Punishment or Not

Document Sample
Punishment or Not Powered By Docstoc
					Punishment or Not?

Alternatives to Punishment
                Defining Punishment
• Traditional approach:
    – definition: application or withdrawal of a stimulus to DECREASE behavior
        • negative punishment: withdrawal, removal of S+
        • positive punishment: application of S+
    – Skinner believed that punishment ineffective
        treatment, particularly w/humans
    – evidence is that should use only in certain situations

• Punishment first really studied by Thorndike:
    – the "negative law of effect": "when a response occurs in a situation, the
      response is followed by an annoying state of affairs, then the connection
      between the situation and the response is weakened, so that the next time
      the situation occurs, the response is less likely to occur."

• Skinner said same thing: any R followed by a P is less likely to occur:
  Punishment reduces behavior
                             Thorndike study
•   First experimental evidence: Thorndike's Spanish Vocabulary Test
     –   presented subjects w/Spanish word and 4 English words
     –   task: choose correct English definition of Spanish word
     –   after every response, S told either "good" or "bad"
     –   good = supposedly satisfying state of affairs
     –   bad = supposedly annoying state of affairs
     –   Results:
           •   responses followed by bad NOT decrease to 0
           •   lessened in probability to about 0.2

•   Thorndike's interpretation: Punishment not work

•   Problem with the experiment:
     –   no control group
     –   not sure "bad" is punisher
     –   if say nothing after a trial, higher probability of responding than when say "bad"
     –   thus: "bad" was somewhat punishing
                  Skinner Experiment
• Skinner and the bar-slap experiment:
    – trained rats to b-press for food
    – instituted 2 conditions
        • bar pressing put on extinction
        • bar pressing first EXT plus if pressed bar it sprang back up and slapped the rat

• found: much lower response rates when animal both on EXT and
  punishment than when animal just on EXT

•    when later taken off punishment and switched to EXT alone,
    responding increased

•    Skinner used this as evidence that punishment doesn't work: only
    that punishment temporarily suppresses responding but not get rid
    of responding
Problems with Skinner’s Interpretation

• Bar slap not that strong of punisher:
   – If had used stronger punisher (shock), less increase in
     responding during EXT after EXT+PUN

• Skinner misinterpreted his own results and his
  own theory:
   – reinforcers are no longer effective if taken away
   – behavior decreases when reinforcers withdrawn
   – why shouldn't withdrawal of punishment result in
     increase in behavior?
         Parameters of Punishment
•   Punishers have same parameters as reinforcement:
•   Acquisition:
•   Extinction:
•   important characteristics of extinction
•   Rate of punishment:
•   Size of the punisher:
•   Delay of punishment
•   Quality of punishment:
•   Percent of punishers that are contingent upon responding:
•   Context of punishment:
•   Generalization and Discrimination:
               Theories of Punishment
• Non-contingent theories:
     – if the punisher automatically elicits a reaction that interferes with the
       response- the R then decreases in frequency
     – incompatible (skeletal) responses:
          • punishers act on behavior by eliciting a skeletal response which interferes with the
            performance of the punished response

• Fowler: attempted to demonstrate this:2 groups of rats
     – GRP 1: trained to run down alleyway to get food, then began shocking rear
       paws when started running
     – GRP 2: same thing, only showed front paws
     – results: rats w/rear paws shocked ran faster than rats w/front paws shocked

•   interpretation: shock elicited incompatible response that interfered
• incompatible emotional response theory:
        – punishers act on behavior by eliciting an emotional response which
          then interferes w/the performance of the punished response

•       Estes: Conditioned Emotional Response:
            CS (light)---> UC (shock) -> UR(fear)
                       CR (fear)
    •    Estes: via Class. Cond., the light should be a S+ of fear, and the
         animal will stop Ring when the light comes on, will start again
         when the light goes off
     Problems with these theories
• interfering skeletal responses are rarely defined and

• fear is hard to define and measure

• the theories cannot explain why punishers that are
  contingent upon responding suppress behavior more than
  punishers that are non-contingent

• Alternative: Contingent theories:
   – punishers are only effective in suppressing behavior when they
     are contingent upon behavior
                      Avoidance Theories
•   Punished responses decrease because
     –   all responses other than the punished response increase,
     –   That is, the organism avoids the punisher
     –   R (driving)---> Punisher (Accident): R decreases
     –   R (walking)---> avoid accident: R increases
     –   R (biking) ---> avoid accident: R increases

