allocates attention through visual fixations to
  various instruments
• the target is known
• Target search context
• scans a region of the visual world
• a targets location and existence is unknown

15

• Supervisory Control Sampling
• Optimality of Selective Attention
• stimulus environment -- channels and critical
  events
• Mental model guides sampling
• form a mental model of the statistical properties
  of events frequency and correlations
• Adjustment to event rate sluggish beta
• the sampling rate is not adjusted with event
  frequency
• Sampling affected by arrangement more likely to
  make horizontal scans than diagonal scans
  simplifying rules and heuristics
• Memory imperfect sampling imperfect sampling
  remainder
• Preview helps as the number of channels
  increases, fail to take advantage of preview
• Processing strategies cognitive tunneling
• failed system, delayed feedback
• Eye Movement in Target Search
• Environmental Expectancies
• fixate most on areas of containing the most
  information (Yarbus, 1967)
• a scan path over same picture dependent upon
  seeking info. (Yarbus, 1967)
• Display Factors and Salience
• neither consistent pattern of display scanning
  nor optimal scan pattern in search

16

• certain display factors to the allocation of
  visual attention
• salience and abrupt stimulus onset in the visual
  periphery may bias decision making
• presence of unique stimuli (singletons) slowed
  the detection of other targets
• physical location in the display the upper
  left, concentrated on the center (edge effect)
• dominated by conceptually or knowledge-driven
  scan strategies
• Display-Driven and Conceptually driven processing
• they commonly interact standardization of
  roadway and sign design
• positive guidance forecast the unexpected event
• Search Coverage and the Useful Field of View
• the highest acuity region of fovea an angel of
  no more than 2 degrees
• UFOV a circular area around the fixation point
  -- 1 to 4 degrees of visual angle
• size determined by the density of information
  the discriminability of the target
• aging restricted UFOV
• training enlarge UFOV, benefits are equal across
  age groups
• reduction in UFOV has serious implication such as
  driving
• UFOV is sensitive to task demand in the foveal
  region
• Fixation Dwells
• survey dwells short, used to establish the
  regions more likely to contain a target
• examination dwells used to provide a detailed
  examination of the region for an embedded target

17

• difficulty of information extraction
• low familiarity, low frequency, and out of
  context higher information content (longer
  dwells)
• expertise
• Visual Search Models
• How long to find a target? What is the
  probability in a given period of time?
• Drurys Model (1975, 1982)
• 1st stage target search stage
• the probability of locating a target increases
  with more search time (fig. 3.2) diminishing
  rate
• 2nd Stage decision stage
• uses the expectancy of flaws to set a decision
  criterion
• Variables affecting search speed
• the number of elements to be searched serial
  search (50 msec/item)
• exceptions to serial search one level along one
  salient dimension
• greater search efficiency for parallel than
  serial
• preattentive (requiring few attentional
  resources) for parallel and attentive for serial
• serial search the target is difficult to
  discriminate from distractors
• exceptions to serial search the presence of
  features rather than absent
• different discriminabilities of targets in the
  two situations

18

• dispersion of targets -- scanning distance and
  visual clutter trade off
• any of several different target types slower than
  only one
• exception a single common feature
• extensive training automaticity ? parallel
  search (consistent mapping not varied mapping)
• Structured Search
• Basics
• information that may help guide the search is
  available
• Application Menus
• target items are reached in the minimum average
  time
• linear visual search model frequently searched
  items positioned toward the top of the menu
• optimal number of items per menu is between three
  and ten (Lee and MacGregor, 1985)
• criterion-based model (Pierce, Sisson, and
  Parkinson, 1992)
• the effect of similarity in menu search
• Directing Attention
• advise an operator in advance where attention
  should be directed
• more accurate as the stimulus-onset asynchrony
  between the warning (cue) and the target
  increases SOA200ms more effective than
  SOA50ms
• peripheral cues (out of foveal area) more
  effective with short SOAs, a transient effect,
  stimulus-driven, automatic process
• central cues more effective with longer SOAs,
  long lasting, goal-directed, controlled
  interpretation

19

• PARALLEL PROCESSING AND DIVIDED ATTENTION
• Preattentive Processing and Perceptual
  Organization
• visual processing of a multiple-element world
  two main phases
• preattentive phase (STSS, automatic, grouping)
  and attentive phase (perception, selecting)
• Gestalt psychologists items to be
  preattentively grouped together on the display --
  proximity, similarity, common fate, good
  continuation, closure ? high redundancy
• all items of an organized display must be
  processed together to reveal the organization
  (parallel processing) -- global or holistic
  processing
• single object within the display -- local
  processing (fig 3.6) -- response conflict
  global precedence
• emergent features global property of a set of
  stimuli (displays) (fig 3.8)
• global processing tends to be preattentive and
  automatic -- reduce attentional demands
• Gestalt principles produce groupings or
  emergent features
• spatial proximity of different elements
  compatible with task demands
• Spatial Proximity
• Overlapping Views The Head-Up Display
• although spatial proximity will allow parallel
  processing, it certainly will not guarantee it
• Neisser and Becklin (1975) separation defined
  not only in terms of differences in visual or
  retinal location but also in terms of the nature
  of the perceived activity
• Wickens and Long (1995) an unexpected obstacle
  was detected more poorly with the HUD than with
  the head-down configuration -- HUD could improve
  control of position during landing, both in view
  and when the runway was obscured by clouds

