Q
of the week
|
Last updated March 6, 2003 Operant
Training vs. Classical Conditioning (If you aren't, pick up a copy of Culture Clash!!) Let's put it in practical use. Here's the scenario: Dog
in shelter run barks and leaps aggressively at the gate when people
walk by. From
Cathy: Classical conditioning is one of the most important ways we can help shelter dog get adopted. This can work with the dogs like the one described above, or shy dogs who sit at the back of their cages looking scared to death. Although initially my motivation for volunteering at Spokanimal was a selfish one (gaining experience and learning from the dogs behavior so I could be a better trainer) it has grown to being able to truly make a difference in these dogs lives and making them more adoptable --- and this after only a couple of times being there. From
Lisa: If the treat slingers wait until the dog is sitting quietly before delivering the treat, the dog operantly learns that sitting quietly earns a treat. From
Diane: The dog in the shelter run has an already established opinion and fear of people, based on his past experiences, breeding, etc. So, tossing a cookie to him in passing is not creating this snarling fearful behavior, it's changing the pattern of what happens, and his opinion of people, by offering something new. He doesn't expect cookies from humans, so at first he has to process the new info and then learn to respond in a new way. If the humans add a nice word and a smile in a non threatening way he'll be able to be less fearful faster. Build on this and soon there will be a dog excited to see people instead of snarling at them. The rate of progress depends on lots of factors, and some dogs may not be able to rehab. Fearful dog snarls---recieves new and positive input from previously scarey situation----questions what the deal is-----positive input repeated-----dog realizes scarey situation no longer happens----looks forward for cookie distrubitors to come by! Yummmmmm
BEAUTIFULLY PUT, LISA AND CATHY - AND HOORAY!! GREAT ANSWER FROM DIANE, TOO! And a great answer from Kim Imel: Operant
Training Vs. Classical Conditioning My response: Classical
conditioning: Learning by association The purpose of the repeated pattern of "human appearance equals food" is to classically condition a new emotional response to humans passing by the kennel run... the theory being that one can in time change the dog's behavioral response to humans passing by. The dog will soon look forward to, instead of becoming alarmed at, the stranger's approach. Of course it would seem that giving food would reward what EVER the behavior the dog is presenting at the time of the food delivery. Here's the answer to why it doesn't:
Scientific studies show that classical conditioning
overrides operant. When a dog is in a limbic, or emotional state, he is unable to learn through operant conditioning. When a dog is totally stressed out, it *can* learn, but only by association. Learning via consequences doesn't happen in this state. So, the dog will associate the person with the treat, but not learn that the treat is a consequence of its behavior. There is learning going on, but not the kind where we say "sit," the dog sits, the dog gets a cookie, the dog *chooses* to sit again. So by dropping treats regardless of the dog's behavior, we CAN classically condition his emotional state by pairing a positive outcome with the appearance of a human at the front of the cage. When this happens, the main or sole effect of dropping treats is to help the dog calm down and begin to have a more pleasant association with humans passing by. The end result, the aggressive behavior decreases as the dog's comfort level increases. Even dogs who are behaving more operantly (able to learn), are probably behaving in this way to drive people away because, for whatever reason, people make them nervous. They repeat the behavior because it works. Thus, dropping the treat still changes the emotional response and the dog will likely feel less need to behave in the repelling manner after several repetitions. * A very few dogs are fully operant when they are barrier-guarding at shelters (will observe that by being aggressive they can cause a person to appear and give treats.). Even in these, we can create a happier emotional response to the presence of humans. As the dog begins to calm down, we need to keep notes, observe, quantify the behavior of the animal: "is it improving or not?" and proceed accordingly. As the dog gains confidence and is able to concentrate on how his behavior affects the outcome, then do more operant conditioning oriented work with him. Mark appropriate behaviors as they happen until he starts to consciously offer them, and then switch to Premack: "If you sit quietly, then I'll give you a treat." Additional comment from Lisa: "Emotional learning overrides all else -- it is why dogs do not generalize well when we train operantly and they DO generalize emotionally driven behaviors." QUESTION FROM DIANE: Okay, here's my question about redirection. Sometimes I see this. Dog snarls at other dogs. Handler redirects, rewards. Dog doing better, gets this down. Most dogs do fine from there. What about the occasional dog who learns the pattern of snarl (or whatever reaction they do), turn to mom and look for cookie, and keeps repeating this pattern specifically for the reward? Some dogs almost literally smile at mom as they turn away from the dog they snarl at, and they seem relaxed even as they do the snarl. They learned to redirect, and that they need not worry about the other dog, just look to mom, but not to omit the snarl. What's the next step? (Gee,
