The Learned Dog Class 10: Aversive control of behavior and putting OC & PC into practice.

The Learned Dog

Class 10: Aversive control of behavior and putting OC & PC into practice

Review...

• Pavlovian Conditioning or Operant Conditioning)?

• Associate sound of clicker with food?

• Dog receives a cookie every time he sits

• Handler stops every time the dog pulls on the leash

• Dog is let out every time he scratches and whines at the door

• Dog comes running when he hears jingle of leash

Review...

• Pavlovian Conditioning or Operant Conditioning)?

• Squirrel heads for tree when he hears sound of sliding door

• Bruce jerks on poor Scuppers’ leash every time a dog barks at Scuppers in class

• Scuppers already sits in response to gesture. Bruce starts saying ‘Sit’ right before he gives the sit gesture

• Scuppers starts getting all excited when we pull into Gemini parking lot

• Scuppers lies down & stares when he wants to go out

Review: its all about associations & prediction

• If the association is between performing an action and some subsequent consequence (good or bad), think Operant Conditioning.

• Dog predicts if they do the action, they will get a good thing or avoid a bad thing

• If the association is between 2 events regardless of the animal’s response to the first event, think Pavlovian Conditioning

• The dog predicts that if they observe the first event, the second event is likely to follow.

Review: what associations get made?

• Those that are easy to make.

• Species-specific associative biases, most common of which is the bias to form associations between things that are close in time and space.

• Those that provide incremental predictive value

• The question to ask: does learning the association & acting on it make the animal any better off?

Review: incremental predictive value...

• Bruce teaches Scuppers to lie-down by saying “down” and pointing down with his hand at the same time. He does this reliably until he is sure Scuppers knows the commands. What will happen if...

• He tries saying ‘sit’ without the ‘down-gesture’ and vice-versa? What are the possibilities and what is likely and why? Is this an example of overshadowing or blocking?

• Bruce teaches Scuppers to spin to the left in response to a hand movement. Once he is sure Scuppers knows this trick, he says “left” at the same time as he does the hand movement. Do you think Scuppers will learn to respond to the verbal cue by itself?

• Why or why not? Is this an example of overshadowing or blocking?

Review: incremental predictive value

• Overshadowing

• When two cues are presented together, without previous experience with either separately, and they predict the same outcome.

• Animal will choose one, and ignore the other. Which they choose will often reflect species-bias:

• dogs probably biased to attend to gesture vs. voice.

• Other is ignored because it is redundant.

Review: incremental predictive value

• Blocking

• Animal learns that one cue reliably predicts an outcome, then another cue is presented at same time as the first cue, and the outcome is the same.

• Animal will ignore new cue. Why? Because the new cue doesn’t tell it anything it didn’t know already.

• How do you prevent blocking?

• Present new cue before existing cue so that it predicts existing cue.

• And/Or vary outcome when new cue is present

Review: role of surprise...

• Surprise is the mother of all learning...

• Learning can be thought of as the process of minimizing the difference between ‘what happened’ and ‘what I expected to happen’. Learning is fastest when...

• Difference is big

• Difference is reliable

• ‘What happened’ matters

• Nature biases animals to learn about bad ‘what happened’ very quickly.

Review: role of surprise

• If you want your dog to learn faster should you...

• Use roast beef all the time

• Use roast beef some of the time, chicken some of the time, kibble some of the time, string cheese some of the time.

• Show your dog what to do & reward all the time

• Make sure your dog never has the opportunity to fail

Review: learned helplessness

• The 3 key questions for a ‘healthy’ animal...

• Is the world a good place or a bad place?

• How good am I at avoiding surprises?

• Is my world a predictable place?

• How good am I at getting things I want and avoiding things I don’t?

• Is my world a controllable place?


• Learned helplessness happens when significant things happen to the animal and the animal does not learn how to control the occurrence of those things...

• Consequences don’t seem tied to actions, so the animal acts as if they stop trying to make the connection between actions and consequences in our situations as well.

