Learning

Document Sample
Learning Powered By Docstoc
					         Learning

Psychology Summer School 2005

          Peter Ward
       Reference:

Gleitman: Chapter on learning

Any other basic psychology textbook
should cover much of the material
          Overview of Lecture
1.   Why is learning important?
2.   Classical conditioning
3.   Instrumental conditioning
4.   Cognitive views on learning
                Why Learning?
   Humans are born with a number of ‘built-in
    programs’
       » Muscle reflexes
       » The Universal Grammar
   But much of what humans do and think is a result
    of learning.
   The things that make individuals, societies and
    nations different are almost all due to learning.
   Obviously a complex process, so psychologists
    have tried to understand it from the most basic
    level upwards
         Learning and Memory
   The simplest form of learning: Habituation
Habituation


     (loud noise)
Habituation




     Goodness me! What a surprise!
Habituation




     That’s still quite surprising…
Habituation




     Okay, I’m less surprised now
Habituation




              Boring.
                  Habituation
   The point being…
    – Leaning involves REMEMBERING
    – Strong link to another crucial area of
      psychology



    – Moving on…
       Part 2

Classical Conditioning
          Classical Conditioning
   First type of learning studied in depth
   Focused on learning associations between events.

   Russian physiologist, Ivan Pavlov (1849-1936),
    worked on digestive reflexes in dogs
       » Realised he could study learned reflexes as well as innate ones
       » Discovered he could actually train, or condition reflexes (hence
         the name)
            How did he do it?
   Pavlov and his amazing drooling dogs.
        How did he do it?
                               The man himself
A dog




               Some Russians
              How did he do it?
 Dogs in the lab salivated when you put meat
  powder in their mouths.
 But they would also salivate when
    –   They saw the meat
    –   They saw the dish
    –   They saw the person who brought it
    –   They heard the person’s footsteps
Classical Conditioning Method

      Food = salivating
      Food + Bell = salivating


        …do it enough times…

        Bell = salivating
     Classical Conditioning Theory
   Different types of response, different types of
    stimulus
    – The innate bit:
       » Unconditioned stimulus = meat
       » Unconditioned response = salivating
    – The learned bit
       » Conditioned stimulus = bell
       » Conditioned response = salivating (to bell alone)
In pictures:
In pictures:
In pictures:
           Classical Conditioning
   Same pattern of learning found in
    –   Crabs
    –   Ants
    –   Ant eaters
    –   Cats
    –   PEOPLE

     Fundamental type of learning!
       The important parts of
       Classical Conditioning
1. Acquisition of Conditioned responses
  –   CS (bell) and UCS (meat) must be paired
      several times. The CS must be presented just
      before the UCS for the association to be
      strongest.
  –   These pairings are called reinforced trials,
      because they reinforce the connection. Trials
      with only the CS are called unreinforced
      trials.
          The important parts of
          Classical Conditioning
    Measuring the strength of the conditioned
     response:
1.   Response amplitude, or strength
     –   E.g. amount of saliva
2.   Probability of response (when CS is
     presented alone)
3.   Response latency (how long a gap is there
     between the CS and the CR?
The important parts of
Classical Conditioning
     The learning Curve:
         The important parts of
         Classical Conditioning
 Second order conditioning
 Bell => salivating.
    – Fair enough, but…

    – Black square + bell => salivating
    –…
    – Black square => salivating
        The important parts of
        Classical Conditioning
2.   Extinction
    ‘Undoing’ the
     connections
    Important so an
     animal doesn’t get
     locked into
     behaviours that
     aren’t useful.
               The important parts of
               Classical Conditioning
3.    Generalisation
     In the real world, two
      stimuli are rarely the
      same
     –   E.g. voices at different
         intonations
     –   footsteps at a different
         pace
     The more different the
      stimulus from the original
      CS, the weaker the
      response
        The important parts of
        Classical Conditioning
4.   Discrimination
    To avoid ‘over-generalisation’
    It may take many trials for an animal to
     tell the difference between e.g. a black
     square and a grey square, but if only the
     black square is reinforced, it will
     eventually tell them apart perfectly.
           Extensions of classical
               conditioning
   Hunger
   Emotions
    – How we feel about certain people
        » Daddy + Disney world => happiness
        Hopefully … daddy => happiness
    – Phobias and their treatment
   Drugs e.g. insulin needles
   Drug addiction
        » Sight of needle => body’s compensatory response => tolerance
        » Danger of overdose if compensatory reaction is not elicited by
          a relevant CS such as location.
        » CS, (such as needle) but without the US (the drug) means the
          compensatory response just carries on. This causes cravings.
              Important point:
   In classical conditioning, psychologists
    thought that an animal learns to respond in a
    reflexive manner to some new stimulus.
    – There was no thinking involved
    – The animal was basically like a machine
    – It matched its old response to a new stimulus
      for no good reason.
    – Learning is a bit ‘stupid’
     Classical Conditioning Quiz
1.   What is the simplest form of learning called?
2.   Give an example of something humans don’t have to
     learn
3.   Who accidentally discovered classical conditioning?
4.   In the experiments, was the bell the conditioned stimulus
     or the unconditioned stimulus?
5.   To get the best effect, should the bell come before or
     after the meat?
6.   What is it called when an association ‘wears off’?
7.   What is it called when an animal cal tell the difference
     between two different bells?
8.   Why can taking drugs in an unusual place be dangerous?
                    Answers
1.   Habituation
2.   The rules of speaking their own language,
     jumping at a loud noise etc.
3.   Ivan Pavlov, Russian Physiologist
4.   The conditioned stimulus
5.   Before
6.   Extinction
7.   Discrimination
8.   The body doesn’t get the conditioned stimulus
     (the location) to prepare for the drugs
        Part 3

Instrumental Conditioning
                   What is it?

