Wednesday, January 31, 2018

Setting Up for Success with Shaping

Kaylee is not a pit bull. Or a bulldog.

This might be a shock to you, but she's not quite like any of my other dogs. This is less because she is deaf, and more because of who she is as an individual. She's a lot "softer" - which I realize is not necessarily a useful label. What I mean by "softer" is that she is much more sensitive to errors in training and less resilient to failure. While the bull and terriers are most likely to give me the metaphorical finger and move on to the next adventure when one of us makes a mistake, Kaylee struggles. When she feels she has made a mistake, she is likely to disengage, usually by wandering away and sniffing.

I have the privileged of not being a cross over training. I have always used the positive reinforcement philosophies when training my dogs, which means that I have never made a big deal out of training errors. I don't actively punish the dogs when they make a mistake. What does happen when there are training errors is that my rate of reinforcement goes down; I don't reward if I don't get the behavior I'm looking for. Kaylee is particularly sensitive this, and I lose her if she doesn't get rewarded when she's trying very hard to be right. This is particularly evident when we try to shape behaviors. Shaping, by definition, includes being wrong - that's how you get to right.

Now, if we go back to that positive reinforcement philosophy, you will hear trainers say that you need to "reward for engagement." I think this is about as useful as saying a dog is "soft." What does engagement look like? How do I reward it? And how do I apply that to a shaping session?

So here are some of the techniques I use to keep Kaylee focused on me, learning, and not wandering away and sniffing during our shaping sessions:

I break down behaviors to their smallest possible criteria. In R+, we say, "be a splitter, not a lumper." What this means is that we want to split behavior down to the smallest possible step. So for example, if the final behavior I am looking for is "pupper runs to a target and remains standing on target until released," I'm not going to start shaping with "runs to target" or "stands on target." I'm going to start with "looks at target" as the smallest first step I can think of and work my way up from there.


I set up the session to make the behavior I am looking for extremely likely to happen. Most people are introduced to shaping with a 101 Things to Do with a Box style game: you have an object, a dog, and a clicker, and you mark the dog for any interactions with with the object until you have a behavior you like. There's nothing really wrong with this game, per se, but it is the purest and hardest form of shaping you can do, particularly if you have a traditionally trained dog who is disinclined to offer behaviors or a dog who has never worked with an object before. You're going to be sitting in front of that box staring at each other for a loooong time! Instead - if my criteria is "look at the object" - I'm going to set Kaylee up so that behavior is extremely likely to happen. In this example, I'm going to pick the object up and wave it around. Dogs naturally look toward movement, Kaylee looks at the object, and BANG - we have a place to start.

I keep shaping sessions short. Really short. Ten treats short. I will admit, this is the hardest rule for me to follow. My several of my other dogs are amazingly resilient and will keep working with me for as long as I want to run the session - but Kaylee doesn't let me get away with those kinds of shenanigans. To keep our sessions short, I count out ten treats in my hand, and when the treats are gone the session is over, no matter how well or poorly it is going. In fact, it's even more important to stop at ten treats when things aren't going well, and I'm using a high proportion of me treats to reset Kaylee - I need to stop and figure out why I'm not getting the behavior I'm looking for (hint: it's probably because I'm lumping multiple criteria together instead of splitting them down).

I never let Kaylee make a mistake twice in a row. If I had to point to one of these tips and say, "This is the most important piece of keeping Kaylee working with me during shaping sessions," it would be this one (followed very closely by "keep sessions short"). In order to keep Kaylee happy and working, I need to keep my rate of reinforcement high. In order to keep my rate of reinforcement up, I need to get reinforceable behaviors.

Okay, what the hell am I talking about now?

Let's work through this one on an example. Say I'm looking for "touch the object" and my dog lays down next to the object. That's a mistake. I don't want to reward that behavior, I don't want it to happen again. Let's talk for a second about what's going through my dog's mind: a dog that understands the training/shaping game is going to understand within a few seconds of offering the down that it was not what you were looking for, and they have made an error. Some very resilient dogs are going to get up and just offer the next behavior - these dogs are a very small minority. Some dogs are going to get stuck and stay in the down and the two of you are going to end up staring at each other for the next million years. Some dogs will get frantic and start flailing about trying to find the behavior that will get them rewarded. And some dogs - like Kaylee - are going get up and start looking for a noose because they have failed the Master and the sky is falling and now they must die. Very tragic. Much drama.

So how I handle this error to help my dog's learning is crucial. If my dog is inclined to offer another behavior, and I am willing to bet $100 that the next behavior will be one I can reward, I can wait for that behavior with a minimal amount of fall out. I don't do this very often. In this situation, I can't say the next behavior that gets offered won't be, say, putting her head on the floor or more duration on the down. (A Note from The Voice of Experience: Duration is the enemy of shaping. If my dog is holding still, then they're not offering behaviors, and if they're not offering behaviors, we can't get closer to our goal behavior. Always add duration last.)


Another way I can respond to this error is to give my dog the opportunity to perform a behavior I know will get them rewarded. I use hand touches for this a lot. Kaylee downs next to the object. I offer a hand touch. She touches the hand, gets rewarded, and offers the next behavior. Maybe this is one I can reward. Or maybe she makes another mistake, I offer another hand touch, she gets another reward, and I need to stop the session and go back to the drawing board to figure out why she's not offering the behavior I want. Either way, her enthusiasm for the game and our working relationship remains intact because she gets to be "right" and I can keep the rate of reinforcement high.

How I respond to errors in training, particularly with shaping, is nearly as important as how I handle success. Errors are a natural part of the learning process, but the more I can do to minimize error and increase the opportunities for success, the happier and more engaged my dog will be with me and with the learning process.

So, how about you? What helps your shaping sessions go more smoothly? What do you and your dog struggle with?


1 comment:

  1. I really like hearing these tips. It's learning about dog psychology and it makes perfect sense. The two dogs we have at my house are not mine but we obviously spend a great deal of time with them and alot of that time is when the owner's aren't around. We struggle with keeping Lincoln and Zeus from pestering us when we are eating. Zeus is much worse (Beagle/Boxer mix). He does not care about a sit or lay command when there is food involved but is great about it otherwise. However, he was trained with snacks to sit and lay. I understand Im not one of your pro trainer friends but thought I'd take advantage with a comment.😎

    ReplyDelete