Treat for position or "reset" the dog?

This forum is all about training and behavior. Everything from potty training to working titles!

Postby pitbullmamaliz » February 17th, 2010, 5:52 pm

I figured this might make an interesting thread, based on Matt's question to Nancy:

maberi wrote:Ha, great video. Vanya is one good looking dog (love the markings). Was there any reason you choose to have Vanya come back to you for the rewards? This was a LOOOONGGG discussion on the clicker forum, that is the only reason I ask :wink:

tiva wrote:Our copy of When Pigs Fly came today. What fun! Just as I started searching for that definition of "biddable", the FedEx truck came down the driveway, hours early. Vanya was snoozing on the back deck in the sun, and decided it was his turn to illustrate what When Pigs Fly is all about. He adores vans, new people, open car doors...so when the FedEx driver opened up her door, Vanya leapt in (ignoring my recall command, of course) and tried to smother her with kisses, in between checking out her van for hidden treats. Luckily, she likes dogs and Vanya decided she's even cuter than his plush dog.

Here's a clip of Vanya playing the shaping game. I'm lazy, so I usually lure, but I felt guilty enough reading the first chapters that I decided to get out a wooden box and shape some interactions with the box:
http://www.youtube.com/watch?v=mhZwq1zYAds&feature=player_embedded

Thanks for the recommendation for this book. It looks great.


So I've heard both treat for position (where you want the dog) or use the treat as a way to "reset" the dog so it can offer the behavior again. Opinions? Does it depend on what you're doing?
"Remember - every time your dog gets somewhere on a tight leash *a fairy dies and it's all your fault.* Think of the fairies." http://www.positivepetzine.com"

http://www.pitbullzen.com
http://inaradog.wordpress.com
User avatar
pitbullmamaliz
Working out in the buff causes chafing
 
Posts: 15437
Location: Cleveland, OH

Postby tiva » February 17th, 2010, 6:26 pm

I wanted to re-set the behavior; otherwise, it's hard to get anything new to shape with Vanya (except whining--he can always vocalize). When I was following Control Unleashed and learning to shape a "go to mat" behavior, Leslie McDevitt suggested tossing the treats to get lots of reps in a shorter time.

For foundation behaviors (sit, watch, LAT, target, heel), I do treat in position.
User avatar
tiva
Snot Nose Bully Pup
 
Posts: 165
Location: WI

Postby maberi » February 17th, 2010, 7:51 pm

I spent a few days reading through the posts on the clicker forum regarding this. There is a lot of good info on it and Greta posted some really interesting information on the topic. The basics were that treating FOR position (and in some cases IN position) drastically decreased the amount of time it took the dog to learn the behavior.

One of the key behaviors she mentioned that she treats IN position for are things such as a "go to mat" behavior. The thought process is that the dog is going to "hang out" where the food is because it reduces the amount of time of the rewards. In this example of training a go to mat behavior the dog is called off the mat to reset the behavior. I found the reading extremely interesting because I've seen so many people treat off the mat to reset the behavior. It is an interesting discussion because it starts to delve into what happens between the marking of the behavior and the reward. Really interesting reading
Look beyond what your own eyes see
User avatar
maberi
I Save My Empty Calories For The Bottle
 
Posts: 2781
Location: rochester, ny

Postby maberi » February 17th, 2010, 8:24 pm

One other note I forgot to mention (this isn't my stuff I'm just regurgitating what I read)

The primary reinforcer is reinforcing the most recent behavior. So in the case of the go to mat exercise if you are calling the dog back to you to treat them you are reinforcing the last thing they did (come back to mom) or if you are throwing the food off the mat you are reinforcing run off the mat and go sniff the ground to find the food. The thinking is why not use that power to reinforce what you want (being on the mat). I honestly never thought about it. I've always rewarded on the mat but that is just because that is how I was taught.

Bob Bailey also had some interesting examples in regards to training chickens and rewarding for position.
Look beyond what your own eyes see
User avatar
maberi
I Save My Empty Calories For The Bottle
 
Posts: 2781
Location: rochester, ny

Postby amazincc » February 17th, 2010, 8:29 pm

Without having ever read any of this stuff before... I always reward for the last thing I want to re-enforce.