   When a seal does a
    trick and gets a fish, it
    learns an instrumental
    response – the trick is
    instrumental to
    getting the fish…
                   What is it?
Different from classical conditioning
 NOT an improvement or replacement of classical
    conditioning.
     – This time the reinforcement (e.g. food) depends
       on making the right response. Pavlov’s dogs
       could not influence when they were fed by
       making a particular response.
     – The response the animal makes has to be
       deliberately selected – Pavlov’s dogs had no
       choice whether or not they salivated.
     Instrumental Conditioning
 Thorndike and the Law
  of Effect (1898)
 Studied animals’
  behaviour to see if they
  showed signs of
  ‘human’ thinking and
  reasoning
Thorndike’s not-so-clever cats
Thorndike’s cats
            • The wrong responses
            led to failure, and the
            right response led to
            success. The wrong ones
            had to be ‘stamped out’,
            the right ones had to be
            ‘stamped in’

            • The cats did learn to
            make the right response,
            but it was a slow
            process. No evidence of
            thinking!
           The law of effect
 If a response is followed by a reward, it will
  be strengthened.
 If a response is not followed by a reward, or
  is followed by punishment, it will be
  weakened.
 The strength of a response is adjusted
  according to the response’s consequences
    Skinner and Operant Conditioning

 B.F. Skinner (1904-1990)
 Insisted Instrumental conditioning was
  different from classical conditioning as an
  animal’s response is basically voluntary
 He called instrumental responses operants
  because they operate on the environment.
 Developed better puzzle boxes
            The main features of
         Instrumental Conditioning
1.   Reinforcement
     –   In classical conditioning, reinforcement
         strengthens a response.
     –   In Instrumental conditioning, this is done by
         presenting some stimulus AFTER the correct
         response has been made
     –   Two ways of doing this…
     1(a) Presentation of
    ‘Appetitive’ Stimulus
Correct response   =>   Nice juicy fish
  1(b) Removal of aversive
         stimulus
     Shock + correct response = no more shock




Evil Psychologist                  Gigantic rat
               Reinforcement
   As in CC, the more reinforcements, the
    more likely the response will be to occur

   As in CC, the response will suffer
    extinction if it is not reinforced.
             2. Generalisation and
                Discrimination
     The responses might be voluntary, but
      outside stimuli can affect it, e.g.
      discrimination.




Pigeon poo




                Red light vs. Green light
            Generalisation and
             Discrimination
   The responses might be voluntary, but
    outside stimuli can affect it, e.g.
    discrimination.

   The difference is, in instrumental
    conditioning, the stimulus tells the animal
    what to do, not just what’s going to happen.
               Generalisation and
                Discrimination
   Generalisation is just the same: if the
    stimulus is quite similar to the training
    stimulus, the animal is quite likely to make
    the same response




    Training         Quite likely       Less likely
    stimulus         to respond         to respond
                        3. Shaping
   The way to train animals…




                           Kung fu hamster


Water-skiing squirrel                        ?
                         How?
   The method of successive approximations
    –   Little bit at a time
    –   Takes a looooonnnnngggggg time
    –   Requires perfect timing
    –   Someone trained a pig to turn on a radio, eat
        breakfast at a kitchen table, put dirty clothes in
        a wash basket, hoover the floor, select a certain
        type of pig food from lots of alternatives
                  And
 Dogs playing the piano
 Pigeons playing ping pong…
 Etc. etc. etc…
             4. Reinforcers

