Halp with free shaping...

Mmmkay, so my power cord on my Mac went dead, so I only have teh interwebz/computer for a few minutes a day on my little sister's ancient desktop. Anywhoo I decided to use all this free time (I am addicted to my computer) to work on free shaping with Frodo.

But I have a problem, the same problem that always prevents me from ever getting into really free shaping anything beyond basics with him. He doesn't advance and offer new behaviors.

For example, I am working on the beginning step of free shaping a pivot to heel, the part where he puts his front feet on the book. Okay, sounds easy. He offers nose touches (started with just looking, then advancing, then lying next to, now the nose touches), and knows that it will get him rewarded, so I need to stop rewarding for nose touches and only reward a new/more "right" behavior. But he isn't offering me any! He just gives me those same behaviors, not any new ones. If I don't reward him at all and try to wait him out he will just walk away and decide training is done. But if I continue to reward things like nose touches and lying down, then we aren't going to progress.

I know I could lure him onto it, but I would really like to be able to do something completely free shaped or I know I am going to give in and end up using luring/other means for the whole trick.

Help por favor?


TIA

I have the same problem with the schnauzers, Stryder more so than Kiba.
I'm interested in well for any help.

Juuuust kidding

Well, in the last 4-5 minutes Frodo learned to put his feet/foot on the book.
Not sure if it is considered free shaping still as I manipulated the envoronment, but it wasn't exactly luring!

I sat on the couch and put the book right in front of me instead of off a little to the front/side and he came up and half laid down on the book so I marked it. Kept marking the down with legs/belly touching book. He got that and started intentionally laying on the book, so I marked him standing over the book before he had time to lie down. Now he is, for the most part, stepping on the book.

But I still have a question, when he offers a new behavior that is closer to what I want, do I ONLY reward that behavior from that point on and ignore the older behavior?

I'm not an expert at this but here's what I would try... I'm assuming he'll shake. Try asking for shake a few times - then put the book in your hand and get him to hit the book instead of your hand. You may have to start with a small book and slip it in at the last moment before he hits your hand. Start moving it down slowly towards the floor, continuing to get him to whack it with his paws, until it's actually on the ground and you can take your hand out from under it entirely. From there you'll start rewarding for longer time whacking it with his paw until he's actually putting a paw on it - then holding out for putting two paws on it.

I have a really easy time with this because Auggie's "touch" is with his paws, not his nose. So free-shaping him to play with his skateboard was easy because his instinct is "hit that with my paw." While teaching back up, in fact, it was hard because he kept WHACKING ME with his paw... or frantically looking around the room trying to see what he was supposed to be touching. With Frodo you just need to teach him to start engaging with his paws and then it will become something he will explore and offer, like his nose-touch. IME once you get a dog to do ONE behaviour they will try and figure out how to apply that same behaviour to other things. Like Auggie who LOVES to offer me "sit" has also done the "back up-sit." *roll eyes*


ETA: WELL NEVERMIND THEN LOL.
What I do is only reward the new behaviour unless I can tell he's really struggling and getting frustrated, which is a sign to me that I need to back up a step and reward the previous one a little more before trying to move forward again.

Oh yeah - and manipulating the environment to get the behaviour you want is still shaping IMHO. I know people who suggest teaching back up by putting the dog between two pieces of furniture and then standing in front of him so he HAS to back up, LOL. It's not luring and it's still marking the dog for changing behaviour. You're just changing the environment to set him up and help him figure out HOW to succeed.

Naw, it's still free shaping.... it's just setting the dog up for success.

Free shaping means that the dog is completely offering new behaviors without any prompting, luring, or any other interference from you, including treat delivery. In the purest form of free shaping, the dog can't see you at all so that you can't possibly give him any cues; and your treat is always delivered in the same way to the same spot.

Of course we can't do a lot of "pure" free shaping, it's just not reasonable to expect to set up every session where we can see the dog but he can't see us. Plus, little tricks like where to stand yourself, where to deliver treats, etc. do help the dog learn the behavior faster; though it will be slower for him to generalize.

One of Karen Pryor's Ten Laws of Shaping is that you shouldn't raise criteria until the current criteria is on a variable schedule of reinforcement. What this means in practice is that the dog can do the behavior multiple times without reinforcement.