•   problem: responses that increase in frequency are not defined and measured

     – whatever the reason is that responses increase when they are followed by a reinforcer is the
       same reason for why responses decrease when followed by a punisher
     – problem: really just passing the buck
     – however: evidence that reinforcers and punishers govern behavior in highly similar ways
Guidelines for positive punishment
• Behavior must be dangerous to person or
• No chance to interrupt and reinforce “good”
• Tried other alternatives

• Typically must go through an ethics board to
  use (for humans OR animals)
•   Technically, time-out from reinforcement opportunities

•   Not so much a punishment, as not reinforcement

•   Works as long as no access to reinforcers during this period

•   Must be careful of escape/avoidance
     – Hate to color, so act up in art class
     – Get sent out for time out
     – Really getting reinforced for acting up

•   Rules for Using Time-Out
     – 1 minute per year of age
     – Must be quiet to get the timer to start
     – Cannot use for dangerous, disruptive or self-stimulatory behavior
     – Must really be “time out” from other rewards
  Alternatives to Positive Punishment:
         Negative Punishment
• Response cost: your response costs you something or
  some behavior

• Two parts:
   – Restitution: reinstatement of environment (clean up)
   – Positive practice: practice better response for situation

• Can also use satiation/habituation
   – But YOU must be willing to wait out organism
   – Their level of satiation/habituation may be higher than
        Differential reinforcement
• Two kinds of differential reinforcement.
   – DR used to decrease inappropriate behavior by ignoring it and
     providing reinforcement for appropriate behavior.

   – DR used to bring behavior under the control of a specific

• The word "differential" means that
   – the animal is taught to differentiate between positive and
     negative behaviors
   – learns that specific behaviors will or will not be reinforced
   – that behaviors are appropriate only when exhibited in certain
     situations, i.e., after certain discriminative stimuli.
          Differential Reinforcement of
              Other Behaviors (DRO)
• DRO: reinforcement delivered when the targeted inappropriate behavior is
  not exhibited.
    – For example, if the targeted behavior is interrupting, then reinforce for NOT
      interrupting, even if other inappropriate behaviors are occurring
    – That is, reinforce anything BUT interrupting

• Because only a single behavior, and not other behaviors, are not being
    – can be used to reinforce an organism after a specific interval of time during
      which a targeted inappropriate behavior was not exhibited.
    – As long as don’t exhibit behavior X for time Y, get the reinforcer

• A DRO-reset schedule resets when:
    – Immediately after the organism exhibits the targeted behavior.
    – Or the interval is completed
      Differential Reinforcement of
       Alternative Behavior (DRA)
• DRA: reinforcement of any behavior which makes
  it impossible for the targeted inappropriate
  behavior to occur.

• E.g., Dog jumping up when greeting people:
  – Dog is rewarded when doing anything BUT jumping up
  – Good for quickly getting rid of a single behavior
  – Side effect: may inadvertently shape up new “bad”
  Differential Reinforcement of Lower
         Rates of Behavior (DRL)
• DRL: reinforcement of a behavior when exhibited at a lower
• Now reinforce the target behavior
    – Not the behavior that is inappropriate.
    – Is the rate or frequency that is problematic

• Inappropriate for student to ask to use the restroom every fifteen
    – not a behavior you would wish to cease
    – Instead, DRL is used to reinforce the

• Give X number of bathroom passes per day; after that is a
  consequence (e.g., lose a privilege).
 Differential Reinforcement of Higher
       Rates of Behavior (DRH)

• DRH: reinforcement of a desired behavior as its
  occurrence increases.
  – Targeted behavior IS reinforced
  – Want to increase rate or frequency of this behavior

• Increase pro-social comments such as "please"
  and "thank you“
  – Occurs at low rate
  – With DRH, reinforce the individual for using these
    targeted words
     How do I decide which schedule of
reinforcement or punishment to implement?

 • Must look at each individual abilities and behavioral
    –   Want to increase or decrease target behavior
    –   What is severity of the target behavior
    –   Can you wait out response?
    –   Are other behaviors available to reinforce?

 • Is the behavior fluent or not?

 • Do you want other behaviors to replace it, or to change the
   rate of THAT behavior

Shared By: