Thursday, January 8, 2009

Premack’s Principle

What is the Premack Principle, Anyway?
If you’re at all interested in behavior modification, you’ve heard of David Premack. He is a psychologist who came up with the relationship between desirable and undesirable behaviors.

Desirable – or high probability – behaviors are those behaviors which the animal wishes to do when given the choice. Undesirable – or low probability – behaviors are those behaviors which, given the choice, the animal seldom, if ever, does.

Premack’s Principle states an animal will perform an undesired behavior in order to engage in a desirable behavior. When a high probability behavior is contingent upon the performance of a low probability behavior, the outcome is the increased frequency of the low probability behavior.

While this might sound hard to enforce, consider the mother who tells her child, “you aren’t allowed to go outside and play until you clean your room.” Cleaning her room is not a preferred behavior for the child, yet since she is extremely motivated to play outside, she will perform the low probability behavior for her “reward”: the high probability, or preferred, behavior.

Training Dogs Using Premack’s Theory
In relation to dog training, Premack’s Principle tells us we can use our dog’s most desired activities as reinforcers for good behavior in operant conditioning. Often, trainers will become frustrated when their dogs lose interest during a training session because they’d rather sniff the floor, play with another dog, chew their favorite toy, etc.

That’s okay! You can use these tendencies to achieve the behavior you want. For example, if your dog would rather fetch a ball, you can use it as a reinforcer during training. If you’re working on “sit” with your dog, wait until she sits, then release her as you’re throwing the ball. Of course, this works best with dogs who are so motivated to play fetch they always bring the ball (or stick, favorite toy, etc.) back to you every time! Mostly, with dogs that enjoy this activity, they understand the principle: I chase the ball, bring it back and the human throws it so I can chase it again!

Or, if your dog would rather sniff around than receive some other form of treat during a training session, wait until your dog performs the behavior you wish, and then let him sniff as his reward! After a few seconds of sniff-time, calmly say “let’s go,” or get your dog’s attention back on you and start over.

I know a dog trainer who used her male dog's extreme desire to mark everything as reinforcement. She taught him to heel at perfect attention merely by waiting for him to perform desired behavior and then rewarding him with allowing him to scent mark. VoilĂ ! Premack's Principle at work.

Premack theory works hand in hand with intermittent reinforcement, allowing you to wean your dog away from “treats” as rewards. Not only that, it allows you to utilize your everyday environment to train your dog. An important thing to always keep in mind is "what motivates my dog the most right now?" If the Premack Principle is used correctly, you’ll quickly have a dog which is always looking for ways to “have fun” by offering you obedient behavior!

Arguments Against “Clicker Training”

Unfortunately, most clicker trainers teach their students to treat their dog each and every time a behavior occurs, or they don’t properly outline how to reinforce intermittently, thus moving away from reinforcing each occurrence. This is the biggest argument against clicker training: people don’t want a dog dependent on receiving treats in order to be obedient. Tied in with that argument is a trainer who doesn’t want to "carry treats around" all the time. They’ll argue, “why does the dog needs a treat every time?”

When trainers use operant conditioning – in which clicker training is based – they not only wean their dogs off of “treats” by using intermittent reinforcement, they also substitute other forms of reinforcers. This means you don’t have to – and actually shouldn’t – carry treats or toys with you wherever you and your dog go.

Another argument against clicker training: the dog’s behavior is dependent on the presence – or absence – of treats. While it’s hard to use much other than dog treats in an actual dog class setting, many clicker trainers fail to teach their students the many types of positive reinforcers, and how to use these other forms of motivation to modify dog behavior outside of the classroom. Also, done correctly, operant conditioning does the opposite of training the dog to only work when "treats" are present!

For example, simply telling your dog “Good dog!” can be reinforcement. Whether or not it's a strong enough reinforcement for your dog depends on if you've correctly conditioned the dog. Also, behaviors themselves can be reinforcing. According to Premack’s Principle, a dog will perform a less desirable behavior in order to be able to perform the more desirable behavior. The dog who is excited to enter the dog park, for instance, must sit calmly and look her trainer in the eyes in order to be able to advance into the park. Although the dog is extremely motivated to play with other dogs, she knows she must offer the obedient behavior (sitting with attention) in order to “earn” the “treat” of entering the dog park.

