Learning Flashcards
Rescorla-Wagner Model
level of conditioning (classical) is a result of an internal comparison between expected and actual strength of US.
- expectation is based on prior experience with US.
- strength of US is fixed
- Assumes CR gets stronger if the CS-US pair is SURPRISING.
∆V = αβ (λ-V)
∆V – change in associative value of CS (rate of learning)
α – salience of the US to promote conditioning
β – strength of the US to promote conditioning
λ – Magnitude of associative value that can be conditioned for CS; actual CS value (how predictive CS is of US)
V – Current associative value of CS (expectations about the CS-US association/CS value)
- previous models assumed that a CR gets stronger the more a CS is paired with a US
- blocking - neutral stimulus doesn’t change how SURPRSED they are of the CS.
- superconditioning - you are SURPRISED that the US is presented when the neutral is paired with the new excitatory.
Reward variables
- drive
- magnitude
- delay
Pavlovian (classical) conditioning
learning by association
Operant conditioning (Skinner)
learning by reinforcement
Escape learning
- like negative reinforcement
- emit a response that terminates an aversive consequence
Avoidance learning
- can lead to phobias
- emit a response to prevent the occurrence of an aversive consequence altogether
Learned helplessness
- long term effect of repeated exposure to punishment
- believe behaviour has no effect on what happens to them
- immunise by an initial experience of control
prototypical Pavlovian conditioning experiment
.
Shaping
selective reinforcement of behaviour resembling the desired target behaviour in order to reach the target behaviour. (eg. going within 5cm from the lever, touching the lever. so that eventually it will pull the lever)
Skinners thee main categories of reinforcers
primary -
secondary -
generalised -
Reward variables
drive
magnitude
delay
Premack Principle
more probable behaviours will reinforce less probable behaviours (eg. eat broccoli before you can eat favourite food).
Escape learning
- like negative reinforcement
- emit a response that terminates an aversive consequence
Avoidance learning
- can lead to phobias
- emit a response to prevent the occurrence of an aversive consequence altogether
Learned helplessness
- long term effect of repeated exposure to punishment
- believe behaviour has no effect on what happens to them
- immunised by an initial experience of control
punishment limitations
- isn’t as permanent as reinforcement
- reduces trust / increases aggression
How to punish effectively
- no escape
- intense as possible (within reason)
- continuous schedule
- no delay
- over short period of time
- no subsequent reinforcement
Other than schedule, what else can affect conditioning?
- Drive - eg. hungry with food reward is faster then sated.
- Size - may learn faster is reward is bigger
- Delay - immediate reward may be better then delayed reward.
The Three Term Contingency
- discriminative stimulus (sets the occasion)
- operant response (behaviour)
- outcome (reinforcer/punisher) that follows (consequence).
Stages of Classical conditioning
- Habituation (CS alone = UR)
- Acquisition (CS + US = UR)
- Extinction (CS alone = CR)