Sunday, August 4, 2013

Kathy Sdao Seminar: The R in Dog Training

To Kathy, the most essential thing to understand about dog training is that consequences drive behavior. Period, end of story. What happens after a behavior happens is the best predictor of whether or not that behavior happens again. There are other important things, of course, and in fact, Kathy has an acronym for them: “Get SMART,” which stands for See, Mark, and Reward/Reinforce Training. (There’s actually a second S- set up- which I’ll talk about in a separate post because there’s a lot of great material there to apply to reactivity.) But the most important is the “R,” so that’s what I’m going to write about today.

The camera caught me mid-reinforcement!
Let’s start with the difference between reinforcement and rewards. Although it might appear that she’s using the two words interchangeably, she’s not. They aren’t the same thing. Rewards are given to an animal; it’s something he earned. Rewards don’t necessarily affect behavior (although they can create good will and enthusiasm). On the other hand, behaviors are reinforced. Reinforcement both causes the behavior to be repeated or occur more often and are contingent on that behavior happening. Reinforcement, Kathy says, is the trainer’s responsibility, not the animal’s.

Obviously, the more reinforcers you have, the better, and the amount of things you can use as a reinforcer is really limited by your own creativity. Classical conditioning will allow you to create a reinforcer. Or, you can use things your dog is distracted by as a reinforcer (this is basically the Premack Principle, and it’s very potent). And, as I’ve discussed on this blog before, cues can also be reinforcing.

I’ve always found this last bit fascinating, if a bit confusing. The truth is, while it’s awesome that cues as reinforcers gives you a lot more options, there are some downsides. You have to put a bit of work into cues-as-reinforcers; the cue must be familiar and the behavior must be fluent. It also needs to have been taught with positive reinforcement only. And then, if you’ve been lucky enough to create reinforcing cues, you need to be careful. If you give them simultaneously with bad behavior- such as when we try to redirect a behavior we dislike- it can reinforce that bad behavior. Oops.

This isn’t the only way reinforcement can go wrong. Remember how Pavlov’s dogs were conditioned to feel happy when they heard a bell that was followed by food? This can happen to anything. So if there are two events that happen sequentially, the way the dog feels about the second event can go backwards in time and contaminate the first. Sometimes this is awesome; dogs learn to love their clickers because they’ve been followed by treats. Sometimes, not so much. Kathy told us that if you reinforce a dog immediately after you’ve punished him, that punishment will become a reinforcer.

Say what? But… yeah, it can happen. It’s just two events getting associated with one another. For example, if you yell at your dog and then immediately praise him for making a better choice, the dog can learn to anticipate being praised after you yell. Or if you give him a collar correction for pulling on leash and then click and treat for heeling, collar corrections can become an opportunity to earn food. If this happens, every time you try to punish your dog by yelling at him or using a collar correction, you’ll actually be reinforcing the behavior and therefore causing it to happen more often!

This also works the other way around. If something bad happens immediately after you’ve offered your dog a toy or some food, then the bad thing can contaminate the good one. This can create a dog that “isn’t food motivated”- not because he doesn’t like food, but rather, because he’s afraid of what it predicts. And this doesn’t have to be punishment. If you try to help a dog get over his fears by luring him into the situation (for example, luring him to you to get a nail trim or to step on a wobble board), you’ll actually make things worse.

But don’t let all this scare you away from using reinforcement! For one thing, even if you aren’t a clicker trainer, it is impossible to avoid (anything that increases a behavior is reinforcement). Instead, avoid the pitfalls by simply separating reinforcement (good things) and punishment (anything scary or bad) with a pause long enough that the dog doesn’t associate the two.

Okay, so you’re ready to reinforce behaviors. You know how to avoid poisoning your treats. So… how do you give them? Experienced trainers know that the way you deliver reinforcement influences the final behavior. Using a marker (like a clicker) will reduce the impact of food delivery because the marker says that’s the behavior. Even so, that marker becomes a sort of cue in itself: it tells the dog that he has earned his reinforcer and that he should go to the location it will likely be delivered. Don’t fool yourself into thinking that only a clicker will tell the dog this; my Maisy has discovered that praise or even just a smile from me means that she should look for her treat. This is why, whether you use a marker or not, the place you give the treat matters so much.

There are three main places to give the treat: in position (while lying down, in heel position, etc.), in order to set up the next repetition of the behavior (for example, tossing the treat away from the dog’s mat when teaching “go to bed”), or “direction sliding” (where you move the dog to the correct location in order to fix a problem such as forging in heel or to further the dog’s learning such as teaching a spin). The option you choose will depend on both the stage of learning your dog is in as well as your final goal. And you may even switch back and forth between locations!

So that’s the down and dirty on reinforcement, AKA, the most important part of dog training. What have you learned about reinforcement? Worse yet, what did you learn the hard way?


A to Z Dals said...

You wrote: "For example, if you yell at your dog and then immediately praise him for making a better choice, the dog can learn to anticipate being praised after you yell. Or if you give him a collar correction for pulling on leash and then click and treat for heeling, collar corrections can become an opportunity to earn food."

Long before I ever saw an explanation for this, I'd noticed it in clicker trained dogs. I told my students it was the "I'll-Be-Bad-So-I-Can-Be-Good-And-Get-Rewarded-Syndrome" Of course now with the explanation I know why it happens, but I'm still using the name because it sticks with my pet dog students better than "tertiary reinforcement".

Dog Training Los Angeles said...

If your puppy is feeling brave or aggressive, he'll try to make himself larger by standing tall, with his ears and tail sticking upright. He'll also push out his chest and raise the hair on his neck and back. He might also growl and wave his tail slowly.