Right now I'm teaching Faust "stay", and I will walk back to him and reward him for obeying instead of calling him to me to get his treat - I think that method is so much less confusing for the dog. :)
User avatar
amazincc
Jessica & Mick
 
Posts: 9814
Location: Holding them both in my heart.

Postby katiek0417 » February 17th, 2010, 9:36 pm

Okay, I'll chime in... :wink:

The purpose of reinforcement is to increase the occurrence of a behavior.

One of the main principles of ANY learning theory (whether it be classical or operant conditioning) is that they, for the most part, follow Aristotle's Laws of Association, of which there are four. Now, many behaviorists maintain that the laws of similarity and contrast don't play as major a role in conditioning as do the other two: the Law of Contiguity and Law of Frequency.

Law of frequency is easy to understand: the higher the frequency in which two things are linked together, the more likely they are to be strongly associated.

Law of contiguity predicts that things that occur close together in space and time, will be more likely to be associated.

Now,take these laws and apply them to learning theory. The more frequently you pair food with a behavior, the more likely the organism will be to perform the behavior. Why? Because the behavior has, in the past, brought about an appetitive consequence.

But, you have to also take into consideration the law of contiguity. Things that occur together in space and time.

So, if the dog gets the treat, right after it comes to you, then you are positively reinforcing THAT behavior. In reality, this is how many of us teach a recall to our dogs. We call dog, dog comes to us, dog gets treat! So, why would we even say that we are "resetting" the dog by getting it to come to us, when, in reality, you're just teaching it to come to you.

Now, is there such a thing as delayed reinforcement? Yes, but there is not a ton of empirical support for the idea that it is as effective. In fact, most researchers cite Skinner's (1938) original study using immediate reinforcement and reinforcement occurring at 2, 3, and 4 second delays in their arguments AGAINST the efficacy of delayed reinforcement. In fact, Skinner found that while the rats COULD acquire the behavior when reinforcement was delayed (2, 3, and 4 seconds), it was severely retarded as compared to immediate reinforcement.

Now, if we look at marker training (which is actually referred to as a conditioned reinforcer by behaviorists), many studies have shown that the conditioned reinforcer functions in much the same was as a primary reinforcer, where a primary reinforcer is simply a reinforcer that is inherently good, and the organism does not have to "learn" it's value (the most common examples of primary reinforcers are food, water, and sex). A conditioned, or secondary, reinforcer on the other hand is one in which the "goodness" of it is learned (for example, the clicker in clicker training) through classical conditioning (the pairing of a previously neutral stimulus with an unconditioned stimulus)...However, this only occurs if the proper classical conditioning trials take place for that previously neutral stimulus.

As a side note, seconary reinforcers are also the primary mechanism in our economic system (i.e., you aren't born knowing that money is good, but you do learn that it can be used to get you other things, which is good) which is basically a larger scale token economy (commonly used in mental hospitals and prisons).

Therefore, to me, you don't need to use this "reset" with a dog if you're also using other consequences in your training with the right frequency and contiguity...and, no, people, I'm not simply talking about a prong collar correction :rolleyes2: ...even something as simple as turning your back - negative punishment - used in conjunction with positive reinforcement can gain the behavior you want (I've done all of Axo's training using these two consequences, and Greg has even made a comment that he's starting to look incredible with his behaviors, all things considered (his age, etc))...

See, this is what happens when you catch me while I'm writing a lecture for learning! :wink:
"Rumor has it, compulsion is evil."

