1. Signal Detection, Information Theory, and Absolute Judgment
Engineering Psychology and Human Performance
Park Young Ho
Dept. of Nuclear & Quantum Engineering
Korea Advanced Institute of Science and Technology
February
Let me start
Good afternoon everyone
Today I’d like to present “Signal detection, information theory and absolute judgment of the Engineering psychology and human performance”

2. Contents 1. Introduction 2. A Model of Human information Processing
3. Signal Detection Theory
4. Information Theory
5. Absolute Judgment
6. Summary and Further Study
7. Reference
This slide shows the outline of my presentation.
First, introduction , I’d like to talk about a model of human information processing
Next Signal detection , information Theory and absolute judgment
Finally summary and further study

3. Introduction
A Model of human information processing stages provides a useful framework for analyzing the different psychological processes used in interaction with systems and for carrying out a task analysis.
Signal Detection theory (SDT) will deal with the situation in which an observer classifies the world into one of two state: a signal is present or it is absent.
The detection process is modeled within the framework of SDT.
How the model can assist engineering psychologists in understanding the complexities of the detection process, in diagnosing what goes wrong when detection fails, and in recommending corrective solution.
The process of detection may involve more than two states of categorization. At this point , introduce information theory, and then use it to describe the simplest form of multilevel categorization, the absolute judgment task.
I’ll talk about a model of human information. This ~ 첫번째줄

4. A Model of Human information Processing(1)
Attention
Resources
Long-term
Memory
Working
Cognition
Sensory
Processing
STSS
Perception
Response
Selection
Execution
System
Environment
(Feedback)
Top-Down
This picture is a model of human information processing stages.
Main stream of model is consist of five stages. first, sensory processing, perception, cognition, response selection and Execution, finally feedback
Sensory Processing
Information and events in the environment must gain to the brain.
All sensory systems have an associated short-term sensory store (STSS) within the brain.
Delay time : Visual STSS: 0.5sec, Auditory : 2-4sec
Perception
Role of the stage is to decode meaning from the raw sensory data.
This processing has two important features.
1) Proceeds automatically and rapidly
2) Bottom-up processing: By sensory input
Top-down processing : By input from long term memory
Cognition and Memory
Important distinction between perception and cognition is that cognitive operations require greater time, mental effort, or attention and are carried out by using working memory.
Resource limited: conscious activities which transform or retain information
Long term memory: It is available to be transferred information in the working memory by learn.
Response Selection
The understanding of a situation, achieved through perception and
augmented by cognitive transformation.
Response Execution
The latter requiring the coordination of the muscles for controlled motion, to assure that the chosen goal is correctly obtained
Feedback
Actions are directly sensed by the human or, those actions influence the system, feedback loop will be observable sooner or later.
Attention
Supply of mental resources
Many mental operations are not carried out automatically but require
the selective application of these limited resources
Bottom-Up
Confirmation (intended goal)
Continuation (flow of information)

5. A Model of Human information Processing(2)
Sensory Processing
Information and events in the environment must gain to the brain.
All sensory systems have an associated short-term sensory store (STSS) within the brain.
Delay time : Visual STSS: 0.5sec, Auditory : 2-4sec
Perception
Role of the stage is to decode meaning from the raw sensory data.
This processing has two important features.
1) Proceeds automatically and rapidly
2) Bottom-up processing: By sensory input
Top-down processing : By input from long term memory
Cognition and Memory
Important distinction between perception and cognition is that cognitive operations require greater time, mental effort, or attention and are carried out by using working memory.
Resource limited: conscious activities which transform or retain information
Long term memory: It is available to be transferred information in the working memory by learn.
Raw sensory data relayed to the brain must be interpreted, or given meaning, through the stage of perception.

6. A Model of Human information Processing(3)
Response Selection
The understanding of a situation, achieved through perception and
augmented by cognitive transformation.
Response Execution
The latter requiring the coordination of the muscles for controlled motion, to assure that the chosen goal is correctly obtained.
Feedback
Actions are directly sensed by the human or, those actions influence the system, feedback loop will be observable sooner or later.
1)Flow of information can be initiated at any point
2)Flow of information is continuous
Attention
Supply of mental resources
Many mental operations are not carried out automatically but require
the selective application of these limited resources.

7. Signal Detection Theory (Paradigm 1)
There are two discrete states of the world:
Signal and Noise
Signal must be detected by the human operator
Two response categories are produced:
(1) YES (I detect a signal)
(2) NO ( I do not)
The combination of two states of the world and two response categories produces the 2 X 2 matrix, generating four classes of joint events
Hit, Miss, False Alarm,
Correct Rejection
This picture is the four outcomes of SDT
Signal detection theory is applicable in any situation in which there are two discrete states of the world (signal and noise) that cannot easily be discriminated
두번째읽고
In the process 세번째 읽음
<The four outcomes of SDT>

