Engineering Psychology PSY 378S

1
Engineering PsychologyPSY 378S

• University of Toronto
• Spring 2005
• L5 Vigilance

2
Outline

• Vigilance
• Definition, typical tasks
• Level versus decrement
• Vigilance situations
• Sensitivity
• Phenomena
• Theories
• Bias
• Phenomena
• Theories
• Practical recommendations

3
Vigilance

• Vigilance task
• Operator is required to detect signals over a
  long period of time
• Signals are typically
• I) infrequent
• II) intermittent
• III) unpredictable
• Signals may also be difficult to detect
• low volume, not very different from background
  etc.
• What is vigilance?
• Operators ability to detect signals over time
  period

4
Level vs. Decrement

• Vigilance Level
• The state of the nervous system presumed to
  mediate performance in vigilance tasks
• High vigilance a state of readiness
• Low vigilance inattentive, fatigued
• Performance at one time
• Vigilance Decrement
• A deterioration in the ability of the observer to
  remain vigilant for critical signals over time

5
Two Vigilance Situations

• 1) Inspection situation
• Events occur at regular intervals
• Event frequency of targets / of events
• Time not taken into account
• e.g., circuit-board inspector
• Some boards are bad--targets
• Each board is an event

6
Two Vigilance Situations

• 2) Free-response situation
• Target event may occur at any time
• Non-events not defined
• Event frequency of targets / unit time (e.g.,
  minute)
• E.g., monitoring for incoming threats, missiles

7
Two Further Distinctions

• 1) Sim-Suc
• In successive situation, observer must remember
  target
• In simultaneous situation, target example shown
• 2) Sensory-Cognitive
• In sensory situation, signals represent changes
  in visual or auditory intensity
• In cognitive situation, signals are changes in
  letters/numbers

8
Measuring Vigilance Performance

• What if use number of targets detected as measure
  of performance?
• Need to specify performance in terms of P(Hit)
  and P(FA)
• Signal Detection Theory (SDT) problem
• Obtain sensitivity and bias measure over the
  course of the watch

9
Factors Affecting Sensitivity

• Target Uncertainty (affects level but not
  decrement)

Hi
Lo Uncertainty
Sensitivity
Hi Uncertainty
Lo
Time
10
Factors Affecting Sensitivity

• Signal Strength

Hi
Strong Signal
Sensitivity
Weak Signal
Lo
Time
11
Factors Affecting Sensitivity

• Event Rate (e.g., speed up conveyer belt)

Hi
Slow
Sensitivity
Fast
Lo
Time
12
Factors Affecting Sensitivity

• Simultaneous tasks produce greater sensitivity,
  no decrement

Hi
Sim
Sensitivity
Suc
Lo
Time
13
Factors Affecting Sensitivity

• Decrement eliminated if task automatized

Hi
Auto
Sensitivity
Controlled
Lo
Time
14
Factors Affecting Sensitivity

• Sensitivity increment with cognitive stimuli

Hi
Cognitive
Sensitivity
Sensory
Lo
Time
15
Theories of Vigilance

• Sensitivity Loss Fatigue (Broadbent, 1971)
• Continued vigilance causes fatigue, attentional
  lapses
• Increases variability of SN distributions
• Sensitivity Loss Sustained Demand (Parasuraman,
  1979)
• Sustained attentional demand, or working memory
  load
• Leads to demand on processing resourcesmay also
  be fatiguing
• Increases variability of SN distributions

16
Bias Increment

• Observers become more conservative over time

Conservative
Bias
Liberal
Time
17
Factors Affecting Bias

• Increases in Target Probability makes observer
  more liberal (note no effect on decrement)

Conservative
Lo Probability
Bias
Hi Probability
Liberal
Time
18
Factors Affecting Bias

• Increases in Payoffs makes observer more liberal
  (note no effect on decrement)

Conservative
Penalize FAs
Bias
Reward Hits
Liberal
Time
19
Theories of Vigilance

• Bias Increment Expectancy Theory (Baker, 1961)
• Upward adjustment of response criterion in
  response to reduction in perceived frequency of
  events
• Ps(S) subjective probability of signal
• If signal missed Ps(S) goes down
• Vicious circle
• Effects of signal probability and payoffs easy to
  account for with this theory

20
Theories of Vigilance

• Bias Increment Arousal Theory (Welford, 1968)
• Decreased arousal reduces both variability and
  mean of N and SN distributions
• Constant sensitivity, effective criterion shift
• Predicts lower P(H) and P(FA)

21
Theories of Vigilance

• Arousal Theory (Welford, 1968)
• Problem Not consistent with results from drug
  studies
• Drugs that modify arousal affect sensitivity but
  not bias
• Factors such as event rate that should affect
  arousal affect sensitivity but not bias

22
Putting it Together

• Some combination of theories may be useful
• Same physiological mechanismincreased fatigue,
  decreased arousalmay account for sensitivity
  decrement and bias increment
• Expectancy theory may kick in once
  fatigue/lowered arousal leads to decreased Ps(S)
• Expectancy theory is necessary also to account
  for prob/payoff effects

23
Vigilance in the Real World

• Signal frequencies in highly reliable systems may
  be much lower than those studied in the lab
• Signals are typically more complex
• Effects of costs and payoffs may be higher
• Motivation level of operator may differ

24
Improving Vigilance Performance

• Increasing Sensitivity
• Show Target examples (reduce memory load)
• Show standard representation of a target
  (picture of defect rather than normal)
• Reduce or vary event rate
• Training
• high amounts of practice with a particular target
  reduces memory load
• develops automatic processing (pop-out)
• Reduce number of potential targets
• each inspector looks for a particular target
• reduces memory load

25
Increasing Sensitivity (contd)

• Signal Enhancement
• Take advantage of naturally attention-grabbing
  stimuli, e.g., perception of motion
• blinking rapid exchange of visual image of
  known good prototype and item to be inspected
  (Drury)
• defect blinks on and off
• Compress motion of long term events
• segregates signal (coherent) from noise (random)
• radar display Scanlan (1975)
• weather satellite images
• Use redundant coding (multiple sensory
  modalities, e.g, both visual and auditory)

26
Adjusting Bias

• Instructions, Training to adjust b to bopt
• Prior probabilities P(S), P(N)
• Payoffs C(M), C(FA), V(H), V(CR)
• Knowledge of Results (KR)
• What is true P(S)?
• False Signals
• Should lead to higher Ps(S)
• Use tagging only applicable in some cases
• Confidence Levels
• Rest periods, phone calls, coffee breaks, noise,
  biofeedback

27
Summary

• Can isolate performance in vigilance task using
  SDT measures
• Theories developed to account for sensitivity
  decrement and bias increment
• Different recommendations for each problem
  described