This is not to say that a trainer shouldn’t use clicker training! Rather, when researching trainers and dog classes to begin on your positive reinforcement journey, be sure to do your homework and find a reputable establishment which utilizes operant conditioning methods correctly. This means you probably should not attend the PetCo/PetSmart dog classes, since the majority of the time these trainers don’t utilize clicker training methods properly.

Intermittent Reinforcement

Positive reinforcement (a common slang term for operant conditioning) is the basis of clicker training. Unfortunately, not all clicker trainers understand the use of reinforcers, namely, when to give them to the dog and when NOT to give them.

Reinforcement Examples
First of all, the trainer needs to evaluate the dog to find out what most motivates him. While dog treats are the most commonly used reinforcer, not all dogs respond the same way to treats. Often dogs won’t respond as well to commercial dog treats as actual meat, so be sure to try various types of food: roast beef, cooked hamburger, hot dogs, salmon, etc.

Of course, the dog might not be as interested in food as she is in playing – fetch or tug-of-war, for example. Anything the dog will work hard for is a reinforcer, and can be used in training to modify behavior. There are many motivators in a dog's life, and all dogs are different, so it's best to evaluate each dog for a list of reinforcers.

Once you’ve conditioned the dog to the meaning of the clicker (by pairing the "click" with a "treat"), you can begin modifying his behavior. The whole premise of clicker training is to create a dog who offers behaviors to the trainer, rather than being forced into them – the dog is the one who is “empowered” to make decisions and wants to work with his human.

Intermittent Reinforcement: Reinforcing Only the Best Behaviors
In order to train correctly and have an obedient dog, a trainer must utilize reinforcement correctly. When the dog is first learning a behavior – “sit,” for example – every instance of the correct behavior (the dog sits down completely) is “treated,” no matter how slowly or sloppily she does it. However, after the behavior is consistently occurring, the trainer needs to be sure only to reinforce the top 80% of the behavior. I.e., those times the dog best completes the behavior, and quickly.

It's also important to note that dog trainers should only work on one behavior at a time. If you're working with your dog on "sit," then work only on that cue for a session. Take a break before moving onto any other cues. This speeds the learning process for the dog.

What Does Constant Reinforcement Actually Teach?
By reinforcing your dog for only her best performances ensures the dog learns HOW you want the behavior; not just that you want any form of it. Dogs which are constantly treated for a behavior actually don’t learn to perform the behavior well. In fact, these dogs for the most part don’t even offer the behavior fully or consistently. For example, with a “sit,” the dog might only crouch down so her haunches are near the ground; not on them. With a “down” cue the dog’s elbows might not even touch the ground. Even if the dog downs correctly, she might take her time doing it.

Why does the dog not perform the cue consistently or to the full completion? Because the dog gets paid every single time the behavior occurs, regardless of the precision or timing. (In real-life terms, think of someone who rides the bus to work. Whether they get there early, on-time, late, the bus waits for them and takes them where they need to go. There's no motivation for getting to the bus-stop on time because the person knows the bus will still be there. If, however, the bus kept its schedule this person would begin arriving on time to get a ride.)

By reinforcing only the dog’s best behaviors, you’re telling the dog, “if you don’t work hard enough, you don’t get paid.” Because you’ve done your research and know which reinforcers your dog finds most appealing, the dog WILL work harder if he doesn’t get paid for something. After the dog is consistently giving you his best behavior, then you begin decreasing the reinforcement – or intermittently reinforcing – for the behavior. There shouldn’t be a schedule to the reinforcement, that is, the dog shouldn’t know that every 3rd or 5th occurrence of the behavior earns him a “treat.”

Because the dog never knows when he's going to be reinforced, he'll consistently offer the correct behavior. This is where many clicker trainers fail to understand the true significance of operant conditioning. They don't understand how to get the behavior happening consistently, then move from consistent reinforcement to intermittent by only rewarding the dog's best behaviors, and they also don't understand the true meaning of reinforcers.

There are many trainers who disregard clicker training because of this very misunderstanding of the true nature of the training methods.