Katrina
Sacha CGC - Dumb Lab
Nisha CGC, PDC, PSA TC, PSA 1 - Crazy Malinois
Drusilla SLUT- Pet
Nemo - Dual-Purpose Narcotics
Cy TC, PSA 1, PSA 2, 2009 PSA Level 3 National Champion
Axo - Psycho Puppy
Rocky - RIP My Baby Boy
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby maberi » February 17th, 2010, 9:45 pm

Nerd
Look beyond what your own eyes see
User avatar
maberi
I Save My Empty Calories For The Bottle
 
Posts: 2781
Location: rochester, ny

Postby amazincc » February 17th, 2010, 9:49 pm

maberi wrote:Nerd


Ditto! :neener: lol

Just TELL me if I'm doing it right... :wink:
User avatar
amazincc
Jessica & Mick
 
Posts: 9814
Location: Holding them both in my heart.

Postby katiek0417 » February 17th, 2010, 9:51 pm

maberi wrote:Nerd


:neener: :wink:

amazincc wrote:
Ditto! :neener: lol

Just TELL me if I'm doing it right... :wink:


Yes :mrgreen:
"Rumor has it, compulsion is evil."

Katrina
Sacha CGC - Dumb Lab
Nisha CGC, PDC, PSA TC, PSA 1 - Crazy Malinois
Drusilla SLUT- Pet
Nemo - Dual-Purpose Narcotics
Cy TC, PSA 1, PSA 2, 2009 PSA Level 3 National Champion
Axo - Psycho Puppy
Rocky - RIP My Baby Boy
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby tiva » February 17th, 2010, 11:11 pm

Katie, interesting points, but the 'reset' in my shaping example is a purely pragmatic decision: I want Vanya to work on the motion of getting into a box, not perfect a position within the box. If I treat at the end of the motion, it takes him a long time to realize that what I'm shaping is the early part of the motion, not just the completed position. If what I wanted to shape was the end position, not the action of hopping into a box, then I'm sure your argument might hold.

Matt, I went back and read that interesting January thread from clickersolutions, and for the life of me, I can't see how what they're doing is really all that different than a nice little lure. If you click when, say, the chicken hits the left side of the target because it's an approximation of the center of the target (your goal for that chicken), but then you offer the treat in the middle of the target, where you really want them to be, anext time they target closer to the middle. Makes sense, but that sounds a lot like you're luring them toward the center of that target. ( I love lures, but I'm being made to feel guilty about them by When Pigs Fly.)

Speaking of chickens, I'd better go lock up the hen house before the coyotes get any bright ideas.
User avatar
tiva
Snot Nose Bully Pup
 
Posts: 165
Location: WI

Postby katiek0417 » February 17th, 2010, 11:31 pm

tiva wrote:Katie, interesting points, but the 'reset' in my shaping example is a purely pragmatic decision: I want Vanya to work on the motion of getting into a box, not perfect a position within the box. If I treat at the end of the motion, it takes him a long time to realize that what I'm shaping is the early part of the motion, not just the completed position. If what I wanted to shape was the end position, not the action of hopping into a box, then I'm sure your argument might hold.

Matt, I went back and read that interesting January thread from clickersolutions, and for the life of me, I can't see how what they're doing is really all that different than a nice little lure. If you click when, say, the chicken hits the left side of the target because it's an approximation of the center of the target (your goal for that chicken), but then you offer the treat in the middle of the target, where you really want them to be, anext time they target closer to the middle. Makes sense, but that sounds a lot like you're luring them toward the center of that target. ( I love lures, but I'm being made to feel guilty about them by When Pigs Fly.)

Speaking of chickens, I'd better go lock up the hen house before the coyotes get any bright ideas.


No, my argument does still hold. Even when you shape by means of successive approximations you have to reinforce as the behavior occurs. Personally, I like to use a marker for this, and have used it successively in training a formal retrieve, jumps, and send aways (just to name a few). These are all more "complex" behaviors that are, in reality, made up of much smaller ones that have been chained together....

For a "reset" why not use some release (other than a reinforcer), letting the dog know that that particular session has ended. That is how conditioning trials are conducted.
"Rumor has it, compulsion is evil."

