Friday, April 5, 2013

Bad Dog?!

One of the biggest motivators for students to come in for training is a dog who is doing things we don’t like.  There are some basic ways that we can help dogs to learn not to do what we don’t want them to do, and most trainers are really good at these.  Once you understand what a trainer is driving at and why they have chosen the method they have, everything becomes much easier to follow along with.

The first thing I do when a client has a dog whose behaviour is a problem is define what the problem behaviour is.  The solution is going to depend upon what the problem is.  There is no single solution to every single problem, and the solutions I choose will depend upon the problems and the situations they occur within that are presented.

Controlling or managing the environment is often a great solution, and is usually the first thing I do in every case.  If you can control what is happening to cause the problem, then you can avoid the problem at least for a period of time.  If the problem is that the dog barks at the window, then the very first thing I am going to do is to get control over that window.  If the dog is confined to the room with the window and is barking six hours out of every eight, then why would I expect the dog to not bark at the window when I am trying to take a quiet moment?  The very first thing to do when looking at a behaviour we don’t want is to avoid the situation where the behaviour occurs. 

In some cases this is a forever solution.  If a client has a dog who jumps into the front seat of the car while she is driving, then my solution is going to be traveling in a crate.  Traveling in a crate is safer for the dog in any event, but beyond that fact is that the dog is safer in a crate when traveling, so that is a permanent solution.

Our puppy traveled to us in an airline crate and has traveled in a crate in vehicles for his whole life.

Some problems cannot be permanently solved by changing the environment for the dog.  Take jumping on people to greet.  Your dog can be kept separate from people for a period of time, but that is not a forever solution.  In this case I am going to choose to reinforce not jumping.  There is a great game you can play with jumping dogs where you have three people who come in and out of a room.  As they enter the room, as the dog approached but before he jumps, you drop a treat and then leave again.  The next person comes in and does the same thing.  You keep doing this until the dog gets the idea that the reward is going to come from below.  At that point you can switch the game up a bit and ask the dog to sit before you drop the treat.  Once the dog understands that everyone coming through the door is going to do the same thing, you can start working with a wider group of people and finally with strangers at the door.

Reinforcing for alternate behaviours (the learning theorists call this a DRA or DRO which stands for a differential reinforcement of alternate or other behaviour) is helpful if the dog is thinking about what he is doing, but not helpful if the dog is behaving badly because of fear.  Consider a dog who barks when approached by a stranger and tries to hide behind the owner.  This dog is afraid and he is not barking because he wants to greet, but because he is concerned about the stranger.  This is the kind of dog whose behaviour can be successfully changed by pairing one thing with another.  To do this, you get a stranger to appear and you give really good treats to the dog.  Then the stranger disappears and you stop treating.  The stranger reappear and you treat again.  The stranger disappears and you stop again.  You keep repeating this until the dog anticipates the treats when the stranger appears.  From there it is a matter of teaching the dog that the stranger can come closer and closer and treats will keep coming, but don’t let the stranger get so close that the dog is overwhelmed.

In our group classical conditioning class, the focus is on pairing the approach of a stranger with food to make a pleasant association while keeping all the dogs in the room below threshold.

This sort of pairing is called Classical Conditioning.  Classical conditioning is an effect originally noticed by Dr. Ivan Pavlov who was studying salivation in dogs.  He would ring a bell and someone would bring in a bowl of meat.  Dr. Pavlov noticed that the dogs began to salivate when the bell rang, before the meat appeared.  They formed an association between the bell and the meat.  Once you understand the effect, you can use classical conditioning in a huge variety of creative ways to teach your dog what is safe and what is dangerous.

Some behaviours are really persistent.  The dog isn’t frightened or concerned, and the dog might not be learning to do a different behaviour instead of the undesired behaviour, and you cannot avoid the problem.  In these cases, we sometimes have to pull out punishment.  Punishment is anything we do to decrease a behaviour.  Previously all the solutions were aimed at either avoiding the situation where the unwanted behaviour happened, changing the unwanted behaviour for a wanted behaviour, or changing the motivation for the unwanted behaviour.  Sometimes none of these alternatives are helpful.  In these cases we may need to consider using a punishment to change the unwanted behaviour.

Punishments come in two flavours.  It is important to understand both of these because one of them can cause more problems than the other.  The first kind of punishment is negative punishment.  That is the kind of punishment where you lose access to something that you want due to your behaviour.  When you use negative punishment the dog may lose the chance to play, or may lose the toy he was involved with or may lose access to the person or other dog he was interacting with.  For the jumping dog, you can sometimes change the situation completely by going into the bathroom every time the dog jumps on you.  You withdraw your presence every time the dog jumps up and pretty soon, if the dog cares about being with you, the jumping stops.  In most cases of negative punishment, the dog is not greatly upset or distressed.  

The other form of punishment is positive punishment.  This is when you DO something unpleasant to the dog when he does something you don’t like.  Positive punishment is the one place where we see the most problems in training and the greatest misuse.  You really cannot go wrong with management of the environment, DRO or classical conditioning.  Negative punishment is difficult to create trauma with, but with positive punishment there are huge risks.  If you are too tough, you can create fear.  If you are not tough enough, you can create tolerance of bad things.  If you give it at the wrong time, you can decrease behaviours you like and want.  If you don’t give it every time that the unwanted behaviour happens, you can teach your dog to gamble that he might get to do the unwanted behaviour this time.  If the positive punisher occurs in the absence of the bad behaviour, the dog won’t learn what not to do, but rather will learn not to try new behaviours that might be better.  There is also the risk that some classical conditioning will happen in reverse to what you thought might happen; so your dog might think that the positive punisher might be related to something that is actually safe. 