• Happens most often when the consequences are aversive, but can happen as well if the consequences (good or bad) don’t vary based on the action.


• Bruce gives Scuppers a treat if he does the jump quickly

• Bruce thinks he is rewarding good performance

• Bruce gives Scuppers a treat if he does the jump tentatively

• Bruce thinks he is giving encouragement

• Bruce gives Scuppers a treat when he is sitting before the jump and when Scuppers just sits there staring at the jump

• Bruce thinks he is giving encouragement

• What is Scuppers learning?

Review: the quadrant...

Add Take-away

Encourage behavior

Positive reinforcemen

t

Negative reinforcemen

t

Discourage behavior

Positive punishment

Negative punishment

Review: the quadrant, what is what?

• What behavior is getting encouraged/discouraged, what is getting added or removed, does it stay the same during the whole protocol?

• Bruce gives Scuppers a treat for walking on a loose leash by his side?

• Bruce stops in his tracks when the leash gets tight and waits until Scuppers circles back and is by his side?

• Bruce turns around and heads in the opposite direction when Scuppers pulls to catch up to a dog ahead of him, and Scuppers is dragged along at the end of the leash?

Review: advanced analysis & critique, what is scuppers learning in each case and why?

• Scuppers sees a strange dog, bruce throws food down on ground, Scuppers eats food...

• Scuppers sees a strange dog, alerts & orients, bruce throws food down on ground & Scuppers eats food...

• Scuppers sees, alerts& orients, barks like a knucklehead, bruce throws food down on ground

• Scuppers sees dog, bruce says “sit” and Scuppers sits, bruce throws food down on ground

• Scuppers sees dog, bruce says “sit” and Scuppers sits, bruce waits & says “ok” then throws food down on ground

Aversive Control

Difference between escape & avoidance

Avoidance behavior relies on responding to cue that predicts future aversive stimuli

çAversive StimuliCue

Time

If dog responds

before this point it is

“avoiding”

If dog responds

before this point it is

“escaping”

Punishment vs. aversive stimuli

• A big question for the animal: is the bad thing that just happened contingent on something I did, or is it independent of my actions...

• If it is contingent, then it is punishment in addition to being aversive

• If it is independent, then it is ‘just’ an aversive stimuli

• In either case, the presence of an aversive stimuli suppresses on-going behavior to a greater or lessor extent...

• The greater the degree to which the aversive stimuli is reliably tied to the performance of a given behavior, the greater degree to which that behavior will be suppressed.

Ignoring side-effects for a minute, punishment is most effective when...

• The punishment is intense, and introduced at its maximum intensity.

• There is little or no delay between the behavior to be suppressed and the punishment.

• The punishment is certain

• The prior reinforcement for the punished behavior is “reduced, eliminated, or made available only when an alternative behavior is performed”

• When it works it can be very fast, BUT...

Those pesky side-effects...

• It requires dispassion and technique, and most of us have neither when it comes to our dogs.

• The greater the delay between behavior & punishment, the less likely it is that the animal will treat it as punishment for the behavior, and the more likely it is that it will treat it as an aversive stimuli predicted by some CS in the environment...

• And that CS could just be YOU!

• And if not you, then something else in the environment

• When an aversive stimuli is delivered in such a way that the animal can’t & doesn’t learn from the experience, it stops being a lesson and starts being something else.

Those pesky side-effects

• When effective, punishment creates a behavioral vacuum. And vacuums always get filled.

• No assurance that the vacuum will get filled by a more desirable behavior, especially since punishment does not affect the underlying emotional state of the animal.

• You may get what you wish for, but not what you want...

• Supressing the growl, but not the bite.

Avoidance Learning

The tricky question...

• In avoidance learning, the animal responds to a cue that predicts something aversive, and by doing so avoids the aversive stimuli...

• What maintains the behavior, since the reinforcement is the absence of something?

• Several theories, but we will focus on just two of them...

SSDRs determine ‘early’ response to a threat...

• Animals respond to dangerous events, or predictions of dangerous events by a set of species-specific defense reactions...