could this be SOFI the BC/Aussie? *G*) The dog is simply doing what it's been taught. It has learned the sequence. The other dog becomes cue to lunge (and the lunge itself is self reinforcing behavior - the dog gets a buzz out of it), handler redirects & from the dog's point of view is rewarding a predictable sequence of chained behaviors. The primary goal of this exercise from start to finish is to redirect BEFORE the lunge - before the dog takes in its breath - while the dog is NOT reactive. The redirection is to PREVENT the lunge, the lunge is not the cue to redirect. This is often the result of poor handler timing. When working with incredibly fast dogs who show little if any body language signals (like Sofi), this can be extremely difficult to catch at first. If the handler has a proactive plan - redirecting *at first notice BEFORE the dog erupts* and jackpots any choice NOT to erupt, she can get rid of this accidentally reinforced behavior. It won't be easy, because part of the behavior is self-reinforcing. I would personally add a verbal NRM (neutral no reward marker - ah-ah!) at the point of lunge and redirect as always but no longer C/T the resulting attention if *I* missed the pre-signals and a lunge happened. (I would take responsibility for being the one who missed it and would not set the dog up to lunge so I could correct it!) If the dog spun around following the lunge and gave me its attention, I would quietly acknowledge the attention but not C/T, increase the distance from what precipitated the lunge in order to set the dog up to succeed, and jackpot success on the next pass. Lisa - perhaps you could share how a correction situation classically conditioned your Bouvier student to hate the approach of other dogs! Lisa's response to Diane's question AND the Bouv story: Diane asked a GREAT question and your resonse, was, like "oh yeah, of course!" Have to redirect before the explosion. How about in a "de-exploding" training set-up -we get a handful of well-trained dogs and handlers to walk by Sofie at varying distances (starting farther away and gradually getting closer). Sophie is tied up. Sophie's mom simply feeds, feeds, feeds as the dog goes by. If Sophie explodes the other handler and dog are to stop with the other dog facing away from Sophie and mom walks away. [Hooray for John Fisher - good plan!] When Sophie turns to look for mom, the other dog/handler can then walk away. OR in a classroom situation, if Sophie explodes, the food goes away and mom turns into a tree. After they are successful at this, the next step is to reward Sophie for voluntarily turning away from the other dog, whether or not she looks at her mom. In other words, click a lookaway when Sophie offers it instead of exploding. The last time I saw Sophie in class her mother was often not even looking at Sophie during the entire explode, re-direct, click and feed sequence. She has actually trained Sophie (thanks to us) to perform the sequence of behaviors. With Denali, I rewarded voluntary look-aways. And the calm looks at other dogs. This is the ticket. Reward a dog's decision to NOT explode. The
Bouv Story: Chloe, the Bouvier, is one of four Bouvier's in this family. They are show/obedience dogs. Chloe is 14 mos old and started exhibiting "aggressive" behavior towards other dogs in the show ring. When getting a history on the dog, I learned that there are many rude dogs where they train. Starting when Chloe was 6 mos old, whenever dogs got loose and/or in Chloe's space her handler would pop the leash and jerk Chloe away. Soon, Chloe was growling as the other dogs approached, which earned her another leash correction. Only now, Chloe's mom, not understanding what was happening also became extremely tense when other dogs approached so whenever other dogs got close she sucked in her breath and tightened her muscles in preparation for whatever was going to happen. Karen's response convinced Chloe that bad things happened when other dogs came close, so she worked harder at chasing them away with bigger growls. She was telling all other dogs, "GET AWAY NOW Pleeeeeeaaaaaase, so I don't get in trouble" In other words, Chloe developed a negative association to dogs approaching when she was on-leash. The biggest lesson I could offer was to teach Karen how to read her dog and be her advocate and re-direct prior to the explosion. Following is an excerpt from a recent email from Karen: "We have had two obedience classes with Barb since we met with you. Chloe was good in the first class, and she seemed to be even more relaxed in last night's class. I spend more time reading her than actual training, but it seems to be paying off. I'm keeping her out of the major crush of dogs, but we still are a part of the class. We had a Golden puppy get loose last night, and I just shielded Chloe with my body. I'm pleased with her progress. I will set up another session after the next class. I read the article that you recommended (He Just Wants to Say Hi, by Suzanne Clothier), and I ordered the book. I recognized myself in the article." Thank you, Lisa!
QUESTION OF THE WEEK, PART TWO: GIVE
EXAMPLES OF HOW THIS THEORY The
reactive behavior of the shelter dog is seen in almost every first
week class. Every
time there is a dog with a stress-based issue in class, we must
use classical
Given the laws of learning theory, we know we can't "train"
them to "feel better" ... WHAT
SPECIFIC WAYS CAN WE USE CLASSICAL CONDITIONING
|