Taking the example of the dog who jumps up; you could use a can of compressed air to stop him from jumping up.  There are big risks with this though, and they aren’t really very obvious.  Let’s say that the dog jumps up on guests at the front door.  You come to the door prepared and the dog predictably jumps up.  You spray him with compressed air and he gets off you right away.  Your dog looks around and notices the spray can in your hand, and figures out that the spray can is the source of the air (this is an example of a classical association).  The dog learns an interesting rule.  “Don’t jump up if the person is holding a can of air.”  This means that the dog may learn to check if the guest has a can of air before he decides to jump up.  The dog hasn’t learned to do anything different, he has just learned that if the guest doesn’t have a can of compressed air, the guest is fair game. 

Then there is the situation where the can of air is not at hand, but the owner wants to punish the dog for jumping up, so after the dog has jumped up, she goes to get the can and spray the dog.  In this situation, the dog, not knowing what the owner is up to, follows the owner to where the spray can is kept and then gets punished for following the owner.  If the owner is the person who applies the air, the risk is that the dog learns that the owner isn’t safe.  This kind of damage happens all the time, and again this is classical conditioning at work; the dog has learned to associate the owner with danger.  Not the desired outcome by a long shot!  The dog may also learn that the rule about jumping up is to only jump up in the absence of the owner.  Again, we have an outcome that looks good, but isn’t quite what we had hoped for.

Finally a bad situation can occur when the dog experiences the same or a similar bad thing when he is behaving the way we want him to behave.  Consider the dog who is lying quietly on the grass by the driveway and someone comes out of the garage with an air compressor to fill the tired.  The whooshing noise that the compressor makes sounds like the compressed air can and we have caused the dog to stop lying on the grass and minding his own business.  We have taken the single event and spread it around to the environment and taught the dog a lot of things we wish he had not learned. 

Positive punishment can be a safe and effective tool, but it is tricky.  In order to use it safely, we need to follow some rules.  It must be tough enough to solve the problem quickly and efficiently.  If the positive punisher is not tough enough, you just create a situation where the dog learns to ignore the unpleasant consequence.  If the dog described above didn’t care about the air can, the jumping wouldn’t stop and he would just learn to tolerate the air can.  On the other hand, if it is too tough, that can cause both physical and emotional harm.  If you used a spray can of mace for instance, you might damage your dog’s eyes and it is quite likely he would be very frightened of spray cans, compressed air and possibly you.  There is a huge responsibility on the part of the trainer to chose the right degree of unpleasant consequence to get a good effect quickly and efficiently, but not so much that you harm or frighten the learner.

You also have a responsibility to use a punisher that is unique.  Compressed air canisters are being sold now as behaviour solutions in pet stores, but people forget that we use compressed air in a number of ways.  We use compressed air to clean our computers, we use it to fill our car tires and if you have someone in your home who uses medical oxygen, the hose might detach and hiss the way a can of air does.  Compressed air canisters are very handy for breaking up dog fights, but should not be used to teach dogs to stop doing specific behaviours.

The punishment must happen every time the behaviour happens and never happen if the behaviour is absent.  This means you cannot allow the undesired behaviour to happen unless you have the punisher at the ready and available.  You also cannot allow the punishment to happen when the behaviour hasn’t happened.  If the punishment is happening when it shouldn’t or isn’t happening when it should, the dog learns to gamble on the punishment, and every time he does the unwanted behaviour and doesn’t get punished, the absence of the punishment is in fact a reinforcer, and makes the undesired behaviour stronger.  This gambling is the route to building the strongest behaviours, which makes the mis-use of punishment potentially the best way to strengthen behaviours.

When used carefully and properly positive punishment is a fast and effective and humane way to stop unwanted behaviours.  When used improperly there are huge risks.  Why aren’t there such risks with managing the environment, DRO or classical conditioning?  Let’s look at that.  When you manage the environment you are avoiding the problem.  This means that you don’t see the problem behaviour, which just isn’t a problem!  When you use differential reinforcement of an other behaviour, then if the dog offers the unwanted behaviour, he doesn’t get anything he wants, but he does get what he wants when he does the other behaviour.  There is no risk to your relationship with the dog, to physical or emotional harm, or that he will gamble that the unwanted behaviour might be safe at a given time.  The only risk you have is that the dog may learn when rewards are available and when they are not and may try the unwanted behaviour when rewards are not available.  We can work around that by using a marker such as a click or yes, to tell the dog when he is making the choice we want him to make and then using that marker to bridge the time between the behaviour and going to get the treat.  Finally with classical conditioning, there is no risk that the dog will get injured or frightened at all; we are just treating him in the presence of a particular stimulus.  The worst thing that will happen is that he will learn that a particular stimulus will produce food.

With so many options available for training, it is important that we keep in mind the old medical adage of “first do no harm” and reach for the least harmful option first.  I would never want a doctor to not learn to do surgery, and I would not want a professional trainer who didn’t understand the use of positive punishment and how it works.  Never the less if your doctor only knows surgery, you may not get the best health care.  The same is true of trainers who only use punishment; they can cause more harm than good.  Before you reach for a positive punisher, think carefully.  Is there a better alternative?  There are times when there isn’t a better alternative, but often there is, and it is worth considering what your alternatives are and develop a training plan before you reach for something that may cause harm.

No comments:

Post a Comment