Chapter 4 Flashcards

1
Q

What is Hull’s theory?

A

When the stimulus conditions are right, the response is made. (You just do it)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is Thorndike’s theory?

A

Law of effect and law of exercise. Thorndike chose the law of effect, but both turned out to be important.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is the Law of effect?

A

Reinforcement is needed to strengthen bonds

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is law of exercise?

A

That mere repetition of behavior is enough (Hebbian Learning)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What is Tolman’s theory?

A

Latent learning, you only turn learning into performance when there is a goal.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What are some relevant brain structures?

A

The basal ganglia
The hippocampal and prefrontal regions
The anterior cingulate cortex (ACC)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What does the basal ganglia do?

A

Responsible for acquisition and application of procedures in the form of reinforcement learning.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What does the hippocampal and prefrontal regions do?

A

Responsible for storage and retrieval of declarative knowledge

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What does the anterior cingulate cortex do?

A

Responsible for cognitive control in the selection of appropriate behavior?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What are the three phases of skill acquisition?

A

Cognitive
Associative
Autonomous

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What is the cognitive phase?

A

The learner computes based on existing facts, such as counting. In ACT-R this is based on the Declarative memory.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is instance-based learning?

A

The problem and answer are learned by repeated computing (exposure). This takes the learner from the cognitive phase to the associative phase. Declarative memory

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is the associative phase?

A

The learner retrieves the problem and its answer from the fact in the declarative memory. The reaction time is reduced.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What is the step from associative phase to autonomous phase?

A

Utility learning. By more exposure you learn the skill in the form of utility learning. (In ACT-R it is also production compilation)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What is the Autonomous phase?

A

The skill is in the procedural memory, meaning the problem can happen “unconsciously” and in a reflex-like matter. The Decalrative memory is not used anymore and the reaction time is reduced a lot.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What is the memory activation in the cognitive phase?

A

The memory activation is related to retrieving count facts (5+4=9, 9+3=12)

17
Q

What is the memory activation in the associative phase?

A

The memory activation related to the answer directly (5$3=12)

18
Q

What is the memory activation in the autonomous phase?

A

no memory activation

19
Q

What is the difference between Instance-based learning and Utility learning on the symbolic level?

A

Instance-based learning: new chunks through vision, imaginal
Utility learning: new production rules through production compilation

20
Q

What is the difference between Instance-based learning and Utility learning on the subsymbolic level?

A

Instance-based learning: through activation, likelihood and speed of retrieval
Utility learning: through utility, likelihood of usage

21
Q

What is the activation for instance-based learning affected by?

A

prior usage, current context and noise

22
Q

What is the utility for utility learning affected by?

A

prior utility, reward, time to reward and noise

23
Q

What is utility learning based on?

A

The expected reward of using a rule, discounted by the time until that reward. The highest utility has then the highest probability to be selected.

24
Q

How is utility learned?

A

By reinforcement learning.

25
What is the utility learning equation?
See slides
26
What is the past tense debate?
It is a debate of different theories of how children learn the past tense of words.
27
How do children learn the English past tense with age?
Stage 1: talk - talked, bring - brought Stage 2: talk - talked, bring - bringed Stage 3: talk - talked, bring - brought
28
Why is there a U-shape?
This is because children learn the rule and then unlearn it and then learn the rule again
29
What are the three theories proposed?
Symbolic theory Connectionist Theory Hybrid in ACT-R
30
What is the Symbolic theory?
Regular verbs are represented by a rule and irregular verbs are memorized The U-shape coincides with the learning of the regular rule. Overgeneralization must be the result of rule learning because it cannot be from experience
31
What are some problems with Symbolic theory?
How is the rule discovered? Does not explain wrong overregularization (bring-brang)
32
What is the Connectionist Theory?
All knowledge is represented in a neural network U-Shape due to growth in vocabulary, decreasing activation threshold of other regular forms, there are more forms ending with "ed" than irregular endings
33
What are some problems with the Connectionist theory?
Only works when regular verbs are a majority Relies on a sudden increase in vocabulary Produces too many types of erros Children learn the past tense without feedback
34
What is production compilation?
Two rules that fire in sequence are combined into one new rule
35
What is the Hybrid in ACT-R theory?
Uses production compilation create 2 rules: 1. Add "-ed" to stem rule 2. Production rule for each irregular verb
36
How does the Hyrbid model explain the U-shape?
The start no production compilation, so retrieval is based on memory. High accuracy till the first production rule is compiled (add -ed stem). Till production rule compiles the second rule, then the accuracy increases again.