Decision Making

1
Decision Making
2
Overview

• Signal Detection Theory
• Information Processing Model of Decision Making
• Decision support system (DSS)
• Decision Making under stress

3
Signal Detection Theory (John A. Swets and David
M. Green, 1966)

• Theory is applicable in any situation in which
  there are two states of the world.
• Signals must be detected by the human operation.

Signal (present)
Yes, I detect the signal.
No, I do not.
Noise (absent)
Stimulus
Processing
Response
4
SDT and Decision Making
Making decisions under conditions of uncertainty,
such as how we would perceive distances in foggy
conditions.
5
SDT
Response
Yes No

Present (Signal) Absent (Noise)
Stimulus
Signal model (Green Swets, 1966)
6
Signal probability

• P(hit) 5 hits 0.25
• 15 misses
• P(hit) P(miss) 1
• P(false alarm) P(correct rejection) 1

7
SDT
Response
Yes No

Present (Signal) Absent (Noise)
Stimulus
Signal model (Green Swets, 1966)
8
Medical Scenario
Response
Yes No

Present (Signal) Absent (Noise)
Stimulus
9
Eyewitness scenario
Response
Yes No

Present (Signal) Absent (Noise)
Stimulus
10

• Information acquisition
• Acquiring more information increases the
  likelihood of either a hit or a correct
  rejection.
• Criterion/bias
• Two people have the same information and good
  training but resulting in miss or false alarm
  differently.

11
Hit and Miss
Criterion response
Response Yes
Response No
Probability
Hit
Miss
0 5 10 15 20 25
Internal response
12
Correct rejection and false alarm
Criterion response
Response No
Response Yes
Probability
Correct reject
False alarm
Internal response
13
Hypothetical distributions
Criterion beta
Response Yes
Response No
Signal Present
Noise Absent
Correct rejection
Hit
Probability
False alarm
Miss
0 5 10 15 20 25
Internal response
14
Sensitivity d

• How hard or easy it is to detect that a target
• stimulus is present from background events.
• Ex 1 remember 30 words rather than 5 makes the
  discrimination harder
• Ex 2 Lack of training of a medical students for
  the task of detecting tumor patterns on an x-ray.
  (property of observer)
• Sensitivity relates to intelligent, experience
  and knowledge.

15
Sensitivity

• d' measures the separation between the means of
  the signal and noise distributions in units of
  the standard deviation of the noise distribution.
  
• Increasing the separation (stronger signal) .
• The signal may be missed because the detection
  process is poor at discriminating signals from
  noise.

No miss or false alarm
D
D
False alarm
Probability
Probability
Miss False alarm
16
Receiver Operating Characteristic - ROC
d1 Lot of overlap
d3 Not much overlap
Probability
Probability
More hits no false alarm.
d3
d2
d1
Hits
d0
False alarms

• Signal strength less overlap

17

• http//wise.cgu.edu/sdtmod/measures4.asp

18
Bias

• Bias is the extent to which one response is more
  probable than another. A receiver may be more
  likely to respond that a stimulus is present or
  more likely to respond that a stimulus is not
  present.

• If there is a penalty for either false alarms or
  misses,
• this may influence bias.
• If Tusinami siren alarms too often and nothing
• happen, people less likely respond for a
  conservative
• bias.

19

• Criterion is a measure of the
• willingness of a respondent to
• say 'Signal Present' in an
• ambiguous situation.

Criterion level
Far left More hit rate More false alarm Say
Yes all times.
Hits
d0
False alarms
20
Signal detection theory game

• Signal probability
• The probability that a signal
• will be presented to the subject.
• Its value must be between 0
• and 1.
• Signal duration
• The signal will be shown to the
• subject in ms.
• (1000ms 1 second).
• Signal strength
• The size in pixels of the signal
• (dot) presented to the subject
• Number of Trials
• how many trials will be
• presented in the experiment.

http//www.mind.ilstu.edu/curriculum/modOverview.p
hp?modGUI224
21
Features and Classes of Decision Making

• Uncertainty
• Familiarity and Expertise
• Time pressure
• Classes of Decision Making

22
Uncertainty

• Risk Uncertain outcomes are unpleasant or
  costly one
• Example of low risk
• - Decision to purchase one of vehicle.
• Example of high risk
• - Decision to proceed with a flight in
  uncertain
• weather.

23
Familiarity and Expertise

• Cue are unfamiliar and little consideration
• Experts make decisions more rapidly and with less
  effort than novices this does not imply that
  they are always more accurate.

24
Time

• Time response under pressure is less than no
  pressure.

People trend to take time to make a decision in
low risk than high risk when they do not have
pressure. When they have time pressure, response
time in high and low risk are nearly the same.
25
Classes of Decision Making
Optimal framework (gold standard) Optimal beta
Decision routines
Making decision in real environment
26
What is good decision making?