8. Signal Detection Theory (Paradigm 2)
The SDT model assumes that
(Green & Swets, 1996) :
(1) Sensory evidence is aggregated
concerning the presence or
absence of the signal.
(2) A decision is made about
whether this evidence indicates
a signal or not.
Neural Evidence (X) : Rate of firing of neurons at a hypothetical “detection center”
Critical Threshold: Response Bias
(XC)
X > XC : Operator Decision (YES)
X < XC : Operator Decision (No)
P(hit) + P(miss) = 1
P(false alarm) + P(correct reject)=1
This picture is Hypothetical distributions underlying SDT
This is noise distribution, this is signal distribution
The figure plots the probability of observing a specific value of X given that a noise trial (left cure) or signal trial (right curve)
When the absolute probability that X was produced by the signal equals the probability that it was produced by only noise, the signal and noise curves intersect.
When the curves intersect, Xc value is critical threshold
All X values to the right will cause the operator to respond yes. All to the left generate no responses.
Sensitivity of the Picture A is higher than Picture B.
Sensitivity is Separation of noise and signal distributions.
<Hypothetical distributions underlying SDT>
High sensitivity
Low sensitivity

9. Signal Detection Theory
Setting the Response Criterion: Optimality in SDT
Where is XC
“liberal” or “risky”
Left ←
Hit ↑
False alarm ↑
Β<1
“conservative”
Right →
Miss ↑
False alarm ↓
Β>1
How to set Beta
Ratio of the two curves
For a given level of XC
This slide shows the setting the response criterion
Operator’s conservative or risky behavior is determined by placing the decision criterion Xc.
If Xc is placed to the left, operator may be liberal or risky: prone to saying “yes” therefore detecting most of the signals that occurs but making many false alarms.
So hit and false alarm are increased
If Xc is placed to the right, they may be conservative :saying on most times and making few false alarms but missing many of the signals.
Beta is the ratio of neural activity produced by signal and noise at Xc or this is the ratio of the height of the two curves for a given level of Xc.
If Xc is placed to the left, beta is less than one.
If Xc is placed to the right bets is more than one.
So Xc and Bets define the response bias or response criterion.
We will first consider the influence of signal probability then payoffs on the optimal setting of beta and finally, human performance in setting beta.
Signal Probability
Payoffs
Human Performance

10. Signal Detection Theory
Signal probability
Defines where βopt should be set, but β is set is
determined by the observer and must be derived
from empirical data.
Signal Probability P(S) ↑ βopt↓ (made riskier),
Signal Probability P(S) ↓ βopt ↑ (conservative)
Payoffs
In this case, βopt is maximizing the
total expected financial gains or
minimizing expected losses.
V : Value of desirable events ( hit: H, correct rejection: CR)
C : Cost of undesirable events (miss: M, false alarm: FA)
When we are setting the response criteria This slide shows the influence of the signal probability and payoffs.

11. Signal Detection Theory
Human Performance in Setting Beta
Operator’s need to respond “creatively”
by introducing the rare response more
often than is optimal, since extreme
values of beta dictate long strings of
either yes (low beta) or no (high beta).
Operator misperceives probabilistic data.
There is evidence that people tend to
overestimate the probability of rare
events and underestimate that of
frequent events.
When we are setting the response criteria This slide shows the influence of the human performance

12. Signal Detection Theory
The Receiver Operating Characteristic (ROC) Curve
This graphic method is used to portray this equivalence of sensitivity across
changing levels of bias
The 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.
conservative
(1) Sensitivity
Separation of noise and signal
distributions.
d’: sensitivity measure
(2) Theoretical Representation
Of the four value, only two are
critical ( hit, false alarm)
The ROC curve plots P(h) against
P(fa) for different settings of the
response criterion
The value of beta at any given point
along the ROC curve is equal to the
slope of a tangent drawn to
the curve at that point.
Low sensitivity
I would like to talk about ROC curve
ROC curve is the receiver operation characteristic.
Β = slope

13. Signal Detection Theory
Applications of Signal Detection Theory (SDT)
Benefits
(1)It provides the ability to compare sensitivity and therefore
the quality of performance between conditions or between
operators that may differ in response bias.
(2) it provides a diagnostic tool that recommends different
corrective actions.
Medical diagnosis
Recognition Memory and Eyewitness Testimony
Industrial Inspection
This slide shows applications of SDT
By partitioning performance into bias and sensitivity components, depending on whether change in performance results from a loss of sensitivity or a shift in response bias.

14. Information Theory
How to quantify this flow of information so that different tasks confronting the human operator can be compared
Measure task difficulty by determining the rate at which information is presented.
Measure processing efficiency, using the amount of information an operator processes per unit of time.
Provides metrics to compare human performance across a wide number of different tasks.
Information Theory
The Number
Of Events
Probabilities Of
events
Sequential
Constraints
and
Context
I’d like to talk about the information theory In this slide
A fundamental issue in engineering psychology is~
Using the information theory We can measure~
We can also ~
Therefore Information theory ~
Also quantification of information is influenced by three variables
The number of possible events that could occur event N
Probabilities of events
The event’s sequential constraints, or the context in which they occur.

