Tuesday, February 28, 2012

Jane Killion Seminar: Reinforcement Training

Jane Killion is a reinforcement trainer.

Notice the words I chose there, because she doesn't call herself a positive trainer. Why? Well, in behavior science, all “positive” means is “add.” You add something in order to manipulate future behavior. Sometimes you add something in order to increase a behavior, and sometimes you add something in order to decrease a behavior. Technically speaking, this label can easily refer to some wildly divergent training methods. But Jane also feels the term “positive” can be divisive. Since so many people misunderstand it to mean upbeat, kind, or virtuous, it implies that people use use other methods aren't. What's more, it also implies that anyone who is nice must automatically be positive, which is misleading.

As a result, Jane favors the term “reinforcement training.” Her methods are based on providing reinforcement to the dog, whether that's by adding something desirable to increase a behavior (positive reinforcement), or by removing something unpleasant in order to increase a behavior (negative reinforcement). What she avoids whenever possible is the use of punishment, as all that does is suppress behavior, not teach the dog what you want him to do.

Training under the umbrella of reinforcement gives a trainer a variety of things to use to create behavior. Without a doubt, the most popular- and Jane's favorite- is to use food. Not only is it a primary reinforcer, but it's also easy to dispense, easy to control, and it tends to tap into the “thinking” part of the dog's brain.

Of course, any one who has worked with dogs knows that sometimes training with treats just doesn't work. Jane identified two main reasons this happens: emotional interference (ie, stress), or competing reinforcers that are more relevant or interesting. After all, when it comes down to a choice between eating a bit of hot dog and chasing a squirrel, most dogs will choose the squirrel. And if you're working with a “Pigs Fly” dog, the chances that he will choose the environment over you is large.

That doesn't mean that you can't use positive reinforcement with a dog who isn't interested in food. You can, it just takes a bit of creativity. Enter Premack, a principle which states that a high probability behavior can be used to reinforce a low probability behavior. If your eyes just glazed over, don't worry; Jane has come up with a very clever way of describing the Premack principle. She calls it ICE: Identify “hot” reinforcers (anything your dog wants, which might include sniffing the ground, rolling in something stinky, or peeing on a bush), Control them, and then Exchange a behavior for the hot reinforcer.

Jane showed us an excellent video of her working with a very distracted young dog that really wanted to do was explore and sniff, which was incompatible with his owner's desire that he walk nicely on the leash. In the video, Jane identified the hot reinforcer (sniffing), controlled that reinforcer by shortening the leash so he couldn't sniff, and then waited for an opportunity to exchange a behavior for sniffing. As soon as the dog looked at her, she let out the slack in the leash and told him to go sniff. After a few moments, she shortened the leash again, and the process repeated. Soon, he was willingly paying attention to her while they walked through the field together. Pretty cool.

As the video demonstrated, Premack is a powerful thing, and I wish she would have demonstrated it live during the seminar. I can understand why she didn't- we had limited time, and Premack takes longer than handing out a cookie- but I was still a bit disappointed. There was even a great opportunity to demonstrate Premack/her ICE system since one of the working dogs had a medical problem that made it difficult to use food for training, but instead Jane had that handler switch out for a different dog.

Another way to train with reinforcement is through negative reinforcement. This often ignored method has us remove something aversive in order to provide relief to the dog. This is a powerful thing, and dogs and people alike will perform behaviors that bring them relief. Jane was clear that she will not add the aversive herself, but that she will use unpleasant things that are already present; a shiny floor, perhaps, or the presence of the judge leaning over the dog to do the stand for the exam. And, Jane said, often the equipment in agility is a stressful thing for dogs.

To that end, she showed us another excellent video demonstrating how reward placement can provide negative reinforcement. On screen, we saw a dog who was hesitant to jump on the table, an agility obstacle that requires the dog to stay on a platform for several seconds before moving on to the next obstacle. The dog in question was slow and reluctant to get on it and lie down. Jane shaped the dog in several steps to move towards, get on, and then lie down on the table, each time throwing the treat away from the table, which allowed the dog relief from the piece of equipment he found unpleasant. Soon, she had a dog eagerly offering the desired behavior.

Again, I was disappointed that she didn't demonstrate the use of negative reinforcement during the seminar, even though there was actually a training problem that could have been set up to take advantage of it. One of the working dogs was having trouble performing recalls in the presence of other people. Jane chose to have people stand in two rows, then instructed the handler to call the dog to run between them. I would have loved to see Jane use some negative reinforcement in this situation through the relief of social pressure by having the people move back when the dog responded. This could have been a powerful reward for this dog, but instead, the dog got a treat when she came.


Now, there's nothing wrong with this. The approach worked; the dog was able to improve her ability to come despite the presence of people, but it was stressful for the dog. Jane was okay with this; learning is stressful, she said. She's right, of course, but after my experiences with Maisy- a dog with clinical anxiety- I have far less tolerance for stress during training. As a result, I just wasn't comfortable watching the amount of stress the working dog endured- especially when Jane could have used negative reinforcement to relieve the dog's stress while still accomplishing her goal.

As these two examples demonstrate, most of what Jane showed us over the course of the weekend was straight positive-reinforcement-with-food. More to the point, Jane has a tendency to shape everything. At times, it felt like this the only tool she has, but then, she's so good at it, she doesn't really need anything else.

I doubt that I will ever attain her level of skill in observation, setting criteria, and timing, but I am hopeful that having the chance to watch her will improve my own abilities. I look forward to telling you about some of what I learned... But it will have to be in the next post, as this one has gotten far too long already.

12 comments:

Jen said...

Thanks so much (again!) for posting your experience at the seminar, and what Jane Killion discussed. Valuable information, to be sure, and I enjoyed reading it.

Crystal Thompson said...

I'm glad you enjoyed it!

Dizzy said...

I've been paying more attention to the negative reinforcement quadrant since Jane's seminar. I realized that I use it quite a bit. In rescue, I've worked with quite a few severely shy dogs. Yesterday night, I was working with a girl who wouldn't come into the same room as me. We marked each time she looked at me, and I had the foster owner throw a cookie behind the dog so she would move away. Before long, she was moving toward me, and eventually was comfortable at a distance of about three feet before I ended the session. Good progress! Plus, the dog was focused on me instead of on the cookie as with luring, so I believe the bite risk was lower because she was only going to come forward as far as she was comfortable. I've done this before with other shy dogs, but this was the first time I actively thought of it as negative reinforcement. Jane certainly got us thinking, didn't she?

andrea said...

fascinating!
thanks :)

Catalina said...

That's Tibby and me! I like reading your review of the seminar - you understood so much more of it than I did :)

Crystal Thompson said...

Dizzy, negative reinforcement is often overlooked, which I understand (something bad has to be present to remove, and who wants to do that), but it makes PERFECT sense for some situations.

Catalina- YES! You and Tibby were so awesome, too. I was so proud of her for working through a difficult situation. Are you going to write more about your experiences?

Andreja said...

Great reminder about negative reinforcement! I usually only think of it in terms of ear pinches in retrieve training and that is just not my kind of thing.

Unlike Jane, I like using negative punishment (taking something reinforcing away to decrease behavior). Like shortening the leash if my dog doesn't respond to cues (too bad, you don't get to sniff). Or leaving the room if he looses focus during training (my favorite!). It's funny how he suddenly remembers that he *does* know how to retrieve a toy if I leave him for just 10 seconds :) And he does it with a lot of energy. Not what someone would expect from 'punishment' at all :)

On the other hand I must admit there was a time when negative punishment backfired. We were on our first agility trial and he didn't complete the weaves. We tried three times, he exited at 10th pole every time. So I told him 'All done'. I even played with him after we exited the ring, but he understood that he didn't get to finish the course because of those stupid weaves. He loves agility, so leaving the ring was punishing. During next run he flatly declined to weave :) Now that was a great lesson for me - negative punishment can only work when the dog knows exactly what to do to avoid it. I thought he knew the weaves, but clearly he did not have enough experience yet.

The Premack video you mentioned sounds very interesting. Are you saying that just by letting him sniff after he looked at her she got voluntary looking at her? Or was there some increase of criteria once he offered the look reliably? How long did it take her to train this?
I played around with this in the past, but it took him weeks to offer attention while walking (and treats because I didn't want to reward with sniffing every time).

Crystal Thompson said...

Andreja,

I don't remember the Premack video mentioning it specifically, but I'm sure she increased the criteria. The beginning showed her rewarding for just fleeting eye contact, and by the end, the dog was walking by her side, making eye contact the whole time, before she released him. So... the criteria was increased to include duration and movement.

I believe she said it all happened in one (or maybe two) session(s), but I'm not sure how long that might have been. The implication was that it happened pretty quickly, though.

Catalina said...

I would like to write about it, but I'm still waiting for the videos from our sessions. I want to review them and see if what I remember happening is what was really going on.

Crystal Thompson said...

It would be interesting to write about the differences between what you experienced and thought happened vs. what showed up on the video. (If they're different.) I know that when I tape my training sessions, I see different things when I watch the video than I did during the sessions.

Andreja said...

Oh, I remembered now why I was having so much trouble Premacking looking up with sniffing. Ruby thinks looking around is almost as fascinating as sniffing. He will just stare at something in the distance... I guess this must be a useful trait in a dog bred to hunt hare. So when I would prevent him from sniffing he would just stand there, looking around. I think the only way to avoid this would be to use a head halter... something that he is not accustomed to.

Joanna said...

Wonderfully detailed review.

I am currently working a couple of times a week with a cavalier mix puppy who has low food and play drive, and VERY high drive to sniff and explore his environment. We did a small amount of recall foundation with food and lured sits and downs, and then I took it on the road using sniffing as a reward.

I have him on a short leash to limit self-reinforcement as much as possible. Initially I would stop and call him, and reward with sniffing when he moved toward me. The next step was coming to me close enough that I could do a collar grab. Then I started holding onto his collar with one finger (to limit his choices so he can be more successful) and asking him to sit. Last session, I also asked for a couple of downs when we were in a low-distraction area.

It's slow going compared to a dog who loves food, but you work with the dog you have. :) I also premack waiting at gates/doors to go through, and saying hello to people and other dogs. Last week I held a havanese in my arms and slowly lowered her to the ground as he had to hold a sit before saying hello to her.

I think I will also add a cue for eye contact.

The fun thing is that he's become much more affiliative with me after a few training sessions. :)