Positive Reinforcement

Choosing a Positive Reinforcer
When using operant conditioning or clicker training with dogs, the trainer first needs to evaluate the dog to find out what he wants most. Anything a dog will work to achieve can be considered a reinforcer. Two things to keep in mind: reinforcers are different for each dog; what WE as trainers think of as 'rewards' for the dog might not be reinforcers*. What motivates one dog might not be important to another. Some dogs are extremely motivated by food, while others prefer play and toys; some dogs merely want attention from their trainer. There are many different types of reinforcement; you must choose the one which is best for your dog.

Main Types of Reinforcement for Dogs
Food: When you’re working with a food-motivated dog, it’s best to ensure you’re using food which the dog cannot resist. While you can certainly train the dog at feeding time – thus using the dog’s actual meal for reinforcement – this won’t have as strong an effect at any other time.

Use the smallest piece of food your dog will work towards; be sure to have soft, quickly-chewable pieces versus big, crunchy treats. The quicker the dog can consume the treat, the quicker he’ll turn his attention back to you instead of taking his time hunting around for the crumbs of the one you just gave him.

If your dog doesn’t seem interested in commercial dog treats, you can try cooked meat, such as hamburger, roast beef, etc. Don’t worry about the dog beginning to beg for “human” food, since the only time this food is used is when training. A dog learns to beg for food only when this behavior is reinforced!

Within the food category, you'll find your dog will have favorites that he'll work for no matter what - whether he's just eaten or he's chewing on a favorite toy, etc - these are your "high-reward" reinforcers, and should be utilized when you're training for very important cues. An example would be the recall, or "come!" cue. Many dog trainers believe this to be one of the most important cues to teach a dog; if it is for you, then you should reward the behavior with an extremely high "paycheck"!

Play/Toys: Many highly energetic, prey-drive dogs actually prefer this method of reinforcement over treats. This can be anything from the dog’s favorite toy to playing tug-of-war, fetch, or any other game the dog loves to play. The trick here is to keep the play sessions very short – no more than 30 seconds – so you leave the dog wanting more. Also, it keeps both of you on task. Again, keep in mind that certain games or toys will motivate your dog more than others.

Attention/Touch: Some dogs crave attention over anything else and a trainer can certainly use this as reinforcement. Petting, scratching, or any other form of attention the dog loves can be used as reinforcers. Again, the key is to keep these sessions short when using them for training. Also, it's good to know in which places your dog really likes to be scratched or petted, versus the places where it's just simple enjoyment. For example, a dog might like getting her belly rubbed, but she absolutely loves having her ears scratched. Keep these things in mind when you're reinforcing for behaviors...the bigger the "duty," the bigger the reinforcement should be.

Similar to petting or scratching, some dogs will work for merely a “Good dog!” or other form of verbal praise. When clicker training or using any other form of positive reinforcement training, it’s important to understand eventually the trainer needs to move away from “treating” the dog every time – this is called intermittent reinforcement – and instead move towards more verbal praise with intermittent reinforcers (such as praise, play, food, etc.).

Territory: Last, but certainly not least, is the use of space as a reinforcer for your dog. Let's look at an example to explain this: think of the puppy who constantly pulls on the leash to get somewhere else. A good dog trainer can actually use this behavior to his/her advantage in order to form new behaviors. Standing quietly and calmly, the trainer waits until the split second in which the pup lets some slack creep into the leash - CLICK! - and reinforce this non-pulling behavior. After a few repetitions of receiving payment for not pulling, the puppy will figure out "if I don't pull, I get rewarded!" While you'd want to start in small enough increments - and with the least amount of distraction - in order to give your puppy a chance to succeed, using space and territory within operant conditioning will work if the trainer is consistent and has good timing.

Reinforcers Are NOT Created Equal
Another important note to make is that not all reinforcers hold the same motivational factor all the time for a dog. For example, if a dog is highly food motivated but has just eaten, she probably won’t work as hard to obtain food – even if it’s her favorite treat you’re using. Also, motivation to work for reinforcement can be affected by the presence of another reinforcer. A dog who will work for both food and play will default to one or the other when both are present. A dog with high prey-drive, for example, will often ignore food in preference of their favorite game.

No matter which reinforcers work best for your dog, keep in mind the proper way to train – to get the best, longest-lasting results – is not to treat for every single behavior all the time. Even when using intermittent reinforcement, you don’t want your dog to know there’s any type of schedule to the reinforcement.