For example, your original shaping plan for the book thing was 1.) looking at book, 2.) moving toward book, 3.) laying next to book, 4.) nose touching, and 5.) pawing. So theoretically, the way the dog would've progressed is that he'd first get really good at looking at the book; as soon as you click, he'd look right back at the book. Once he has that behavior figured out, you can wait until he looks at the book twice before clicking; maybe even stretch it to three times. Putting a behavior on variable reinforcement usually makes that behavior stronger, meaning that the dog will give an extra *uumph* when doing the behavior. For looking, the extra *uumph* would be looking at and moving toward the book.... moving toward is your next criteria, so then you can click the moving toward. I'd spend about 10 trials (10 clicks) where your criteria is "Two looks or one movement toward" - that's the behavior he has to do to get a click. By the end of the 10 clicks, he should be offering the moving toward at least 70% of the time, which is when you can start only accepting the moving toward.

That's a lot of words to describe pretty much what you were probably doing without really knowing it. Plus your dog probably progressed so quickly that you didn't really notice.

Fast forward to the point where you got stuck: he was only offering nose touches and gave up when you stopped clicking them altogether. So at that point, I'd change my criteria to where you're accepting "Two nose touches, or lifting paw." Then he won't get so frustrated that he gives up, but hopefully gets a little bit frustrated to where he'll try to offer something better. If you don't see improvement in about 10 trials, then wait for him to offer three nose touches. And jackpot the paw lifting until he gets more consistent with that behavior. Continue clicking multiple nose touches until he's doing the pawing about 70% of the time, then raise your criteria to only clicking pawing.

Others have given you good advice for shaping. I just wanted to say that not every dog is going to be great at free shaping. If I leave Gavroche to offer behaviors, he'll just wave at me. But I can get him from a food lure to not having a lure in 3 repetitions or less, so I just lure. He's very reliable, food or not.

Even clicker training at all isn't for every dog. I've grown to adore clicker training, but Logan is SO focused on the food that for some things the clicker just doesn't work. For example when we do guide work, the food has to be left at home or there's no way he'll lead out.

This is true, but also IMO not every trainer is going to be good at using a particular method. Personally, I can teach many behaviors with free shaping faster than a lot of other trainers; but at the same time, I'm not very good at luring, it takes forever for me to get a lured behavior on a verbal-only cue, while I know some trainers who are much faster at luring than I am.

I had trouble dropping a lure too, but once I saw the Kikopup video everything clicked, and now I just think it's so cool to watch as I wean a dog off a lure lol

What kind of treats are you using? With Fozzie, there's delicious training treats.... and then there's shaping treats!!! I save his highest value treat (cheetos mixed in with chicken heart bits) for shaping, because he offers all sorts of crazy behaviors in hopes of a piece.

It sounds like you're just now getting really into shaping, so the issue is probably that Frodo doesn't have much confidence in offering behaviors. The more you practice the more he'll offer!

Free shaping is just that. I never use free shaping with a goal in mind. ( I know that sounds weird) because you're waiting for the dog to do something you like. My best free shaping moments with my dogs are bait bag sitting on the coffee table for a couple hours while I'm watching TV and my dogs are wandering around the living room. I usually watch out of the corner of my eye and when I catch something I like, mark (I use a consistant YES! instead of a clicker) then reward. Doing this I have taught one of my dogs to chase his tail, offer left foot, right foot, hind foot, and then his tail LOL, find a specific item, and look for (Dad, another dog, or a food item.)

So I guess my advice is for free shaping to stop forcing it. Either use a lure if you want a specific move or allow free shaping to be free.

Thanks everybody!! As always, learned a bunch from you!

We've just been using his kibble, but I read this and used some sliced ham this morning. He definitely is trying harder, but he still doesn't seem to get it completely. He will stand on the book multiple times in a row for clicks and treats, then randomly (it seems) just offer something like a down. I like that he is offering a different behavior, but why wouldn't he just step on the book again, he knows it works :lol-sign:

Auggie does the SAME THING, LOL. It's really weird. I was just commenting to my mom about that. If something is working (like backing up) why all of a sudden offer me a sit or a down or try pawing at stuff??
I think Auggie is a) making it too complicated for himself. He thinks there must be something else he SHOULD be doing so he's trying that. or b) he gets bored doing the same thing over and over so he starts offering something else to try and make it more fun for himself. Or he's just stupid, LMAO.

I have two theories:

1.) laying down also works... he's gotten clicks and treats for that in the past, how's he supposed to know he won't get clicks and treats for that now?

I've seen a lot of dogs do this if stimulus control (you don't get reinforced for the behavior unless I ask for it) isn't a priority for the trainer. When I started with my own dog, if we were in one session and she suddenly offered some other cute behavior, I'd click/treat that cute behavior even though it wasn't what I was trying to shape. I ended up with a creative dog who can't really do a whole lot on cue. Now, I'm very careful to stay focused on the behavior I'm trying to shape, no matter what else the dog does.... and now my dogs can figure out pretty quickly which behaviors don't get reinforced during a shaping session and don't waste our time offering those behaviors.

