Tuesday, May 27, 2014

Schedules of Reinforcement


Discussions about training--animals, humans, aliens, spouses--always come around to the idea of positive and negative reinforcement.   It's pretty much a given that most people understand that if the trainee does the right thing, whatever that might be, that s/he/it deserves to be rewarded.   Most folks don't get up and go to work in the morning unless there's something in it for them.  Yes, even volunteers get something for their trouble, usually in the form of satisfaction as defined in their personal story.

There are fine points that are often overlooked.  The schedule on which the reinforcement is delivered is critical to the success of the training or learning effort.  The article above explains all of the options in a clear and concise manner, so I highly recommend that you print out a copy and staple it to your partner's shirt for reference.  Or if you're training non-shirt-wearing animals, a copy tacked to the wall near your work space will do just fine.  Do it in a BIG font so you can see it from the floor where Fuzz Butt laid you out when you tried to explain the "give me your right hoof over here on your left side" behavior.  You won't want to screw up the schedule just because of a concussion or slight coma.

When you're starting a new training effort, it's imperative to have a specific goal in mind.  If you want Fluffy to move a particular body part, and he does something else, no matter how cute that other behavior might be, no reward should be forthcoming.  You have to focus as much as you want him to.

Of course, this is the point where I always bring up using a clicker or some other attention-grabbing device to indicate to the horse both 1) that training is about to occur, and 2) he did the thing you asked for at exactly the moment you clicked.  The payoff comes after the click.

First, though, you have to be sure you've trained the horse to understand that clicking has a meaning.  The simple way to do this is to walk up to him (leave him in his stall so he doesn't wander off, but be sure he can reach over the door or stall guard so you don't have to open the door to reward him), say his name, tell him something he's used to hearing like "Good boy!", then click followed by the specified reward.  I like food rewards.  My horses, cat, late dogs, partner, and child agree.  But do what you want that suits your belief system.  Research has clearly shown that food works best, but it's your trainee.  You'll come to that conclusion on your own when you find it's taken you six months to train him to lift his foot by rewarding him with glimpses of the Horse Illustrated centerfold.  Believe me, he'll only be lifting his foot to stomp it down on your toes so you'll leave him alone.
As you've read in the linked article, when you're trying to teach something new, rewarding the behavior every time it's offered will work best.  You don't need to fret over what happens while you're sleeping and he's in the pasture repeatedly lifting his foot.  He'll get over it.  He needs to connect your presence to the process anyway or it's not really a trained-in stimulus/response situation.  You stimulate with the cue you've chosen, and he responds.  Bingo Bango, the kid stops throwing his socks on the floor under the couch.  Yay!

Maxine, not entirely engaged in the training process.
Once you're getting a goodly number of correct responses, that's the time to cut back to fewer rewards and more repetitions required to earn them.  He'll be okay.  Don't feel bad for him.  Don't indulge in Guilt Cookie-ing.  Move on.

Now that he's got the picture about learning how to learn and what the deal is going to be, you can introduce other behaviors one at a time.  It's important to keep them separate at first until each behavior is thoroughly ingrained.  If you add too many at once--like trying to teach your dog to do the Boot Skootin' Boogie when he's only just figured out how not to pee on the floor--you'll just confuse him and, in turn, you'll get frustrated.  One step at a time.  Baby steps. That's the key to real training.

A trainer once told me that a horse has to repeat the behavior correctly at least 100 times before he can be considered trained in that behavior. And I can guarantee that on effort 101, he'll mess up.  Just start counting again.  The end result you're looking for is absolute recall of the behavior connected to the command.  So don't think that just because he got on the trailer three times in a row this afternoon he's trained to trailer load.  Nope.  Not even.  If he does it ten times in a row for ten days in a month, then again two months later, then you're on the road to his being trustworthy to load without incident.  Or pick up your son's socks off the floor.  Or whatever behavior floats your boat.

It's a little deceiving to watch demos at horse expos and gatherings. The clinician/trainer usually shows off his horse's skills then has some other person bring a horse in and do the same thing.  The result is often that the new horse seems to pick up the behavior miraculously quickly using the Big Name Trainer's method.  But unless his owner takes him home and can get the same behavior with the same cue 100 times without the BNT present, it's not real learning.  It's showmanship.  Party tricks.

Aim for the best, and you'll get it.

No comments: