Premack Principle AKA Premacking and the power of prediction in creating value transfer
Premack Principle AKA Premacking
and the power of prediction in creating value transfer.
I had a warm piece of chocolate brownie cake on my birthday. The frosting was fudgy; the cake was moist. I ate it slowly with a bit of vanilla ice cream. If I had been given broccoli for my birthday treat, when I was expecting a cake, my disappointment level would have been defcon 5 but I ate my greens with dinner.
I have used an example of a choice between broccoli and chocolate cake as a training analogy in my classes for years. It's a clear example of many of the problems people face in training their dogs and their children and even forming the good healthy habits we need to function in our personal lives.
When there is no cake, broccoli is just fine as a reward, It’s crunchy and tasty and even healthy… However, given a free choice, a dog (and often ourselves) is “likely” to choose the cake. Not the Broccoli. Too often we go into a training session holding (figuratively) broccoli and try to make the dog ignore the figurative cake (don’t feed your dog chocolate.) We are asking our dog to make an impossible choice. You see, we have this low-value thing and we expect them to pass on the high-value rewards that are FREE to them in the environment. Those things your dog wants to do and can’t seem to ignore… Those are the “likely behaviors.” These behaviors contain their own high-value rewards. We also call these “self reinforcing behaviors.” Sniffing, chasing, barking, digging, chewing, possessing the toy and so on.
Where many articles on this topic and trainers who use this principle miss the boat is that the broccoli doesn't have to be broccoli. They can both be cake! Let's call them “small cake” and “big cake.” Given a free choice between the two the dog always picks “big cake,” until it gets premacked and sees “small cake” as the on-switch for “big cake.”
The psych textbook might say:
A Likely Behavior May Reinforce An Unlikely Behavior.
I find people often have a hard time understanding Premack’s Principle when described that way but it's entirely accurate. Another way of saying this might be: performing the unlikely choice becomes a prediction or an “on-switch” that the high value reward becomes available. In this way we achieve a transfer of value from the likely behavior to the unlikely behavior. In that way the dog sees the small choice as the on-switch for the thing they wanted much, much more.
Premack's Principle could also be more simply stated as: A low value reward can predict a high value reward.
Alarms should be going off in your head right now at the power of that statement. Let it sink in for a moment if you need and reread it.
Where we see an amazing result with this idea is in Bite-work. I know not everyone wants to do bite-work but bear with me as it is a great analogy for all those things your dog really wants to do that you have issues getting them to ignore. Bite-work, AKA “protection dog” is challenging work to say the least. It is not dogs biting whatever and whoever, being over aggressive or over protective. Good bite-work is precise, focused, structured and reliable. It is an example of all of the traits we need from good dogs. It just includes behaviors that not every owner wants. Try to see past these differences if you are not familiar and think of these things as obedience in the face of extreme competing motivation.
For a protection dog to function in the job, he or she, MUST WANT THE BITE. A lot. During the training process we make biting the bite equipment highly rewarding and fun for the dog. We allow the dog to feel powerful, often even more powerful than they actually are during the bite. As helpers and decoys, we over-act to the dog’s power and get them to think they are super dogs. Soon they start to believe it and act like super dogs. Flying after the bite and moving at incredible speeds. Braving even the most fierce opposition and capturing and defeating foes many times their size.
I ran some reps with a personal protection dog client a while back. The dog, Piper, a female working-lines German Shepherd, was a rock star on the decoy (Me.) She bit like a beast. However, she started getting salty on the out, failing to let go on command and started to leave the handler without cue. Making her own free choice to take what she wanted. She also had failures to recall, losing handler focus... but biting well.
We see this as too much drive without enough control. It is bad news. This is not functional and can even become dangerous. The options are few and most trainers apply pressure, mask drive by punishing the dog for the failures and they sometimes get where they want, maybe... with the right dog. Piper however, suppresses with correction to the point it hurts the work. Punishing would send us backwards a lot.
Sometimes the punishment and pressure approach creates conflict. Conflict can be the result of confusion about how the dog’s job works. If that conflict arises, too often we see a shift in the dog from seeing the handler as a teammate to seeing the handler as the referee in a fixed game. Nobody likes that ref. A dog who thinks it has super powers may redirect those powers onto the handler out of frustration. We always strive for the dog to see the handler as a teammate. The team captain. Both the dog and the handler, striving for the same goals, to win the game. This is essential to maintaining the proper relationship with the working dog.