15. Information Theory The number of Events: Probabilities of Events
When all alternatives are equally likely, the information conveyed by an
event Hs, in bits, can be expressed by the formula.
where N is the number of equally likely alternatives.
Information theory has a quality of optimal performance.
This is based on the minimum number of questions therefore arrives at a
solution in a minimum time.
Probabilities of Events
The probabilistic element of information is quantified by making rare events convey more bits.
where Pi is the probability of occurrence of event i.
First talk about the number of events
Next probability of evet

16. Information Theory
Psychologists are often more interested in measuring the average information conveyed by a series of events with differing probabilities that occur over time.
Event A B C D Pi
0.5
0.25
0.125 2 4 8 1 3
0.375
If probabilities of events are different, using this formula.
For example there are four events a,b,c,d, with probabilities of
The computation of the average information conveyed by each event in a series of such events would proceed as follows
This value is less than log4=2 if probabilities of events is the same
This means ~
Low –probability events convey more information because they occur infrequently. However, the fact that low-probability events are in frequent causes their high-information content to contribute less to the average.

17. Information Theory Sequential Constraints and Context Redundancy
Given a particular context, it may be highly expected, and therefore its occurrence conveys very little information in that context.
Absolute probability of the event Pi is now replaced by a contingent probability Pi|X (the probability of event I given context X)
Redundancy
Defines this potential loss in information
where Have is the actual average information conveyed taking onto account all three variable. Hmax is the maximum possible information that would be conveyed by the N alternatives if they were equally likely.
The term Redundancy formally
Percent redundancy is quantified by the formula

18. Information transmitted through the system No information transmitted
Information Theory
Information Transmission of Discrete Signals
Concept of information channel
How much information is presented to an operator
Channel Capacity : How much information is transmitted from stimulus to
response
Bandwidth : How rapidly information is transmitted. Hs HR HT
Hloss
Noise Hs HR HT
Hloss
Noise
Information transmitted through the system
No information transmitted
This slide shows information transmission of discrete signals
Concept of information channel is consist of three part
First
HS: Value of stimulus information
HR: Response information
HT: Information faithfully transmitted
HL: Information loss

19. Information Theory
Stimulus A B C D 2
(0.25)
Response
Stimulus A B C D 1 2
This slide show the example that quantities Hs,Hr and Hsr are calculated.
If each stimulus generates consistently only one response
If there is greater dispersion within the matrix, there are more bit within hsr more than hs hs more than ht
Response

20. Absolute Judgment Absolute Judgment:
An observer assigns a stimulus into one of multiple categories along a
sensory dimension.
Inspector of wool quality who must categorize a given specimen into one of several quality levels.
Driver who must interpret and recognize the color of a display symbol appearing on his map display.
I would like to talk about absolute judgment
This is an~
For example
Absolute judgment is consist of the these factors
I will talk about those factors next slide

21. Typical human performance in absolute judgment tasks
Single Dimensions
Channel Capacity (Experimental Results)
The level of the late part indicates the channel capacity of the operator. (2~3bits)
Errors began to occur (HT<HS)
George miller
Edge Effect (Experimental Results)
Stimuli located in the middle of the range of presented stimuli are generally identified with poorer accuracy than those at extremes (Shiffrin & Nosofsky, 1994) 1 2 3 4 5 Hs Ht
Perfect Performance
Human Performance
Typical human performance in absolute judgment tasks

22. Number of combined dimensions (Increasing Hs)
Absolute Judgment
Multidimensional Judgment
Most of our recognition is based on the
identification of some combination of two
or more stimulus dimensions rather than
levels along a single dimension.
Orthogonal dimensions
The level of the stimulus on one dimension can take on any value, independent of the other
For example, Weight and Hair color
As more dimensions are added, more total information is transmitted, but less information is transmitted per dimension. 1 2 3 4 5
Number of combined dimensions (Increasing Hs)
Total Ht
Perfect Performance
Human Performance 6 7
Bits/dimension
Human performance in absolute judgment of multidimensional auditory stimulus

23. Absolute Judgment
Correlated Dimensions
The level on one constrains the level on another.
For example, Height and weight
As more dimensions are added, the security of the channel improves, but HS limits the amount of information that can be transmitted.
Orthogonal dimensions maximize HT, the efficiency of the channel.
Correlated dimensions minimize Hloss; that is, they maximize the security of the channel.
Number of combined dimensions (Increasing Hs)
Information transmitted, Ht Hs
Hloss
Human performance in absolute judgment of multidimensional auditory stimulus

24. Summary and Further Study
The model of that we have described for Human information processing is not a computational one that can provide estimates, say, of the time required to perform certain tasks, or the error rates expected. Rather, it provides a general framework for analyzing human performance.
We have seen how people classify stimuli into two levels along one dimension, several levels along one dimension, and several levels along several dimensions.
Signal detection was characterized by the probabilistic element of decisions under uncertainty.
Further Study
Engineering Psychology and Human Performance
Manual Control
Attention, Time-Sharing, and Workload
Stress and Human Error
Complex Systems, Process Control, and Automation

25. Reference Engineering Psychology and Human Performance
Third Edition, Wickens & Hollands