Don't shoot the dog book notes

Reinformances: Better than Rewards

Reinformance vs. reward

  • Reinforce increases the chance of something happening/not happening again.

Timing reinforcers

  • Too late and too early train unexpected behaviour

Scheduling reinforcers

  • Keep interest with 20 / session, 80 / day;
  • Small as allowable

Jackpots

  • Oppressions/resentment/sullenness may be overcome with large (10x) suprize positive reinforcer

Conditioned Reinforcers

  • A signal deliberatly presented before of during delivery of a primary reinforcer
  • Separates & pinpoints timing of an action even when giving primary reinforcer is impossible
  • Shouldn't be used unless training
  • Train it by pairing it with out reinforcers
  • Conditioned negative reinforcer should be a warning prior to punishedment, not during.

Reinforcement Schedules

  • Constant reinformce is while learning
  • Switch to variable reinforcement to maintain a behaviour
  • Gradually move to longer variable reinforcement
  • Once on variable reinforcement, the behaviour can be shaped with variable reinforcement

Exceptions to Variable Reinforcement

  • Puzzle/tests always need feedback to confirm correct solutions

Long reinforcement schedules

  • Variable instead of fixed schedules are best.
  • Too long may lead to extinction. (Metabolic boundary: reward is not worth the effort)
  • Long reinforcement schedule can lead to slow start, like procrastination, before starting the first action
  • A slow start can be overcome with a reinforcer for starting

Reinforce yourself

  • People are vastly under-reinforced for our behaviour

Shaping: Developing Super Performance without Strain or Pain

Shaping
Teach successive approximations to desied behaviour.
(no term)
Establish imtermediate goals. starting with a behaviour that already occurs sometimes

Ten Laws of shaping

  1. Raise criteria in encrements small enough that the subject always has a realistic chance for reinforcement
  2. Train one aspect of a behaviour at a time; don't try to shape 2 criteria simultaneously
  3. Put the current response on a veriable schedule of reinforement before adding or raising criteria
  4. When When introducing new criterion, temporarily relax that old one
  5. Plan ahead of your subject, so you're prepared
  6. Stick to one shaper per behaviour
  7. If one shaping procedure is not eliciting progress, find another
  8. Don't interrupt a training session gratuitously; that constitutes punishment
  9. If behaviour deteriorates, "go back to kindergardge"; quickly review the whole process with a series of easily earned reinforcers
  10. End each session on a high note, if possible; but in any case quit qile you're ahead

Shortcuts: Targeting, Mimicry, and Modiling

  • Mimicry: kids can learn through watching and mimicry, dogs are bad, cats are good
  • Targeting: Teach an animal to touch it's nose to a target. Then you can elicit lots of other behaviour by moving the target appropriatly
  • Modeling: Physically putting the subject through the motions. Needs to have shaping included to shape the skill and fade the modeling

Special Subjects (yourself)

  • Record keeping is a useful self-reinforcer. (Streaks)

Shaping without words

  • Formally people will accept instructions then shaping.
  • Informally, people resent being shaped. Especially with behavours they are currenlty endorsing through emotions
  • Don't talk about reinforcement unless tou have the subject's permission
  • If you talk about it, you are bribing them and they learn to take actions for rewards. Not learning the impusle preverberally.
  • Don't brag about it

Stimulus Control: Cooperation without Coercion

  • Triggers/commands are just a stimuli/cue
  • Once a behaviour is trained, reinforce it only when it's cued
  • Introduce a cue, the cue as the behaviour is starting & reinforce completing, alternate between cue and no-cue, reinforcing only cued behaviour, then shape the behaviour
  • Once your learning understands the rules, new cues can be attached to new behaviours practically instantly this way.

Rules of Stimulus control

  • Four conditions to perfect stimulus control
    1. Behaviour always occurs immediatly after the stimulus.
    2. Behaviour never occurs, intraining or work, without the stimulus.
    3. The behaviour never occurs on response to a different stimulus
    4. No other behaviours occur in reposnse to this stimulus
  • Pick signlas that can be easily perceived
  • Possible for, multiple signals for the same stimulus, and multiple behaviours for the same stimulus
  • Nonprimary signals just need to have enough magnitude to be noticed
  • Shouting isn't useful unless it's intentional unpleasant aversion
  • Once signal is trained you can fade it until it's barely perceptible, and still get the same response. Like a counting horse
  • Speed of response can be trained, but a steady response-time criterion should be used.
  • Behaviour chains should be trained backwards. Give > return > pickup > Go go > Sit and wait for frizbee to be thrown

Untraining: Using Reinforcement to het reid of behaviour you don't want

Eight solutions/methods to get rid of behaviour

  1. Shoot the dog
  2. Punishment
  3. Negative reinforcement
  4. Extintion
  5. Training incompatible behaviour
  6. Put the behaviour on cue.
  7. Shape the absence - train the oppsite action
  8. Change the motivation - remove the cause
  • Some behaviours have multiple causes
  • Some behaviours require multiple solutions