That day with Piper, we ran maybe a dozen premack reps with a small tug, and a bigger recall wedge, Let's call them small cake and big cake. The handler posts the dog on a harness and leash and offers the small tug to the dog. All in the presence of a bigger toy that is held by the helper in neutral. At first Piper did not want the small tug and was doing everything to avoid it and handler focus. At some point she started to give a half hearted mouthing to the tug. Then, 8 or so reps in... click, she gets it. Bite the “little cake” with the handler, get the “big cake!”
We put in a couple more reps, great, finished up and moved on to other stuff, put that equipment off to the side 40 feet and worked on some other skills and took a break. Then I went and got some big bite equipment to put on. The small tug was still on the ground, so was the larger wedge.
Now remember this dog WANTED the big cake and the on-body-worn equipment is the biggest cake of all! As I'm getting in position Piper runs over, grabs the small tug off the ground, runs it to her handler and waits.... like: "you forgot the on-switch boss."
At this point everything got easier. Piper increased her focus to her handler and performed obedience to get the small tug reward and then be released to the line of departure for attacking the decoy. This required good teamwork between the two of us. I had to be able to switch between neutral and attacking at just the right moments so the dog remained cued by my behavior to bite.
But wait, remember the problems that brought us to this session? Piper was failing to out and recall off the bite and going early on her own. We didn't even practice that. In fact we practiced nothing about outs and not really any recalls. All those “big cake” rewards ended with Piper winning the toy/equipment and carrying it off. What did we practice though?
We practiced the power of prediction. In our premacking reps, biting the small tug became an “on-switch” that turned on the decoy from neutral to available. Then we practiced some obedience skills with the small tug as a reward. During that obedience practice it was clear that value transfer had occurred from the big cake to the small cake. Why? Again: Prediction. When we finally asked Piper for an out off the “big cake” that day, she had won the game a dozen of times or more. Well what happened? Clean outs happened… Why? We had created an idea within Piper’s mind that only the small cake would turn on the big cake. Any obedience behavior including out, recall, heeling, etc. could be inserted into the queue of behaviors that lead to the on-switch of the small tug being offered, then being released to the big cake. Her direct attack cues were also preserved, don’t worry. Value transfer happens when the dog starts to see the value of the big cake in the small cake. Great news, this transfer happens with or without the big cake present in later sessions! Our obedience practice for “small cake” looks much better and more enthusiastic even without the “big cake” reward in the mix. This transfer needs some maintenance but it remains powerful with periodic practice and grows session after session.
Now let's think about some non-bite-work related ways to apply premack principle. Some dog’s struggle to crate. Some to the point they get destructive to the crate and even injure themselves. Premack’s principle can be applied to crate training with interesting results. We often play crate games with our dogs and the obvious method is to throw food rewards in the crate and get the dog to chase them in. Is this premacking? No. It's chasing a reward. What about rewarding the dog for coming out of the crate. Wait what? Yes. Reward for coming out. Now, not to the point we are rewarding the dog for trying to come out, that will build an escape artist dog. But if we can get the dog to go in by chasing “small cake,” why not give the “big cake” reward as soon as the dog comes back out? I usually like to speed up this value transfer by giving the dog’s bridge or conditioned reinforcer upon entering the crate. This helps them to understand step one. But on step two, I give the big cake reward for exiting. Why? Because the dog had to be inside to exit. Clearly. After a few reps of this I wait. And if my big cake was big enough, I just might get the dog to make a choice to enter to try to trigger big-cake even with no small cake to chase. When this works it really works, the dog will jump back in the crate as soon as the big reward from coming out is finished. They choose to go in to switch the big cake reward on. In this way we create value transfer to the crate. Applying this method we can get the dog to choose to take longer and longer sessions waiting in the crate for the reward. Usually we have them wait with the door wide open. No pressure. Many of them will go in and lay down. I think they usually see pressure on the floor of the crate as the “on-switch.” We use the same method in creating the place behavior with chasing games for coming off the bed, we use the same method in creating the touch-pad behavior, in retrieve, in recall, in heeling and so on. The possibilities are endless. Lets train dogs. Semper Audax.