[More thoughts on the terms 2Q, 4Q, and instinctive drift from articles posted to the PGD list, somewhat edited for republication here]
Posted December 17, 2009, at 1:05 AM
On Dec 17, 2009, at 12:02 AM, higgins10000 wrote:
> Sorry to plead ignorance here but what is 2Q?
Hi, Brad. I'm not Randy, but I can answer that. "2Q" is a term that was invented by Alice Woodyard to distinguish between traditional dog training methods ("4Q") and those that don't use aversives.
Both 4Q and 2Q refer to the four operant conditioning quadrants, outcomes for behavior that affect the probability of future choices under the same circumstances: positive and negative reinforcement, positive and negative punishment.
"2Q" provides a useful alternative for those who feel the term "positive" is not an accurate description of our approach, since we do NOT use positive punishment, and we DO use negative punishment.
LL&L
Posted December 17, 2009, at 10:55 AM
On Dec 17, 2009, at 9:46 AM, Alice Woodyard wrote:
Another reason "positive" is not accurate is that field training using
traditional methods is also strongly positive due to all the rewards (birds)
that are intrinsic to the dog work. You can't get away from that positive
aspect of field work even if aversives are used to shape behaviors.
Great point, and one that I'm aware of though I sometimes forget to mention it.
If it weren't true, they'd both be 2Q methods, just different Qs. :0)
I'd like to put even more emphasis on "strongly" in your first sentence for the information of those who are not aware of it. Successful 2Q and 4Q retriever training are both based above all on the dog's intrinsic desire and ability to perform field retrieves, so the lion's share of the outcome that shapes future choices comes not from extrinsic reinforcers like treats and games of tug, but from the intrinsic reinforcement of doing the work itself. The trainer's extrinsic reinforcers add to that, especially with puppies, but they don't add much, and they add less the more the developing dog discovers within herself the joy of retrieving.
The idea of training a retriever for advanced field work and somehow eliminating intrinsic reinforcement, so that the dog is only responding to extrinsic reinforcers -- a concept that seems to dawn on every newbie "positive" trainer as a lightning bolt, one example having recently been discussed at length on this list -- is totally out of touch with the actual experience of training a retriever. First, the idea of "desensitizing" a dog to intrinsic reinforcers is as absurd as the idea of "desensitizing" a teenager to music, video games, and sexual awakening by flooding the teenager with those stimuli. Second, if you somehow did manage to kill birdiness in an individual dog, you'd have a broken dog who would stand no chance of meeting the standards of dogs who retrieve for the pure love of it. At Junior level tests and breed certifications, we occasionally see dogs who don't show intrinsic field-retrieving desire. This probably isn't because they were "desensitized"; they either weren't born with it or they haven't had enough experience for it to come out. We don't see dogs like that at the more advanced levels, because they can't do the work.
Unfortunately, the developing dog doesn't only discover within herself the joy of retrieving correctly. She also discovers within herself an instinct for undesirable behaviors that were never trained, never practiced, but rise up from inside and become increasingly attractive to her. Those are the behaviors that I call "instinctive drift", borrowing that term from the Keller's famous article "Misbehavior of Organisms". From my experience, instinctive drift is a far more pervasive and difficult challenge in training retrievers for field work than in other canine sports and other animal-training endeavors, precisely because the sport is based upon the dog doing what it was bred to do, inviting those instincts to awaken and grow with every retrieve.
LL&L
Thursday, December 17, 2009
Monday, December 14, 2009
Shaping Interaction with a Duck
[I sent the following post to PGD list on May 18, 2007]
In the thread "Forced Retrieve":
On May 17, 2007, at 6:04 PM, Linda Conrad wrote:
So, Lindsay, care to share how you got her to love to pick up
birds in 4 weeks time? I only ask because I'm struggling with this
issue right now - I jokingly said as soon as they start running hunt
tests with bumpers and Dokken ducks, we'll be MORE than ready!!!
Hi, Linda. Aren't Dokken's cool? We have a bunch of them, all different "species".
I should mention that it didn't take four weeks to get Lumi to love to pick up ducks, that took only a few minutes, but it took those same few minutes several times over a period of days, during which we had to start at the beginning of the process each time. Eventually, we didn't have to do again, since now Lumi seems all too ready to pick up ducks, but who knows, we may need to do it again some day if she loses that.
Other barriers also needed to be crossed in those four weeks. The actual progression was: 1st week, she wouldn't pick up a duck; 2nd week, she dawdled coming back on land retrieves; 3rd week, she picked up her first flyer (flier?); 4th week, Katie bar the door.
As for picking up ducks, we used shaping, or as it's also known, freeshaping. Although Lumi is a clicker dog, I didn't use a clicker because I felt that for this it would be superfluous. In addition, I didn't shape with food as I usually do (though perhaps I could have) because I felt we needed the best possible reinforcer, and for Lumi, that's swimming (an example of Premack reinforcement).
So each session went like this:
(1) Take a dead duck (preferably not too cold and in the best possible condition) plus a dummy and/or Dokken (I used both, I'll explain below) to the shoreline of some favorite swimming location. We used a fast-running creek, deep enough for Lumi to swim in, one of Lumi's favorite places in the world, about five minutes drive from home.
(2) Toss the duck on the ground near the water.
(3) Shape increasing interaction with the duck, reinforcing each correct response as follows:
(a) At the very instant of the response, Lumi receives a rousing send-out into the water (cue as marker), followed by me throwing the Dokken into the water just ahead of her, or upstream, or somewhere to make the swim the most fun possible. I used the Dokken for the water retrieves to help Lumi practice how to carry a duck.
(b) When Lumi comes back with the Dokken, bring her to heel, take the Dokken, and toss the dummy in the grass a few feet away, cueing "shake". In theory, this should keep me slightly drier than if she shakes as soon as I take the toy.
(c) When Lumi then brings the dummy back to me (again coming to heel), suddenly offer a game of tug (we use only the dummy for tug) for a few seconds.
(d) Finally, toss both toys on the ground, cue "leave it", and draw Lumi's attention back to the dead duck.
(e) Occasionally, I'd reverse the game of tug or the send out, or use only one of two. Lumi likes tug almost as much as retrieving. I don't want Lumi to know what reinforcer she'll get -- her motivation seems highest when I surprise her. (NOTE: Pretty much every field trainer I've met has warned me not to play tug with a retriever. Maybe I'll regret it someday. So far in our brief time in field training, I've seen no disadvantage and it has always had huge motivational benefit in other training.)
(4) As with any shaping I do with Lumi, the key is how I define "correct response" from one successive approximation to the next. For Lumi and me, it was something like:
* Glance in general direction of the duck -- "GO OUT!" (I might repeat this definition of a "correct response" or others two or three times, or go immediately to the next level on the next rep.)
* Glance at duck itself without lowering head
* Glance at the duck, lowering head slightly
* Seem to sniff the duck a little
* Touch nose to the duck -- (Sometimes after some progress, Lumi stops giving the current level of response, and we'll back up to an easier level and resume from there.)
* Place mouth on duck -- (When starting, you wonder if this will ever happen, after seeing how Lumi felt about the duck at the beginning. But she had a strong incentive -- those go-outs, to figure out what it was going to take to be sent out again.)
* Drag the duck one inch -- (When Lumi finally did that, I knew that it would be smooth sailing from there, and it was.)
* Drag the duck three inches, five inches
* Slightly lift the duck
* Lift the duck higher
* Put the duck in my hand, which of course I made as easy as possible
* "Retrieve" the duck after a toss about one foot away from my hand
* Retrieve after a toss two feet away, then longer and longer until I reach my throwing distance limit
* Put down a "sight blind" with Lumi beside me, walk a short distance away, send Lumi to retrieve (by now, she's coming to heel with the bird)
* Put down a sight blind with Lumi waiting some distance away in a sit, walk back to her, give her a treat her for waiting, then send her
* Increase the distance on sight blinds
* If available, recruit a bird boy for some longer marks (good luck finding a bird boy, that's one of my biggest problems)
I should mention that one of my criteria for correct response is enthusiasm. I might reinforce one or two tentative responses if necessary, but I want to extinguish those (by not reinforcing them) as quickly as possible.
I have no idea if that exact progression will work on other dogs. Lumi and I have done a good bit of shaping in other activities over the years, so she knows this game well. That is, she knows that reinforcement is contingent on her own actions. I suspect that dogs who have not been shaped in the past may need time to learn that concept, but maybe I'm wrong, maybe any retriever would respond as Lumi did to the above sequence regardless of training background.
I know from other lists that people often have questions about shaping, such as, What if the dog just looks at me?, or What if the dog gets interested in something else instead of interacting with me? I'm sorry if no one on this list has such questions, but I'll give a brief answer in case someone does.
First, I'm not a shaping purist. Dogs (and all animals) are most drawn to salient stimuli, so I'm not above moving the duck around to make it more interesting, even though that's not pure shaping. I'm also not above moving around myself, such as moving to the other side of the duck, so that Lumi might brush the duck as she comes toward me. ("You touched the duck! YAY! Go OUT!")
Perhaps most important, I set my criteria at whatever level it takes for me to get a response I can reinforce quickly. I don't want twenty seconds to go by with Lumi just wondering what's going on. Did her eyebrow flicker? Good enough! "GO OUT!"
Then there's timing. Sometimes I watch people trying to shape and there's too much delay between the moment the dog committed to whatever the current criteria are, and the reinforcement. I consider half a second too long. At the very instant Lumi does whatever it is I'm looking for, that's the moment I burst into motion and voice.
So if the dog is just looking at you, move around, move the duck around, take any glance away from you as your criteria until you're getting some fluency with that. If the dog isn't in the game, make the game as easy as necessary by lowering your criteria so that she can get reinforcement quickly (three seconds is what I aim for), and of course use a high value reinforcer.
Sorry if I've given too much detail, just trying to answer as best I could. I hope if you try it you'll let us know how it goes.
L&L
In the thread "Forced Retrieve":
On May 17, 2007, at 6:04 PM, Linda Conrad wrote:
So, Lindsay, care to share how you got her to love to pick up
birds in 4 weeks time? I only ask because I'm struggling with this
issue right now - I jokingly said as soon as they start running hunt
tests with bumpers and Dokken ducks, we'll be MORE than ready!!!
Hi, Linda. Aren't Dokken's cool? We have a bunch of them, all different "species".
I should mention that it didn't take four weeks to get Lumi to love to pick up ducks, that took only a few minutes, but it took those same few minutes several times over a period of days, during which we had to start at the beginning of the process each time. Eventually, we didn't have to do again, since now Lumi seems all too ready to pick up ducks, but who knows, we may need to do it again some day if she loses that.
Other barriers also needed to be crossed in those four weeks. The actual progression was: 1st week, she wouldn't pick up a duck; 2nd week, she dawdled coming back on land retrieves; 3rd week, she picked up her first flyer (flier?); 4th week, Katie bar the door.
As for picking up ducks, we used shaping, or as it's also known, freeshaping. Although Lumi is a clicker dog, I didn't use a clicker because I felt that for this it would be superfluous. In addition, I didn't shape with food as I usually do (though perhaps I could have) because I felt we needed the best possible reinforcer, and for Lumi, that's swimming (an example of Premack reinforcement).
So each session went like this:
(1) Take a dead duck (preferably not too cold and in the best possible condition) plus a dummy and/or Dokken (I used both, I'll explain below) to the shoreline of some favorite swimming location. We used a fast-running creek, deep enough for Lumi to swim in, one of Lumi's favorite places in the world, about five minutes drive from home.
(2) Toss the duck on the ground near the water.
(3) Shape increasing interaction with the duck, reinforcing each correct response as follows:
(a) At the very instant of the response, Lumi receives a rousing send-out into the water (cue as marker), followed by me throwing the Dokken into the water just ahead of her, or upstream, or somewhere to make the swim the most fun possible. I used the Dokken for the water retrieves to help Lumi practice how to carry a duck.
(b) When Lumi comes back with the Dokken, bring her to heel, take the Dokken, and toss the dummy in the grass a few feet away, cueing "shake". In theory, this should keep me slightly drier than if she shakes as soon as I take the toy.
(c) When Lumi then brings the dummy back to me (again coming to heel), suddenly offer a game of tug (we use only the dummy for tug) for a few seconds.
