Signal Detection Theory

Question 1

What is signal detection theory?

Answer 1

Model of how humans separate “noise” (all stimuli, whether relevant or not) from “signal” – the relevant, important information.

(examples: Part defect, vs. just a light reflection; Photo has “red eye” or it doesn’t (photo processing/inspection problem))

Question 2

Describe the Signal Detection Theory and Information Theory (IT) diagram.

Answer 2

Question 3

What happens in signal detection theory?

Answer 3

Collect observation (through sensory systems), then determine if there is enough evidence to constitute a signal. The observer discriminates signals from noise.

(Example: Crack is large enough; Eyes are red enough)

Question 4

What assumption is made in singal detection theory?

Answer 4

Normally distributed, random noise - the signal increases the mean of the distribution.

Question 5

What is the signal detection theory paradigm?

Answer 5

Question 6

What are the 2 error types?

Answer 6

1) Type 1 Error: Accidently concluding that independent or causal variables had an effect when it was really just chance.
2) Type 2 Error: Concluding that the independent or causal variables did not have an effect when in fact they did.

Question 7

What is involved in the collection of a signal as it relates to the process of singal detection?

Answer 7

Sensory evidence (X) is aggregated concerning the presence or absence of a signal (with corresponding neural activity)

Question 8

What is typically found in signal detection?

Answer 8

1) low levels of sensory evidence
2) Random variation in environmental levels of X, and in baseline neural firings

Question 9

What can issues related to random variations lead to?

Answer 9

1) False alarms, if random variation leads to higher than normal X
2) Misses, if baseline, environmental levels lower than normal

Question 10

When to errors in detecting between signal and noise occur?

Answer 10

With smaller differences in intensity between signal and noise, their is a greater likelihood of errors.

Question 11

Describe signal theory demonstration graph.

Answer 11

Question 12

What two important performance parameters of signal detection?

Answer 12

1) response bias b
2) sensitivity d”

Question 13

What is the ROC curve?

Answer 13

Representation of performance at one level of sensitivity (a characteristic of the receiver), across all possible criterion levels.

Question 14

What is signal detection influenced by?

Answer 14

Signal detection is influenced by sensitivity and response bias.

Question 15

What are the two states of the world, as it relates to the signal detection theory matrix?

Answer 15

Two states of the world are:

1) Signal Present (+ Noise)
2) Singal Absent (Noice Only)

Question 16

What are the two response from the observer in the signal detection theory model?

Answer 16

Yes = signal is present.

No = signal absent, only noise.

Question 17

What is the Signal Detection Matrix?

Answer 17

The data from a signal detection environment represented in the form of a matrix. It combines the two stsates of the world and the two respones that make up the Signal Detection Matrix.

Question 18

What are the four cells that make up the signal detection matrix?

Answer 18

1) Hits
2) False alarms
3) Misses
4) Correct rejections

Question 19

How do error types apply to the signal detection theory paradigm?

Answer 19

Question 20

How are decisions made for signals?

Answer 20

Decision regarding signal present vs. absent - if eveidence exceeds criteria Xc (criterial threshold).

Question 21

What are the two performance parameters of signal detection?

Answer 21

• Repsonse Bias (denoted as b)
• Sensitivity d’ (d prime)

Question 22

What is criterion level?

Answer 22

Criterion level, often abbreciated as Lc, is the steady noise level permitted for a full eight hour work shift.

Question 23

If there is a liberal response bias, what is the response likely to be?

Answer 23

For liberal response bias, people are prone to say “signal”

Question 24

If there is a conservative response bias, what response likely to be?

Answer 24

A conservative reponse bias is not prone to say “signal”.

Question 25

Which is better: a false alaem, or a miss

Answer 25

It depends on the task and the consequences. For example an air traffic controler will have a higher consequence for a miss that a false alarm, so they are more likely to have false alarms.

Question 26

What is the impact of types of errors made in response bias?

Answer 26

False alarms and misses.

Question 27

What is the signal detection equation?

Answer 27

Question 28

The goal is to reduce the overlap in the distrubtions. What are the two ways in which that can be accomplished?

Answer 28

1) Increase intensity of signal (collect better and stronger evidence at a criminal case).
2) Imrpove the ability of the observer to discriminate through training and practice, or thorugh increasing the DT (smaller probe tip)

Question 29

Is sensitivity considered a bottom-up or top-down approach?

Answer 29

It is a bottom-up approach.

Question 30

When is sensitivity considered to be higher?

Answer 30

Sensitivity is higher if there are more correct responses and fewer errors.

Question 31

How can sensitivity be degraded?

Answer 31

Through the degrading of certain conditions.

Question 32

Can sesnitivity be improved?

Answer 32

Yes.

Question 33

Who can sensitivity be improved?

Answer 33

Sensitivity can be improved through on the job training (to a point) and through experience.

Question 34

What does thte curve on the ROC represent?

Answer 34

The curve represents differences in neural activity:

1. Difference due to physical properties of the signal.
2. Differences in the observer (perceptual capabilities, training)

Question 35

What does a low d’ lead to?

Answer 35

Low signal detection performance.

(example: failure to detect defects in lumber; lace of memeory to memorize signal)

Question 36

What is a counter measure to low d’?

Answer 36

Memory aid.

Question 37

How is bias represented in the signal detection equation?

Answer 37

It is represented as beta.

Question 38

In deteriming bias (beta), what does a beta value less that 1 indicate?

Answer 38

It indicates a bias toward the yes response.

Question 39

In determining bias (beta) , what does a response greater than 1 indicate?

Answer 39

It idicates a bias toward the no response.

Question 40

What does a bias (beta) response of 1 indicate?

Answer 40

A bias response of 1 indicates no bias toward the no or yes response.

Question 41

What does a response bias reflect?

Answer 41

The response bias refelcts the bias of the operator to respond to “yes, signal” versus “no, noise”.

Question 42

Is the response bias a top-down, or bottom-up process?

Answer 42

The response bias is a top-down process.

Question 43

Response bias is typcially affected by which two variables?

Answer 43

1) Operator expectancy (if the operator expects a “yes” they end up having misses because they are bias to yes.)
2) Changes in the values of the four different kinds of events, which can shift the criterion.

Question 44

What is the outcome of a signal detection analysis plotted along?

Answer 44

The outcome of a signal detection analysis is plotted along a receiver operator characteristic (ROC) curve.

Question 45

What is absolute judgement?

Answer 45

Absolute judgement refers to the limited human capacity to judge the absolute value of the variable signaled by a coded stimuli.

Question 46

What does discrimination refer to?

Answer 46

Discrimination refers to judgement of differences between sources of information that are actually or potentially present. Generally people are good at this task as long as the differences are not small and the viewing conditions are favorable.