 Many different types (food, money, less
  pain etc.)
 Not just things that meet immediate needs.
                 Reinforcers
 It seems a reinforcer is anything that
  motivates behaviour
 Reinforcers will motivate behaviour
  differently depending on what other rewards
  are available.
     » If a kid gets no pocket money, £5 is a good
       reinforcer
     » If they get £30 pocket money, suddenly £5 is not so
       interesting.
                Reinforcers
   Some behaviours have their own reinforcers
    ‘built-in’ (intrinsic motivation)
    – Art
    – Sport
    – Sex
    5. Schedules of Reinforcement
 In the lab and in real life, reinforcement is
  usually partial
 Reinforcement usually comes in patterns or
  after a delay.
    – The rules for reinforcing are known as
      schedules of reinforcement
          5(a). Ratio Schedules
   Fixed ratio:
    – E.g. 10 responses for each reward
   Variable ratio:
    – Roughly 10 responses for a reward.
    – The uncertainty produces a high response rate
    – This is how gambling machines make money
        5(b). Interval Schedules
   Reinforcement only comes at a fixed time
    since the last one.
    – Responses slow down after a reward, and then
      start to pick up near the end of the 5 minute
      period. So you can train an animal to respond
      every 5 mins.
    – Example: you only check for mail in the
      morning, because that’s when the effort is
      rewarded.
            6. Punishment
 The opposite of reward
 Timing is crucial
 Escape and avoidance
 The point is to
  weaken responses.
           In the real world
 Table manners
 Prisons
 Token economies
 Mental health treatments
      Instrumental Conditioning
1.   Why do we use the term ‘instrumental’?
2.   Why do we use the term ‘operant’
3.   Are instrumental responses voluntary or reflexive?
4.   Does the reinforcer come before or after the response?
5.   Can an instrumental response become extinct?
6.   What is the correct term for the way you train animals?
7.   How do we define a reinforcer?
8.   Give an example of intrinsic motivation
9.   What reinforcement schedule do gambling machines
     use?
                      Answers
1.   The response is instrumental to getting some reward
2.   The response is a way of operating on the environment
3.   Voluntary
4.   After
5.   Yes
6.   Shaping, or Successive Approximations
7.   Anything that motivates behaviour
8.   Drinking
9.   Variable ratio
     Part 4

Cognitive Learning
             Cognitive learning
 ‘Cognitive’ = information processing
 Psychology today, in this university, is
  cognitive psychology.
 More concerned with what goes on in the
  mind than simple reflex actions.
   Earlier psychologists treated people almost like
    machines – they believed you couldn’t study
    mental processes because you couldn’t see them.
             Cognitive learning
 Learning cannot just be a change in
  behaviour – it involves new KNOWLEDGE
 The proof ?
    – Carry a rat round a maze and it will learn the
      layout of the maze
       » Without doing anything!
Cognitive views of classical conditioning.

   The conditioned stimulus (bell) doesn’t become a
    substitute for the unconditioned stimulus (meat).
    Often the CR and the UR can be different, as in
    the drug example.

    – The dog learns that the bell means the food is
      on it’s way, so it responds appropriately.
    – It learns the relationship between the two
      stimuli.
Cognitive views of classical conditioning.

 – It learns the relationship between the two
   stimuli.
                             Ah, I see…
                               Ah, I see…
    Why does conditioning work?
   Pavlov thought it was just because the the bell and
    the meat occurred close to each other.
   But we now understand in terms of signals: the
    dog hears the bell and knows it signals
    approaching food.
   Presenting the stimuli the other way around is
    poor at producing the response, and having them
    at the same time is pretty useless…
    Why does conditioning work?
 Think about road signs
 Signs have to come before the thing they
  warn us about. It’s no good having them at
  the same time, or after the thing itself!
    Why does conditioning work?
 The important thing is that the bell predicts
  the food, not just that they happen closely
  together in time.
 Storm clouds – we look at them and predict
  storms. We don’t immediately dive for
  cover.
    – Storm clouds are not a substitute for an actual
      storm; they serve as a signal
            The role of surprise
   The CS – US pairing has to be NEW, it has
    to make an animal sit up and pay attention.
    – If the dog was already expecting the food, it
      would ignore the bell because it doesn’t tell it
      anything new.
       » Animals tend not to make new connections when the
         old ones are good enough
       » Temperature example
    Cognitive view of Instrumental Conditioning

   An animal doesn’t just learn to perform a response – it
    learns the relationship between a response and it’s outcome
     – This is called an act-outcome representation



                                  **Deep see…
                                 Ah, I thoughts**
               Latent learning
   Rats know their way around a maze, but
    only bother demonstrating this when it gets
    them some food!
    – So what they have learned is knowledge, not
      just behaviour
Cognitive view of Instrumental Conditioning

   Animals can learn which action leads to
    which reward.
    –   Press lever => food
    –   Pull chain => sweet water
    –   …
    –   Poison the water…
    –   No more chain pulling
         » Points to quite complex knowledge
Cognitive view of Instrumental Conditioning

   Animals (and humans) know when there is
    a relationship between two events and when
    there isn’t
    – And they seem to prefer it when there is one!
Cognitive view of Instrumental Conditioning

   Learned
    helplessness
    – When we fail to
      learn the relationship
    – How to make a dog
      depressed
    Link to depression in humans
   People may be in situations when they
    really were helpless
    –   Losing job
    –   Losing a court case
    –   Someone dying
    –   Illness
         » They end up learning that there is no relationship
           between what they do, and the things that happen to
           them
                    Summary
   In the cognitive view, animals and people actually
    gain knowledge when they learn.
   They don’t just learn new reflex actions, or new
    complex behaviours. They don’t just act like
    machines.
   Learning about relationships is highly useful, but
    can also be damaging, as in the case of depression.
   Psychologists only work with dogs, pigeons and
    rats because they’re too scared to work with tigers
 The End.

Thanks for listening!

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:8/25/2011
language:English
pages:78