• Freeze,

• Flight,

• Fight

• Form, order and frequency varies

• The closer a desired response is to an innate SSDR, the easier it will be for the animal to learn it. SWR’s example of pigeon flapping wing vs. rat standing on head.

Cognitive theory explains why avoidance behaviors persist...

• Basic tenets of Seligman & Johnson theory

• CS precedes shock

• Animal prefers ‘no shock’ to ‘shock’

• Animal forms expectations that are strengthened when confirmed and weakened when disconfirmed.

• Respond to cue -> no shock

• Don’t respond to cue -> shock

Cognitive theory explains why avoidance behaviors persist... (cont)

• Basic tenets of Seligman & Johnson theory

• Probability of an avoidance response increases as animal becomes more confident of its expectations...

• Animal undergoes fear conditioning to CS associated with the shock

• Fear is extinguished when CS is not paired with the shock


• The entire picture...

• Pavlovian conditioning causes CS to become associated with the aversive stimuli.

• Animal performs SSDR biased behaviors and variants and forms expectations as to their efficacy with respect to avoiding the aversive stimuli.

• Self-confirming & conservative. Every time animal performs response and avoids aversive, become more confident in expectation so more likely to do it. They don’t experiment (geez, I wonder what would happen if I don’t respond...). This is why they persist in the absence of feedback


• The entire picture...

• Only if they are forced into a situation that disconfirms their expectations will they alter their behavior...

• For example, prevented from performing avoidance behavior and do not experience aversive...

Avoidance learning is key!!!!

• A cognitive explanation is the easiest explanation of what one sees, and that is cool...

• Animals form expectations associated with the consequences of their actions and use these expectations to guide their behavior. And that is a big deal.

The Learned Dog – class 10

Dog Training & Operant Conditioning


Dog Training Methods – OC review

reinforcement increases behavior

punishment decreases behavior


Dog Training Methods – Positive Reinforcement

Lure – Reward

Capture

Shape

Target

Combinations of above


Dog Training Methods – Other

Model

Negative Reinforcement

Negative Punishment

Positive Punishment



Lure – Reward Use an object of attraction the dog wants that

becomes the reward Place the lure where you want the dog Give lure to dog when behavior is accomplished



Lure – Reward Pro’s: Easy to get basic behaviors started if dog is

interested in the lure Quick results early on



Lure – Reward Con’s: Easy to get hung up and stuck on the lure Can increase fear or create mistrust if used in

fearful situations Dog afraid of the lure ? what is learned or more accurately how quickly is it

learned with a lure

example: car/ directions



LURE-REWARD VIDEO

Kodi - Spin, Sit, Down, Stand

Kodi – Feet on box



Capture – simply + reinforce (click/treat) when the dog does a behavior in its repertoire

Example – set it up to teach down Bring dog in a small room such as the bathroom with a clicker and treats, and a

book if you want!Wait (do absolutely nothing) and be

ready to click the dog lying down, and treatToss a treat when you release with ok to get dog up again to start over



Capture –

More exotic examples:

Sneeze

Play Bow (stretch)

Tail Chase



Capture Pro’s:

You start with a completed behavior the dog does readily

Easy – just takes being ready



Capture Con’s:

Limited to behaviors offered regularly or that are easy to elicit the full behavior like a down in a small room



Shaping + Reinforcing small pieces of a behavior in successive

approximations “Shaping” the behavior into what you want it to look

like Break a behavior into pieces, reward the pieces and

then put them together Simple to complex behaviors can be shaped



Shaping Hot and Cold game, ok so I am dating myself…

Reinforce when getting closer (hot) and ignore when not on track (cold)

Keep a high rate of reinforcement to keep the behavior on track

Place reinforcement to give you maximum benefit Remember those chickens!