Katrina
Sacha CGC - Dumb Lab
Nisha CGC, PDC, PSA TC, PSA 1 - Crazy Malinois
Drusilla SLUT- Pet
Nemo - Dual-Purpose Narcotics
Cy TC, PSA 1, PSA 2, 2009 PSA Level 3 National Champion
Axo - Psycho Puppy
Rocky - RIP My Baby Boy
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby tiva » February 18th, 2010, 12:44 am

katiek0417 wrote: No, my argument does still hold. Even when you shape by means of successive approximations you have to reinforce as the behavior occurs. Personally, I like to use a marker for this


I believe you are misunderstanding me. Of course I'm using a clicker as a marker to reinforce the moment the desired behavior occurs (it may not be clear from the video, but the clicks can be heard in there). We're talking about when to deliver the reward when a marker has already been used to mark the moment of the desired behavior, not when to deliver a treat when no marker is being used.

The session hasn't ended between each successive treat. I end the actual session with a cue to let him know we're done and he can go play.
User avatar
tiva
Snot Nose Bully Pup
 
Posts: 165
Location: WI

Postby TheRedQueen » February 18th, 2010, 1:31 am

I found one thing I do now...I throw treats a LOT more for Fig than I ever did with the other dogs...seriously, he's only inches off of the ground (unless he's leaping for them). I find that I toss more of his treats to the ground rather than handing them to him. lol Perhaps not always the best idea for what I'm doing...but that's the artsy side of me... :dance:

I'm a lazy crap with an aching back...so sue me. :wink:
"I don't have any idea if my dogs respect me or not, but they're greedy and I have their stuff." -- Patty Ruzzo

"Dogs don't want to control people. They want to control their own lives." --John Bradshaw
User avatar
TheRedQueen
I thought I lost my Wiener... but then I found him.
 
Posts: 7184
Location: Maryland

Postby Malli » February 18th, 2010, 3:31 am

I think that treating out of position works if you are just "practicing" the behavior and the dog is fairly well versed in what you are asking (see Erin's dog Fig in the "mat" video). I think when you are specifically trying to catch something or mark it, reward asap is best :|
thoughts?
I can only please one person per day. Today is not your day, tomorrow doesn't look good either.
_______________________________________
"You didn't know of the magical powers of the break stick? It's up there with genies and Harry Potter as far as magic levels go." SisMorphine 01/07/07
User avatar
Malli
E-I-E-I-O!
 
Posts: 6341
Location: CANADA EH?

Postby maberi » February 18th, 2010, 8:44 am

tiva wrote:Matt, I went back and read that interesting January thread from clickersolutions, and for the life of me, I can't see how what they're doing is really all that different than a nice little lure. If you click when, say, the chicken hits the left side of the target because it's an approximation of the center of the target (your goal for that chicken), but then you offer the treat in the middle of the target, where you really want them to be, anext time they target closer to the middle. Makes sense, but that sounds a lot like you're luring them toward the center of that target. ( I love lures, but I'm being made to feel guilty about them by When Pigs Fly.)


I'm not sure if I would categorize the placement of the treat AFTER the chickens have been marked luring. If that is luring, than I would argue that you are also using luring by luring Vanya off of the box and back to you to treat her. They are no doubt free shaping the behavior with the chickens as you are with Vayna and the box. The only difference is that in the example of the chickens they are using the "power" of the reward placement to increase the learning curve of the chickens. Rewards occur in the center of the target after the chickens behavior has been marked, so the chickens begin to gravitate to the center more.
Look beyond what your own eyes see
User avatar
maberi
I Save My Empty Calories For The Bottle
 
Posts: 2781
Location: rochester, ny

Postby katiek0417 » February 18th, 2010, 11:10 am

tiva wrote:
katiek0417 wrote: No, my argument does still hold. Even when you shape by means of successive approximations you have to reinforce as the behavior occurs. Personally, I like to use a marker for this


I believe you are misunderstanding me. Of course I'm using a clicker as a marker to reinforce the moment the desired behavior occurs (it may not be clear from the video, but the clicks can be heard in there). We're talking about when to deliver the reward when a marker has already been used to mark the moment of the desired behavior, not when to deliver a treat when no marker is being used.

The session hasn't ended between each successive treat. I end the actual session with a cue to let him know we're done and he can go play.


