Download 16 - Angelfire

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Theoretical ecology wikipedia , lookup

Reinforcement learning wikipedia , lookup

Transcript
Psy 356
04/08/2010
I. Extinction Review
-
Extinction produces:
o Multiple attempts
o Increased response variability
o Frustration
- Extinction does not reverse acquisition!
So how could spontaneous recovery contribute to getting back together
Renewal rekindle the old flame
o CONTEXT IS IMPORTANT IN RENEWAL
Reinstatement leads to “another try”
o Seeing an individual can renew U.S
Extinction
II.
Reinforcer Devaluation in Extinction
Rescorla (1993) Results
-
Devaluation in non-extinguished rats decreased FR3
Devaluation in extinguished rats decrease R1
The R-O association persists despite extinction
This is also true of the CS-US relationship and the S-O relationship
What is learned during extinction?
Pavlovian Conditioning:
S-S appears more important than this S-R
Instrumental Conditioning
R-O and S-O associations are more important than S-R
- S-R associations are most important in extinction
-S-R association can account for renewal and reinstatement
-
Non-reinforcement produces an inhibitory S-R association
Therefore, non- reinforcement in the presence of specific stimuli (S) produces an
inhibition of response ®
Ex: Renewal- the extinction context is often the specific stimulus (S)
Courtney Love example
III.
What is learned in Extinction
A.
The S-R association
- The inhibitory S-R relationship is dependent upon a history of reinforcement:
- Every response that is not reinforced leads to frustration
Psy 356
04/08/2010
- This aversive emotional reaction produces the inhibitory S-R relationship (“Why even
try?”)
B.
Inhibitory conditioning
( Does extinction make the CS and inhibitor?)
1.
Train
A+/B+
2.
Train
A-
Summation Test
Ext
A-
Test
Test: AB
Retardation Test
Ext
A-
Train
A+
 Does A inhibit responding to B?
YES!!
<= Does A inhibit manipulation?
YES!
Question1
Extensive reinforcement training should slow the rate of extinction?
False, because of overtraining
Question 2
Responding declines more rapidly in extinction following reinforcement with a larger reinforcer
True, because the organism becomes frustrate
Question 3
Continuous reinforcement produces more resistance to extinction than partial reinforcement
False, because
IV.
Paradoxical Effects of Extinction
A.
Overtraining: If decreased responding is due to frustration of unexpected lack of
reinforcement, more rapid extinction should occur when expectancy is higher
Paradox: A true statement leading to a contradiction of intuition
- Does more extensive training produce resistance to extinction? NO!
Ex: Puppy training
B.
Magnitude Reinforcement extinction effect: More rapid extinction should occur
when reinforce expectancy is larger
- Does responding decrease more rapidly in extinction following training with a larger
reinforcer? YES!
Ex: Allowance and chores
$10 decrease rate of extinction
$100 increase rate in extinction
- The larger the reward the higher the rate of extinction ( High and low contrast)
Psy 356
04/08/2010
C.
Partial Reinforcement Extinction Effects (PREE): Extinction is slower if
partial rather than continuous reinforcement was previously in effect
Ex:
Continuous reinforcement
ATM machine
Response
Dip card in
Partial reinforcement
Playing roulette
Sometimes you win and sometimes you lose
Reinforcer
Get money
V.
Mechanisms of PREE
- PREE in rat pups: Chen and Ansel, (1996)
- Subjects rats
- 10 D old
-12 D olds
Training Alley running (speed)
Reinforcer: Milk
- Continuous reinforcement (CRF)
- Partial reinforcement (PR)
Extinction
Removal of reinforcer
Findings:
- Rats start running faster for CRF
- P13 rat extinction
CRF > PR
P11 rat extinction
CRF=PR
- PREE is acquired by P13 and is likely based on brain maturation
A.
Discrimination Hypothesis: Introduction of extinction is easier to detect after
continuous reinforcement vs. partial reinforcement
Jenkins (1962) and Thelps (1962): Test of discrimination theory
Group 1
Group 2
Training 1:
PR
CRF
Training 2:
CRF
CRF
Extinction
Immediately after training 2
Findings
Group 1: Demonstrated more resistance to extinction
B.
Frustration Hypothesis: Persistence in extinction results from learning to respond
when you expect to be non-reinforced or frustrated
Psy 356
04/08/2010
- Dependent upon intermittent reinforcement the presentation of random reinforced and
nonreinforced trials
Early training
=>
Reinforcement increases responding
Reinforcement decreases responding
Late Training
Reinforcement increases responding
Reinforcement increases responding
- Non reinforcement leads to expectation of no reward, so when they are unexpectedly reinforced
with training responding increases
C.
-
Sequential Hypothesis: Subjects remember rewarded trials followed by
nonrewarded trials and the memory of the nonrewarded trials serve as
cue for the next reward
The memory of reward created a “It’s got to happen sooner or later” idea
VI.
Behavioral Momentum Hypothesis” Not a paradoxical reward effect suggesting
persistence in extinction may represent resistance to change in the face of extinction
-
Suggests that behavior with momentum is difficult to stop
P= m*v
P= momentum
M= mass
V= velocity
While this theory is useful for comparing rates of reinforcement, it has difficult for
accounting for PREE