2.) maybe he's bored/tired/frustrated/whatever... like his brain is fried from the session.

MOST of the time when I have a dog that does this, it's because I kept the session going too long. Most of my shaping sessions are only 2 minutes long (I set a timer and force myself to stop!); sometimes longer but certainly never more than 5 minutes straight. For a lot of dogs, I do even shorter sessions; one dog I had couldn't go longer than 30 seconds for the first few months until his endurance was built up. Remember that it's way more efficient (the behavior will be learned faster in the long run) if you stop while the dog is still being successful, rather than waiting to stop when he's getting tired.

Lol, I personally think Frodo just likes to screw with me. I mean, if he does it right, then I am going to start fading the rewards, so throwing in stuff like this makes me continue to reward because he thinks that I think that he doesn't know it!

Makes sense.
I have been doing VERY well about clicking only the behavior that I want though. I should get cookies for that!

I do keep the sessions very short. Usually less than 2-3 minutes. I could agree that maybe he is frustrated though. He can do a very forceful down when he wants to!
He usually follows me to put treats/props/etc away, so I would say he is still engaged and wanting more at the end.

Now...another question
(this is EXACTLY why I need to be in a class and paying someone so I can bother them :lol-sign

I've gotten to the point where he KNOWS his front paws need to be on the book. So my next criteria would be movement of back legs. Which he does. Not purposefully, but he does shift/take small steps/etc and I click them. And click them. And click them.

He still doesn't seem to do it purposefully. I am thinking maybe he just has really bad hind end awareness and doesn't even think about what is going on behind his shoulders?
I am totally willing to continue clicking movement of his back legs, but I also don't want to waste time clicking and giving him mixed signals about what he is doing correctly.

An alternative I could do (that I saw on a luring video of teaching the pivot as opposed to the shaped one) would be to walk into him while his front paws are on the book so that he would have to walk his back legs around. But I feel like that is going to make it hard to then make him do that on his own while starting at a front position...

In my experience, the book pivot thing really helps teach hind end awareness. I've seen a HUGE difference in Gavroche since I taught him to do it. I need to get going on it with Logan too. The poor dog is so clumsy, he needs all the help he can get lol

Sael, do you just click for them moving their back feet or do you use your own body movement to get them to move, then click?

With Gavroche I use my body movement - that is, I stand with him in a front, and move a little, and expect him to follow in the front position. I click for movement then. However after doing this a few times he started offering to move around the book before I ever moved.

I used it mainly to help solidify fronts, but the improved hind end awareness was a big plus, as was a tighter heel/LLW.

I would use body movements as well.

What's your goal? For him to move from a front to heel position? If that's it, then I'd suggest starting with him in heel, then move away from him just a touch so that he learns to move his back feet in towards you. You want him to move in a tight circle, so that his front feet stay in the same spot but his back feet move into you. And click any movement of the back feet coming in, even if it's a tiny movement.

I don't feel like that's very clear (it's much easier to explain if you can demo it in person!), so I'll explain another way. Start with him standing in heel position at your left side with his front feet on the book. Preferably it's a small book so his feet are pretty much in the middle. Imagine a little tiny circle around his front feet, and a much larger circle that his back feet will move in if he pivots the whole circle with his front feet in that little circle.... like a compass kids use in school to draw perfect circles, the pointy end would be his front feet, they stay in the spot, and the pencil end would be his back feet, they rotate around. Ok, so he's in heel. Now you move about 1/8 of the way around the circle, so that when he gets in heel position at your new spot, his front feet will have stayed in the same spot and his back feet will be the only thing to move. Click when he moves his back feet. Then move another 1/8 of the way around the circle. When he starts getting the idea, you can move 1/4 of the way at a time, then 1/2 way, 3/4, whatever. Halfway around, though, would be him moving from front to heel, which would be your finished behavior.

But if your goal is to work on straighter fronts, you'd do a similar thing except you start with him in front instead of heel. Then you just rotate around the circle, clicking his back feet moving into a straight front.

I'm pretty sure you're joking (can never tell for sure over the internet!), but it is definately possible that that's what he's figured out here. If so, then you really need to figure out for Frodo, what is the purpose of switching to a variable reinforcement schedule and how to make it more effective. 'Cause really, you should be pretty constantly switching the reward schedule based on what behaviors he's offering. I know trainers often describe shaping as a step-by-step system, but in practice it is very fluid; you don't necessarily "finish" one step before going on to the next.