Last, a trainer should always be thinking outside the dog training box. The aforementioned “popular” methods of positive reinforcement are effective, but don’t forget the Premack Principle – situations and the environment can be used as reinforcers, too!

*Think of a 'reinforcer' as something the DOG wishes to obtain/is motivated to work for, while a 'reward' is what the TRAINER thinks the dog wants. Seems confusing, doesn't it? If a trainer has done her homework, she knows the reinforcers her dog most enjoys, and when to use them, while a trainer who hasn't evaluated the dog's motivational factors will think "my dog will work for these" when in fact, the dog might not want to!

Thursday, January 1, 2009

Clicker Training Dogs

The past 10 years or so has seen a dramatic increase in a method of dog training called clicker training. "Clicker training" is actually jargon for the lesser-known scientific terminology, operant conditioning. While the concept certainly isn't new - it was developed under B.F. Skinner's tutelage as early as World War II - it seems to have swept the dog training world as a highly effective training method to achieve long-term effects.

Operant Conditioning
In operant conditioning, consequences are used to modify the frequency of behaviors. Generally speaking, animals will continue performing a behavior which results in a “positive” outcome, and discontinue behaviors which result in either no consequence or a “negative” outcome. In this article, "positive" and "negative" refer to the emotional connotation. In clicker training, trainers use positive reinforcement (the addition of a reinforcer) to get the dog to repeat desired behaviors as well as decrease undesired behaviors.

In order to get the dog repeating a behavior, she must know exactly which behavior earned the reinforcement. Through the use of a clicker, the trainer can “tell” the dog the instant a desired behavior occurs (this is called marking the behavior). This tool is used because of its sharp, quick sound – the click – which is much more precise than the trainer’s voice. It takes a lot longer to say “Good boy!” than it does to simply click. Once the dog understands the clicker signifies she’s done something good, the trainer can begin shaping behaviors.

Conditioning a Dog to the Clicker
The first step in beginning this training is conditioning the dog to the meaning of the clicker. To start, the trainer must find a motivator for the dog – something this specific dog greatly wants. The most popular form of reinforcement is food in the form of dog treats since all dogs, at some point, are food motivated. If your dog isn’t particularly interested in treats, there are many different positive reinforcers you can use. The important thing is to find what motivates your dog at that moment.

With the dog on leash or in a small space (in order to minimize distractions and the dog’s ability to leave), the trainer clicks the clicker, pauses, then “treats” the dog, regardless of what he's doing. This step is repeated several times in a row in order to form the association between the clicker and the positive reinforcement. At this point, the dog is simply learning that after he hears a click, he will receive a reward.

Getting Started Clicker Training
Once the association between clicker and positive reinforcement is made, you can get started with the actual training. At this point, the dog knows that she needs to hear a click in order to receive the reinforcement. Since she's still very new at this, start with something simple so she can succeed. You can begin with waiting until the dog looks away, and when she turns back and makes eye contact (as she eventually will), click, pause and then treat her. Repeat this step - dog looks away; dog makes eye contact, CLICK, pause, treat - until the dog is consistently looking at you. This teaches eye contact and attention but also ensures the dog learns the ultimate goal is NOT the treat; rather, working for the click itself.

When the dog is consistently making eye contact, start lengthening the amount of time he needs to hold your gaze before clicking. Count to 2 seconds, the next time wait 5 seconds, etc. This teaches the dog he needs to HOLD eye contact. Whatever you've read or heard on the Dog Whisperer about humans needing to "win the staring contest" doesn't apply here. Wouldn't you rather have a dog eagerly looking into your eyes for the next cue, rather than a dog who is afraid to hold eye contact?

Leave Your Dog Wanting More Training
For every training session, it’s best to leave the dog wanting more, rather than working with the dog until she gets bored with the training. In the beginning, 5-minute sessions are best. Work with the dog, then give a “release” word and take a break to pet or scratch the dog. When you’re ready to start back in, merely stand still and wait until the dog makes eye contact again, or you can call the dog’s name to get her attention. When you're done with the training session, make sure to end on a positive note.

Remember, this is a game and should be treated as such – it’s fun but you and your dog are learning at the same time!

Yellow Puppies Blogger Template | Template Design | Elque 2008