(d) Finally, toss both toys on the ground, cue "leave it", and draw Lumi's attention back to the dead duck.
(e) Occasionally, I'd reverse the game of tug or the send out, or use only one of two. Lumi likes tug almost as much as retrieving. I don't want Lumi to know what reinforcer she'll get -- her motivation seems highest when I surprise her. (NOTE: Pretty much every field trainer I've met has warned me not to play tug with a retriever. Maybe I'll regret it someday. So far in our brief time in field training, I've seen no disadvantage and it has always had huge motivational benefit in other training.)
(4) As with any shaping I do with Lumi, the key is how I define "correct response" from one successive approximation to the next. For Lumi and me, it was something like:
* Glance in general direction of the duck -- "GO OUT!" (I might repeat this definition of a "correct response" or others two or three times, or go immediately to the next level on the next rep.)
* Glance at duck itself without lowering head
* Glance at the duck, lowering head slightly
* Seem to sniff the duck a little
* Touch nose to the duck -- (Sometimes after some progress, Lumi stops giving the current level of response, and we'll back up to an easier level and resume from there.)
* Place mouth on duck -- (When starting, you wonder if this will ever happen, after seeing how Lumi felt about the duck at the beginning. But she had a strong incentive -- those go-outs, to figure out what it was going to take to be sent out again.)
* Drag the duck one inch -- (When Lumi finally did that, I knew that it would be smooth sailing from there, and it was.)
* Drag the duck three inches, five inches
* Slightly lift the duck
* Lift the duck higher
* Put the duck in my hand, which of course I made as easy as possible
* "Retrieve" the duck after a toss about one foot away from my hand
* Retrieve after a toss two feet away, then longer and longer until I reach my throwing distance limit
* Put down a "sight blind" with Lumi beside me, walk a short distance away, send Lumi to retrieve (by now, she's coming to heel with the bird)
* Put down a sight blind with Lumi waiting some distance away in a sit, walk back to her, give her a treat her for waiting, then send her
* Increase the distance on sight blinds
* If available, recruit a bird boy for some longer marks (good luck finding a bird boy, that's one of my biggest problems)
I should mention that one of my criteria for correct response is enthusiasm. I might reinforce one or two tentative responses if necessary, but I want to extinguish those (by not reinforcing them) as quickly as possible.
I have no idea if that exact progression will work on other dogs. Lumi and I have done a good bit of shaping in other activities over the years, so she knows this game well. That is, she knows that reinforcement is contingent on her own actions. I suspect that dogs who have not been shaped in the past may need time to learn that concept, but maybe I'm wrong, maybe any retriever would respond as Lumi did to the above sequence regardless of training background.
I know from other lists that people often have questions about shaping, such as, What if the dog just looks at me?, or What if the dog gets interested in something else instead of interacting with me? I'm sorry if no one on this list has such questions, but I'll give a brief answer in case someone does.
First, I'm not a shaping purist. Dogs (and all animals) are most drawn to salient stimuli, so I'm not above moving the duck around to make it more interesting, even though that's not pure shaping. I'm also not above moving around myself, such as moving to the other side of the duck, so that Lumi might brush the duck as she comes toward me. ("You touched the duck! YAY! Go OUT!")
Perhaps most important, I set my criteria at whatever level it takes for me to get a response I can reinforce quickly. I don't want twenty seconds to go by with Lumi just wondering what's going on. Did her eyebrow flicker? Good enough! "GO OUT!"
Then there's timing. Sometimes I watch people trying to shape and there's too much delay between the moment the dog committed to whatever the current criteria are, and the reinforcement. I consider half a second too long. At the very instant Lumi does whatever it is I'm looking for, that's the moment I burst into motion and voice.
So if the dog is just looking at you, move around, move the duck around, take any glance away from you as your criteria until you're getting some fluency with that. If the dog isn't in the game, make the game as easy as necessary by lowering your criteria so that she can get reinforcement quickly (three seconds is what I aim for), and of course use a high value reinforcer.
Sorry if I've given too much detail, just trying to answer as best I could. I hope if you try it you'll let us know how it goes.
L&L
Subscribe to:
Posts (Atom)
Posts