• Expected values of gold standard
• Minimize the maximum loss
• Good outcomes
• Decision to launch Challenger
• 3. Expertise

27
An Information Processing Model of Decision Making
Attention Resource Effort
Meta-cognition
Long term Memory
Long term Working Memory
Confirmation
Options
Risks(Values)
Selective attention Clue filtering
Working Memory and Cognition DIAGNOSIS H1 H2
Situation Awareness
Cues
CHOICE Response selection
Response Execution
Perception
Senses
Environment
28
Cue seeking and Hypothesis Formation

• Cue diagnosticity
• High see rain drops falling (100 diagnostic
  cue).
• Low Dark clouds on the horizon.
• Undiagnostic - 50 chance of showers
• Cue reliability or creditability
• Reliability from 0 1
• An eyewitness to a crime is a liar.
• The physical features
• Cue that make it salient

29
Representationof the process of information
Hypothesis Belief
H1 H2
H1

• Diagnosticity
• (D)

0-1
Info Value RxD
H2
(2) Reliability0-1 (R)
(3) Physical features (salience)
Cues
H2
H1
Truth (True state of the world)
30
Attention and Cue Integration

• Information cue are missing
• Decision maker does not have all information
  at hand to make an accurate diagnosis.

The maintenance makes the sensor functioning
wrong.
31

• Cues are numerous Information overload
• More information leads to more time-consuming
  filtering at the expense of decision quality.

• Cue in three miles island
• Light turn on
• Sound alarm
• Thousand gauges and switches

32
Choice of Action

• Certain choice
• Rank order the important of each attribute
• Assess the value of each object on each attribute
• Assess the sum of the products of (value x
  importance)
• Chose to purchase the object with the highest
  sum.
• Object A 2x1 3x4 14
• Object B 3x1 1x4 7

33
Training Decision Making Practice and Debiasing

• Feedback is critical for any form of learning or
  skill acquisition but
• Feedback is often ambiguous
• Feedback is often delayed decision maker may
  have forgotten the processes and strategies used
  to make decision.
• Feedback is processed selectively bias.

34
Domain of Decision Making
35
Decision support system (DSS)

• Decision support systems are a class of
  computer-based information systems including
  knowledge base.
• Model driven - statistical, optimization or
  simulation model.
• Knowledge driven expertise stored as facts,
  rules, procedures, or in similar structures.
• Communication driven more than one person
  working on a shared task.

36
Knowledge Driven
37
Expert System

• Contains some of the subject-specific
• knowledge, and contains the knowledge and
• analytical skills of one or more human experts.
• Inference Engine
• Forward chaining rule
• Backward chaining rule

38
Military Decision Making process
This process consists of six broad steps, which
are the foundation of decision making1.
Recognize and define the problem. 2. Gather
facts and make assumptions. 3. Develop possible
solutions. 4. Analyze each solution. 5. Compare
the outcome of each solution. 6. Select the best
solution available.
http//www.globalsecurity.org/military/library/pol
icy/army/fm/17-95/c2.htm
39
Decision Making Template (location base)
Visual Display
40
Decision tree analysis
41
Communication Driven
42
Challenger Space Shuttle disaster

• Engineering problems design of the joint seal.
• Decision making procedures relates to launch
  shuttle. Fail as a design criterion

43
Flawed in Decision Making

• A group decision support system (GDSS) did exist
  between NASA and Morton-Thiokol.
• An agenda was never defined, hence NASA were
  surprised by the Thiokol O-ring presentation and
  appalled by their decision not to launch.
• Conflict management was avoided by NASAs
  domination of the meeting, and hence conflict was
  not satisfactorily resolved.
• The GDSS put safety last and operational goals
  first. Note that shuttle crew were not
  represented at the meeting, although they had the
  most to lose.

44
Stress and Decision Making

• Stressor includes
• Influence noise, vibration, heat, dim, lighting,
  and high acceleration.
• Psychological factors anxiety, fatigue,
  frustration, and anger.

45
A Representation of Stress Effects
Stressor
Physiological arousal
Direct (e.g. vibration)
Indirect
Direct (e.g. lighting, noise)
Information Processing
Performance
Input

• Vibration will reduce the quality of visual
  input.
• Time stress can curtail the amount of
• information.

46
Strategic Control of Stress
Appraisal Difference between required and
available Resources to cope
Stressors
CHOICE
Accuracy
1 Resource recruitment Accelerated effort
mobilization
2 Remove stressor
4 Do nothing
3 Change the Goals of the task
Performance
Physiology effort cost
47
1. Recruitment of More Resources

• Try harder - mobilize more resources
• Strategy can be adaptive but it has risks.
• Three Mile Island incident was to shut down an
  automated device that had in fact been properly
  doing its job.

48
2. Remove Stressor

• Eliminate the source of stress
• Postpone a deadline to remove time pressure

49
3. Change the Goals of the Task

• Select a simpler or less effortful strategy
• Chose to process fewer cues in decision tasks
  carried out under time pressure.

50
4. Do Nothing!

• Allowing the stress effects to influence
  performance in a more predictable way.

51
Clean up the stress

• Design Solutions
• Reducing unnecessary information(visual clutter)
• Degrading time pressure by using graphic rather
  than a digital display (Schwartz, 1985)
• Instruction for emergency should be phrased in
  direct statements of what action to take rather
  than as statements of what not to do.

52

• 2. Training
• Training for emergency procedure
• Retrieve the habits from long-term memory when
  stress imposes that bias.

53

• Stress inoculation training (SIT)
• SIT is a form of cognitive restructuring as
  it is a method of changing an individuals
  thinking patterns about themselves and their
  lives. The aim is to change their emotional
  responses and their behaviour ideally before the
  individual becomes very anxious or depressed as a
  result of stress.
• Conceptualization
• Skill acquisition
• Application

54
Summary

• Decision Making in human factor is based on the
  signal detection theory hit, miss, false alarm
  and correct rejection.
• Key issues are cue, bias, probability,
  uncertainty, and outcome from decision making.