Keeping the Sucker in the Game: A Poker Lesson for Dog Trainers

dogsTo create a behavior with positive reinforcement there must be a minimum pay-off to keep the animal working. No matter what else you do, you must always remember to pay attention to the dog’s willingness to work. If you do that, you will have great latitude in how you run the process. For instance, many trainers attempt to always create “clear communication” so the dog learns in the least number of repetitions and is never confused. This goes back to B.F. Skinner’s only suggestion for creating behaviors – successive approximation. The idea is that you give tiny bits of information from repetition to repetition and the animal eventually gets to more complex behavior. That is like playing solitaire rather than Poker.
Successive approximation is a tedious, boring and rarely necessary way to craft a behavior. Because it is done ‘on the cheap’, giving many tiny rewards solitaireover many repetitions, the animal comes to think of the process as “penny ante”. Over time it becomes harder and harder to keep the animal motivated. Think about learning foreign language vocabulary and you’ll see the problem. If the only learning you do is limited to baby steps, you will never take adult steps.

EG: I once fixed the behavior of two, ten-year-old coyotes at the Phoenix Zoo. That is a different story altogether, but over the months I was offering my services I was introduced to a female squirrel monkey who was suffering from Valley Fever. The medication she was supposed to take was incredibly bitter. In six months of “stair step”, successive approximation they hadn’t gotten her to drink any of the meds. At the end of about 20 minutes I got her to drink 3, 3ml syringes full of the noxious liquid. I did it by “upping the ante”, pissing her off and triggering monkey aggression – all things that zoo people and behavior analysts don’t do.2008-08-09-zoo-sq-monkey-21121 Her biggest problem was that they were nickel and diming her, literally, to death. Because the aggregate reinforcement was being doled out with an assumption of many repetitions, they never made it worth it to her to take a gulp. No, I didn’t simply offer jackpots. I made the predictable outcome completely unpredictable. Meaning sometimes I gave her a jackpot for just putting her hand around the syringe. Sometimes she got a click and no treat. Sometimes she got nothing at all – no feedback about whether her behavior was correct or not. If you have ever been in a poker game where the stakes meant something to you, you will understand this model.

For those of you who have played a great deal of poker, you’ll recognize this as “keeping the sucker in the game.” You have to let the sucker win, sometimes, in order to skin him of his poke. You have to build his confidence and then shatter it. You have to make him feel like he’s going to run the table – and then jerk it out from under him. That is a great model of nature – which is why this process works so effectively. If you can arrange the wins and losses to keep the sucker in the game, you win. If you do it enough times, the sucker will become a much better player – or quit. You can’t shear the sheep if he jumps the fence.

Rules of the Game:
Rarely end a session on a high note.
Many modern animal trainers preach a motto of “always end on a high note.” This displays a very shallow understanding of how behavioral repertoires are built. Consider if you worked as a laborer and were completely exhausted at the end of each day. Would an ice-cream sundae at the end of a work day actually change your mind about the daily grind? If you become predictable about high notes at the end of each session it will lead to predictable planning on the part of the animal. The idea is to teach the animal that reality doesn’t always lead to an ice-cream sundae. Sometimes it’s work. People who play cards eventually lose. If they lose $100 and win a $50 jackpot on the last hand, they are still in the hole when they count their money.

Inject unpredictable consequences.
Making the training process interesting is more important than grinding out minute advances, repetition after tedious repetition working for chump-change. Many people play “penny ante” poker – but nobody becomes a master that way. However, people play the lottery because of the allure of a huge pay-off. Injecting unpredictable consequences keeps the animal guessing and metaphorically hungry for the periodic jackpot. Caution: Jackpots are not “super reinforcers” for any specific behavior. They affect the overall tendency of the animal to keep playing the game. When you give a big jackpot you are actually triggering variability and ironically insuring that the next repetition will be other than what you just liked. (I sometimes use a jackpot to make an animal stop doing the same error, over and over.)

Nurse the sucker through the dry spells. Remember, if the sucker quits or starts to lose interest in the game you can’t shear him. The same is true of a dog that learns that training is a plodding affair that never pays off. If you see the animal flag, be willing to give him an easy win. If that has to be some behavior he already knows, don’t worry about it. You will have plenty of repetitions to get back on track, vs. a dog that shuts down.

There has to be wins and losses.
The concept of “positive only” training is stupid. If you had an open line of credit at the Bellagio in Vegas and all your losses were covered, it would take all the fun out of gambling. The reason we love to gamble is because it mirrors nature so perfectly. Meaning “positive” training isn’t natural and will never build a passion for learning. Technically, every time the animal isn’t reinforced is a form of punishment, but it’s not sufficient to create a manic devotion.

Be willing to let the sucker win some big jackpots. Many people are reluctant to inject big jackpots because they are cheap. The want to stretch a training session like an all-night game of Texas Hold ’em. If you get 10 excellent minutes of training from a dog because you gave a jackpot in the middle, do it. The alternative runs the risk of making training a drudge. The game is supposed to get your blood pumping!

2 thoughts on “Keeping the Sucker in the Game: A Poker Lesson for Dog Trainers

  1. I’ve just been turned on to you from a older podcast of Jeff Gellman’s I listened to today. I’m a new be in dog training but have trained and or worked with all my dogs and many fosters. My wife and I have been told we should do it professionally by the many people that we have helped along the way. If I can work with what I love and fight against what I hate at the same time (animals being given up to shelters) seems like a great plan to me. Your insight of the animal dog mind is very intriguing. The more I read the more I want to keep reading in hopes it will all take root. Thank you for your thoughts and taking the time to make them available for people like myself.

  2. Adam, good luck in your journey. Before you go any further I would recommend that you work in a shelter as a grunt, not a trainer rescuing dogs. Do it enough so that you get your hands on as many dogs as possible. Do every job no one else wants to do so that you will not be perceived as an alien. If you are not perceived as family you will be left out of important events that will elevate your knowledge. Never mention you want the information to better be a trainer. Assist in euthanasia. If you can’t do that you will never be able to recommend it with conviction when it is the right thing to do. If you are not comfortable around dogs that can hurt you and know how to handle them with calm assurance your safety will be in jeopardy – and the effectiveness of your service. Pick the worst municipal shelter you can find. That way you will have the freedom to learn things without serious scrutiny. They are usually desperate to get volunteers. In the end, that is the best place to learn dog behavior. You won’t have to rely on hearsay from people who are styled experts but actually are simply repeating hearsay themselves. Good luck!

Leave a Reply to Gary Wilkes Cancel reply

Your email address will not be published. Required fields are marked *