Shaping Pro’s

Engages the animal – they control the consequence Best method for fearful situations; dog stays in control

of choices – much quicker and sometimes the only way to get over a fear

Can be a great way to tire a dog’s mind and body will follow



Shaping Con’s

Requires some mechanical skill practice Getting a behavior may be slower in beginning but quicker

in the end Less behavior repair later

Smaller slices of behavior when learning More reinforcement history along the way



Target – teaching dog to go to and touch an object

Can use multiple objects hand yogurt lids target stick (dowels work well) almost anything can be a target



Target Pro’s:

Fun easy behavior

Eliminates need for lures early on

Use the target to teach other behaviors

Target Con’s:

Need to fade (pretty minor)



SHAPING and TARGETING VIDEO

Kodi - Shape nose touch lid target and target stick

Steiff – Shape feet on the box

All first sessions


Dog Training Methods - Positive Reinforcement

Combinations of above

Examples: Use lures to get shaping started Use targets to teach other behaviors

Great for distance

Example: hearing dog client go to door


Dog Training Methods – Model

A blast from the past… This was a step towards more positive training in

the 80’s

Basically placing dog in position and praise

Example: SitPlace one hand on dogs chest or collar and other hand

behind back legs above the back of the knees and fold into a sit. Now demo this with a dog you don’t know that might want to bite you. Just kidding! But that was a typical demo.


Dog Training Methods – Model

Model Pro’s: Better than punishment

Model Con’s: Slow learning It was punishment to some dogs Potentially dangerous to trainer


Dog Training Methods - Negative Reinforcement

Remove an unpleasant stimulus to increase behavior being repeated

Often positive punishment as well or at least an aversive stimulus

Have to add it to stop it

Examples: Stop collar pressure when dog sits Stop pushing on hindquarters when dog sits


Dog Training Methods – Negative ReinforcementNegative Reinforcement Pro’s:

May get some behaviors happening quickly If dog will not or cannot eat treats Teaching self control

Negative Reinforcement Con’s: Dog associating something unpleasant with

trainer and or situation Physical damage to dog depending on technique Requires excellent timing


Dog Training Methods – Negative Reinforcement

CAT: Constructional Aggression TreatmentA method developed by Jesus Rosales-Ruis & Kellie Snyder

Uses Negative Reinforcement

To shape calm behaviors to replace aggressive displays

towards dogs and humans

Seminars and a 10 hr DVD is available

Very new so all the pro’s and con’s are not yet known

Many use this with R+ (more on this when I talk on reactive dogs)


Dog Training Methods – Negative Punishment

Remove a pleasant stimulus to decrease a behavior being repeated

Examples: Puppy biting hands: leave puppy Jumping up: turn back ignore dog

Advanced training examples: Dog breaks a start line stay for agility: dog does not get

to run the course (assuming lots of stay work)


Dog Training Methods – Negative PunishmentNegative Punishment Pro’s

A benign way to stop unwanted behavior Generally not harmful physically or mentally

Negative Punishment Con’s Must be consistent otherwise you may be rewarding the unwanted behavior sometimes which is unfair


Dog Training Methods – Positive PunishmentAdding an unpleasant stimulus directed at the dog to decrease/stop a behavior

Examples: Pulling or jerking on the leash for pulling

Choke Collars, Pinch Collars Electronic collars, spray or shock

Running away/not coming, barking, fence Spray bottles for barking


Dog Training Methods – Positive PunishmentPositive Punishment Pro’s:

Stop unwanted behavior quickly at least short termPositive Punishment Con’s:

Dog associating something unpleasant with trainer and or situation

May increase or create fear and or aggression Physical damage to dog depending on the technique Learned helplessness Excellent timing required


Dog Training Methods – Summary

Positive Reinforcement is least damaging

Negative Punishment a safe addition

Negative Reinforcement occasionally useful

Positive Punishment mostly avoid

When training behaviors I use mostly PR some NR

When stopping unwanted behaviors I use NP when

possible and PP if needed.

When I use PP it is usually citronella spray

The Learned Dog Class 10: Aversive control of behavior and putting OC & PC into practice.

Documents

class scuppers

sure scuppers

existing cue

sit gesture scuppers

poor scuppers

leash dog

verbal cue

door dog