Ohhh...okay...an no, I didn't hear the clicks when I watched it (I watched it last night and Greg had the TV on, and was playing his (stupid) video game)... :oops: sorry!

Malli wrote:I think that treating out of position works if you are just "practicing" the behavior and the dog is fairly well versed in what you are asking (see Erin's dog Fig in the "mat" video). I think when you are specifically trying to catch something or mark it, reward asap is best :|
thoughts?


Absolutely...with Axo, I mark behaviors as soon as they occur, then I treat immediately. With Cy, however, I mark behaviors, but he doesn't get his reward (usually a bite) until the series is over...

maberi wrote:
tiva wrote:Matt, I went back and read that interesting January thread from clickersolutions, and for the life of me, I can't see how what they're doing is really all that different than a nice little lure. If you click when, say, the chicken hits the left side of the target because it's an approximation of the center of the target (your goal for that chicken), but then you offer the treat in the middle of the target, where you really want them to be, anext time they target closer to the middle. Makes sense, but that sounds a lot like you're luring them toward the center of that target. ( I love lures, but I'm being made to feel guilty about them by When Pigs Fly.)


I'm not sure if I would categorize the placement of the treat AFTER the chickens have been marked luring. If that is luring, than I would argue that you are also using luring by luring Vanya off of the box and back to you to treat her. They are no doubt free shaping the behavior with the chickens as you are with Vayna and the box. The only difference is that in the example of the chickens they are using the "power" of the reward placement to increase the learning curve of the chickens. Rewards occur in the center of the target after the chickens behavior has been marked, so the chickens begin to gravitate to the center more.


I'm not a member of clickersolutions...but I did want to chime in that Matt is right. The technical term for clicking when the chicken goes to the left side, then clicking when it goes to the middle is reinforcement of successive approximations, most people know it as shaping. With this method, you are allowing the organism to figure out what to do, but you are rewarding it for each step it takes closer to the goal (desired behavior).

Luring suggests baiting. There's more intervention with luring. Think about lure coursing: dogs run as fast as they can because they are following a lure...they are lured into running. When I first teach puppies to go into a kennel, I use luring. I throw the treat in, when they run in, I praise them. But this is MUCH different than me waiting for the dog to offer behaviors closer and closer to what I want, and then rewarding them...
"Rumor has it, compulsion is evil."

Katrina
Sacha CGC - Dumb Lab
Nisha CGC, PDC, PSA TC, PSA 1 - Crazy Malinois
Drusilla SLUT- Pet
Nemo - Dual-Purpose Narcotics
Cy TC, PSA 1, PSA 2, 2009 PSA Level 3 National Champion
Axo - Psycho Puppy
Rocky - RIP My Baby Boy
User avatar
katiek0417
pointy ear hoarder
 
Posts: 6280
Location: Glen Burnie, MD

Postby TheRedQueen » February 18th, 2010, 11:43 am

and to offer my two cents on luring...I do lure occasionally if I'm having trouble getting the behavior any other way...but it's few and far between. To bring up Fig again, I lured him the first few times to "down" because he just wasn't offering the behavior during training sessions (he's clicker savvy and offers MANY behaviors...just not a down...lol) So I helped him by luring maybe 5 times total, and then the lightbulb went off..."ohhhh...that's what you want!"

My biggest problem with luring is that it often becomes a crutch for people...they don't fade the lure quickly (or ever). :nono: I was watching the Therapy Dogs in our organization (we have weekly classes...usually around 30-50 dogs in one big group) last night...and these are all adult dogs, trained already (they have to pass an obedience test to get in the program, etc)...and EVERY owner when told to "Down your dog" took their right arms, outstretched, palm down, bent over and swept their arms down to the floor and said "DOWN" to their dogs. :rolleyes2: :bs: That's seriously not fading a lure! I wanted to jump in and call class and say..."now try that while standing upright and keeping your arms at your sides..." The reason I get really annoyed with it, is that I see my Service Dog puppy raisers doing the same things...and trust me, if you're paralyzed in a wheelchair, that movement isn't going to happen. :rolleyes2: I have to remind them all of the time...and push them to fade lures.
"I don't have any idea if my dogs respect me or not, but they're greedy and I have their stuff." -- Patty Ruzzo

"Dogs don't want to control people. They want to control their own lives." --John Bradshaw
User avatar
TheRedQueen
I thought I lost my Wiener... but then I found him.
 
