Learning - Villanova University

Download Report

Transcript Learning - Villanova University

Learning
Learning
Behaviorism
Psychology is the “science of behavior.”
Emphasis on what can be directly observed. Ignore the mind (unobservable).B. F. Skinner
Emphasis in Reinforcements / Rewards
Mental Processes
Stimulus
Behavior
(Response)
Classical conditioning
Ivan Pavlov (early 1900s)
• Russian physiologist
• Nobel prize winner for his work on the digestive system
» Interested in the salivary reflex
» Do different types of food elicit different amount of saliva?
http://www.mphy.lu.se/avd/nf/hesslow/bilder/pavlov.jpg
Classical conditioning
Food
Salivation
Reflex
Salivation
Not Reflex
Learned behavior!
Footsteps
Food bowl
Bell
Classical conditioning
How does classical conditioning work?
Step 1: Find a stimulus that elicits a reflexive behavior
Step 2: Pair stimulus with a neutral stimulus that does not normally elicit the
same reflexive behavior
Step 3: Over time, if neutral stimulus always signals the original stimulus,
the previously neutral stimulus will now elicit the reflexive behavior
Neutral Stimulus
(i.e., bell)
Stimulus
(i.e., food)
Reflexive response
(i.e., salivation)
Classical conditioning
Neutral Stimulus
(i.e., bell)
Stimulus
(i.e., food)
Reflexive response
(i.e., salivation)
Food — Unconditioned stimulus (UCS)
Salivation — Unconditioned response (UCR)
Bell — Conditioned stimulus (CS)
Salivation (in response to bell) — Conditioned response (CR)
Classical conditioning
What are the requirements for classical conditioning to work?
• The conditioned stimulus (i.e., bell) must be previously neutral
• Contiguity during acquisition
»
the CS must come immediately before the UCS
» The CS should signal the UCS
• Contingency
» If the CS no longer signals the UCS reliably, the CR will stop (i.e., extinction)
• Spontaneous recovery
» Even after extinction, it is possible to re-establish the CS - CR relationship
If all of these conditions are followed, any neutral stimulus
has the potential to become a conditioned stimulus
Classical conditioning
Different stages of classical conditioning
Classical conditioning
Little Albert (8-month old infant)
Neutral Stimulus
(i.e., white rat)
Stimulus
(i.e., loud noise)
Reflexive response
(i.e., startled & cry)
Stimulus generalization
• Previously neutral stimuli that are similar to the CS will now elicit the CR
Click here for video
Classical conditioning
Classical conditioning in humans
“Give me a dozen healthy infants, well formed, and
my own specified world to bring them up in and I’ll
guarantee to take any one at random and train him to
become any type of specialist I might select —
doctor, lawyer, artist, merchant-chief, and yes,
even beggar-man and thief, regardless of his
talents, penchants, tendencies, abilities, vocations
and race of his ancestors.”
- John B. Watson, 1924, Behaviorism
What a freak!
Classical conditioning
Application of classical conditioning theories: Drug addiction
Neutral Stimulus
(i.e., needle, back alley)
Stimulus
(i.e., drug)
Reflexive response
(i.e., physiological response)
Classical conditioning
But ... classical conditioning doesn’t explain all learning
• Not all unconditioned stimuli are equal
» Some fears are easier to condition than others
• Does not account for reward and punishment
http://www.theage.com.au/ffximage/2007/04/16/snake,0.jpg; http://www.sunbeamhospitality.com/resources/images/products/3278-500-600/3278int_xlg.jpg
Operant conditioning
Operant (instrumental) conditioning
• All actions serve some purpose
• We learn to behave in a certain way because those behaviors lead to rewards
• We learn NOT to behave in a certain way because those behaviors lead to
punishment (or no reward)
• Consequences matter
Operant conditioning
Operant conditioning
Reward
Increase in behavior
Activates reward system (n. accumbens)
Punishment
(or no reward)
Decrease in behavior
Operant conditioning
B. F. Skinner
• The organism behaves in a certain way in order to change the environment
» not merely a passive recipient of the conditioned stimulus
» behaviors repeat because they have been reinforced
» consequence of behavior matters
http://www.bfskinner.org/images.asp
Operant conditioning
Reinforcement — a positive consequence that follows a behavior
Punishment — a negative consequence that follows a behavior
Change in environment
Likelihood to repeat behavior
increase
decrease
positive
reinforcement
positive
punishment
(get to use car more often)
(additional hassle)
stimulus
negative
reinforcement
negative
punishment
removed
(no longer dependent on
(take away driving
parents)
privileges)
stimulus
added
Operant conditioning
How to measure effects of reinforcement/punishment?
Change in environment
Likelihood to repeat behavior
http://www.cs.tcd.ie/research_groups/crite/personal/imgs/skinner3.gif
stimulus
added
stimulus
removed
increase
decrease
positive
reinforcement
positive
punishment
(water)
(foot shock
introduced)
negative
reinforcement
(foot shock
removed)
negative
punishment
(water
removed)
Operant conditioning
How do you get the animal to do what you want to begin with?
• Shaping
» Successive approximation — reward behavior that is similar to the desired
behavior
Initial behavior
Sitting
Standing
Standing on hind legs
Standing near lever
Sniffing lever
Desired behavior
Pressing lever
Operant conditioning
What does it take for operant conditioning to work?
• Reinforcers must be presented immediately after the behavior
» Primary reinforcer — fulfills basic needs (e.g., food, water)