20

• the HUD appears to facilitate parallel processing
  of scene and symbology when the pilot expects the
  stimulus and interferes when the stimulus is
  quite unexpected
• conformal nature of the symbology
• Visual Confusion, Conflict, and Focused Attention
• spatial proximity confusion
• spatial density little effect on visual search
  time
• Wickens and Andre (1990) -- critical variable in
  predicting performance is the degree of spatial
  separation of relevant item from irrelevant, not
  the spatial separation between the relevant items
  themselves
• Holahan, Culler, and Wilcox (1978) to locate
  and respond to a stop sign in a cluttered display
  is directly inhibited by the proximity of other
  irrelevant signs in the UFOV
• Eriksen and Eriksen (1974) -- perceptual
  competition a failure of focused attention
  caused by the competition (ex. UHP)
• response conflict (ex. FHF), redundant gain (ex.
  HHH)
• Object-Based Proximity
• different attributes of a single stimulus object
  at one spatial location
• concurrent processing of elements lying close
  together in space (space-based model of
  attention)
• concurrent processing occurs when elements lie
  within a single object (object-based model)
• Stroop task subject is asked to report the
  color of a series of stimuli as rapidly as
  possible
• multiple dimensions belonging to s single object
  are likely to be processed in parallel
• integral dimensions produce a cost for a
  filtering task and a benefit with redundant
  dimensions

21

• Applications of Object-Based Processing
• in cognitive psychology, an object has three
  features
• surrounding contours or connectedness between
  parts
• rigidity of motion of the part
• familiarity
• benefits of objects in two contexts
• conformal symbology mapping of display objects
  to real-world objects
• object displays multiple information sources
  are encoded as the stimulus dimensions of a
  single object
• Conformal Symbology and Augmented Reality
• conformal symbology helped the pilot divide
  attention between the display and the world
  beyond, align the display object to the real
  object, and reduce tracking error
• Object Displays
• parallel processing of object features to create
  multidimensional object displays
• The Proximity Compatibility Principle
• three ways in which multiple display channels can
  be integrated emergent features, spatial
  proximity, object integration
• proximity-compatibility principle whether
  different tasks are served differently by more or
  less integrated displays
• to the extent that information sources must be
  integrated, there will be a benefit to presenting
  those sources either close together, in an
  object-like format, or by configuring them to
  create emergent features

22

• to the extent that information sources must be
  treated separately, the benefit of the
  high-proximity object display will be reduced, if
  not sometimes reversed
• close proximity increases the possibility of
  parallel processing
• close proximity and objectness can create useful
  emergent feature that help information
  integration if they correspond to the key
  variables of the task
• emergent features can hurt performance if they
  are not mapped into the task
• response conflict can result if proximity
  combines variables that require focused attention
  
• Color Coding
• benefits
• rapid localization
• capitalizes on population stereotypes
• tie together spatially separated display elements
• redundancy in combination with shape, size, or
  location
• limitations
• failure of absolute judgment five or six
  colors, glare or low illumination (affected by
  ambient light)
• no ordered continuum brightness (saturation)
  rather than hue
• population stereotype poor design with a
  conflict meaning
• irrelevant color coding can be distracting --
  display-cognitive compatibility

23

• ATTENTION IN THE AUDITORY MODALITY
• omnidirectional no analog to visual scanning as
  an index of selective attention
• transient
• Auditory Divided Attention
• an unattended channel of auditory input remains
  in preattentive STAS (3 6 sec)
• attention switch
• on examined
• off LTM (preattentive) pertinent enough
  ?focused attention
• negative priming -- information presented in an
  unattended channel is temporarily inhibited for
  several seconds following presentation slower
• auditory object as a sound with several
  dimensions -- parallel processing
• Focusing Auditory Attention
• monaural and dichotic listening large benefits
  of dichotic over monaural listening in filtering
  out the unwanted channel
• cocktail party effect auditory selective
  attention (pitch, intensity, semantic properties)
• auditory attention can be directed by cueing
• Cross-Modality Attention
• parallel inputs across modalities
• redundancy gain -- speeds up processing
• dividing attention between modalities may be
  better than dividing attention within a modality
• visual dominance over auditory and proprioceptive