Posts: 7184
Location: Maryland

Postby tiva » February 18th, 2010, 11:46 am

katiek0417 wrote: The technical term for clicking when the chicken goes to the left side, then clicking when it goes to the middle is reinforcement of successive approximations, most people know it as shaping. With this method, you are allowing the organism to figure out what to do, but you are rewarding it for each step it takes closer to the goal (desired behavior).


Hi Katie,
Again, I'm not disagreeing with you on your terms, but we're talking about something a bit different--clicking when it goes to the left, and rewarding in the middle (for a click on the left).

When I earned my PhD in behavioral ecology, we focused less on learning theory and more on evolutionary theory, which shows when I try to remember the precise language for learning theory!
User avatar
tiva
Snot Nose Bully Pup
 
Posts: 165
Location: WI

Postby tiva » February 18th, 2010, 12:03 pm

maberi wrote:I'm not sure if I would categorize the placement of the treat AFTER the chickens have been marked luring.
Oh right--I forgot--lures occur first. This afternoon, I'll try to think of a good behavior to shape and reward closer to the position I'm ultimately aiming for. Hardly a scientific experiment, but fun all the same.

When we use a target to shape behaviors, does that count as a lure, or as some other kind of information? Last night, I taught Vanya to leap into the air to bop my hand, using his target. We shaped this by clicking first for his standard nose touch to target, then I raised the target higher and higher so he had to leap into the air to touch it. Then I faded the target, so he was bopping my hand, not the target. If I had spread cheese on the target, that would have been luring, presumably. But a cheese-free target? With the jump, I couldn't treat in position (could I?)
User avatar
tiva
Snot Nose Bully Pup
 
Posts: 165
Location: WI

Postby maberi » February 18th, 2010, 12:24 pm

Right on!! I think you hit on a really good point. You can't treat IN position for all behaviors nor would you want to. You having Vanya target your hand is a great example. I think that feeding in position is good for stationary behaviors (sit, stay, heel position, go to mat, etc..) but definitely doesn't work for many other behaviors.

The discussion on luring is interesting as well. I totally agree with Erin. Most people use luring as a crutch and don't fade it quickly enough. Oddly enough luring doesn't have to be food only. I taught Kayden targeting a while back and use my hand as a lure when teaching a lot of behaviors for disc. I just recently taught him a scoot which basically consists of having the dog stand in front of you, then do a 180 degree turn so their back is towards you, and from there, back up until they go through your legs. I had to teach a couple different behaviors (backing up, turning 180 degrees, backing up between my legs) before I could chain them all together. I used a LOT of luring when teaching that trick, but started fading the lures as he learned the different components. I'm still luring him a little with my hand for the 180 degree turn, but I'm slowly fading that and incorporating a cue with the disc for the half turn.

Again, I think you have to go with whatever is working. I know I'm always analyzing the techniques I am using because more often than not there are things I could change to help him understand what I'm looking for.

tiva wrote:When we use a target to shape behaviors, does that count as a lure, or as some other kind of information? Last night, I taught Vanya to leap into the air to bop my hand, using his target. We shaped this by clicking first for his standard nose touch to target, then I raised the target higher and higher so he had to leap into the air to touch it. Then I faded the target, so he was bopping my hand, not the target. If I had spread cheese on the target, that would have been luring, presumably. But a cheese-free target? With the jump, I couldn't treat in position (could I?)
Look beyond what your own eyes see
User avatar
maberi
I Save My Empty Calories For The Bottle
 
Posts: 2781
Location: rochester, ny

Next

Return to Training & Behavior

Who is online

Users browsing this forum: Majestic-12 [Bot]

cron