» Secondary reinforcer — more abstract (e.g., money, fame)
• Delivery of the reinforcer must be consistent
Operant conditioning
How often should a reinforcer be given?
Reinforcement schedules
continuous reinforcement
partial reinforcement
(i.e., desired behavior is
(i.e., desired behavior is
reinforced every time)
reinforced intermittently)
Operant conditioning
Reinforcement schedule
Partial reinforcement
Ratio
Fixed
Interval
Variable
Ratio
Fixed
Variable
Fixed
Variable
Interval
every 5th lever press
every 20 seconds
3rd, 8th, 19th, 20th, 35th
2s, 8s, 34s, 36s, 40s
Predictable
Unpredictable
Operant conditioning
How is behavior influenced by the different schedules?
Casinos use variable ratio reinforcement
Limitations of conditioning theories
But ... conditioning theories do not explain all learning
• One trial learning: taste aversion
» A rat given sweetened water before radiation (which causes nausea) will later avoid
sweetened water, even after just one exposure
• Insight learning
» Kohler’s experiment
Video _pigeon
Figure 6.6
Limitations of conditioning theories
But ... conditioning theories do not explain all learning
• Remember ... Skinner claimed that learning is a result of reinforcement
• If that’s true, then unreinforced behaviors should not be learned
» Latent Learning (Edward Tolman)
» Trained rats to run mazes
» Will rats learn without reinforcements?
Limitations of conditioning theories
Latent learning: Will rats learn without reinforcements?
• Train rats to run a maze under 3 conditions:
» continuous reinforcement
» no food reward until the 11th day
» no food reward ever
Remember, conditioning theories state that ...
• Reinforcement is necessary for learning to
occur
http://psychclassics.yorku.ca/Tolman/Maps/FIG1.gif
Limitations of conditioning theories
Latent learning
No food
No food (first 10 days)
No food
Continuous food
Continuous food
No food (first 10 days)
Learning took place even when there was no reinforcement
Learning was revealed after behavior was reinforced
Limitations of conditioning theories
Learning through observation (in humans):
• Albert Bandura’s Bobo doll experiment
Results:
Children who observed violent adult interactions
with bobo doll displayed more aggressive
interactions with the doll than children who
observed non-violent adult interactions with doll
Observational learning video
http://academic.evergreen.edu/h/hiljus01/bobo.gif
Bobo doll
Limitations of conditioning theories
Learning through observation (in monkeys):
• Monkeys raised in the lab environment did not show fearful responses towards
snakes
• Shown footage of other monkeys being fearful of snakes
• Monkeys raised in lab developed a fearful response towards snakes
Conditioning: Phobias
- overgeneralization: a dog bit me -> afraid of all dogs  afraid of animals 
afraid of going outside
- role of classical conditioning: CS (dog), US (dog bite)  UR (pain, fear)
=> CS (dog) – fear response;
role of operant conditioning: avoid CS (dog)  reduces anxiety => continue
avoiding CS
- treatment: pair fear stimuli with relaxation (or other pleasurable stimuli)
other things: fear of some stimuli (snakes, heights, dark) more common than
others (knives, electrical outlets): argues against equipotentiality, in favor of
evolution
-illusory correlation between feared stimulus and negative outcome
Conditioning: Physical punishment by parents
- Should we use punishment to eliminate unwanted behavior? No.
but if you do, do immediate punishment (more effective than delayed
punishment, this is true even for your pets!)
Why not use punishment?
- more punishment => more aggressive behavior by the child
=> child likely to use physical punishment when he grows up
Does punishment really cause aggression?
Alternatives hypotheses:
- aggression causes punishment
- a third variable (e.g., genetics) causes aggression & punishment
External vs. Intrinsic rewards
Nursery school children
Paper & crayons: draw whatever you want
External reward group: “Good Player Award”
Intrinsic reward group: nothing
External reward  increased amount of time drawing, BUT
Ending external reward reduces drawing time below the intrinsic
reward group!
should we give students extra-credit for attending talks in their major?
Lepper & Greene, 1978
Operant Conditioning: Infant sleep
Infant Sleep
Extinction is harder to achieve after intermittent reward (partial
reinforcement)
- baby cries, dad breaks down and goes in to console the baby:
- baby learns that crying leads to reward (daddy comes)
- to make matters worse, dad only breaks down sometime (partial
reinforcement)
Operant Conditioning: other applied issues
- ‘variable ratio’ reinforcement also occurs in casinos
-Operant conditioning is used
- to train animals (flipper, sea world)
- for Token economies in psychiatric hospital: target behavior,
reinforcers (sometimes fail to transfer to outside world)
- in Applied behavioral analysis (for autism)
- for Behavior modification (kazdin, slate)
Take home message
Operant conditioning
Classical conditioning
• Association between neutral and
reflexive stimuli leads to learning
• Passive process
• Consequences matter
• Takes rewards and punishments
into account
• Active engagement with the
environment
Limitations to conditioning theories:
• Not all unconditioned stimuli are the same
• One-trial learning
• Insight learning
• Latent learning
• Learning through observation
Other things to include:
Classical conditioning example: baby bottle suck
Species specific behavior: racoons & pigs
Figure 6.11
Figure 6.12