Real Man’s Guide: Positive Reinforcement

Last time in the Real Man’s Guide I introduced some of the basic aspects of learning theory, and promised to go over examples of how to use this stuff.

You’re probably reading this on a web browser written in C++. C++ (pronounced see-plus-plus) has 63 “keywords” and about 30 or so arithmetic and logical “operators.” Compare this to the roughly 600,000 words in the last edition of the Oxford English Dictionary! With less than 100 basic building blocks, C++ can produce Firefox (for better or worse), Chrome, Internet Explorer (definitely worse), and most of the software you use every day. You don’t need a large complicated system to get things done. You’re better off with a smaller and simpler one. (As a matter of fact many programmers, myself included, think that C++ could probably lose 20 or so keywords and be the better for it.)

The four basic building blocks of operant conditioning, and thus dog training, can take us a long way toward describing and changing behavior. There’s a lot more going on than they describe, and sometimes they tend to oversimplify, but they still tend to get the job done.

When we want to train a new behavior we use positive reinforcement (R+ for short.) We are adding something (a treat, a toy, praise….) that reinforces the behavior.

This means our dog found whatever we added as a consequence of the behavior reinforcing or, as we mortal humans without white lab coats might say, rewarding. This is the first area where some people get tripped up with positive reinforcement. Just because you think it’s reinforcing, that doesn’t mean your dog will.

For example, 45 minutes into a one hour class, that yummy dog treat might start getting boring. As one of my dog trainer friends is fond of saying, “I love cheesecake! But after an hour it gets a little old.” You have to be willing to reward honest work with appropriate pay.

But the biggest mistake people make with positive reinforcement has to do with a slightly more complicated concept: reinforcement schedules. How frequently someone is rewarded for a behavior is just as important, if not more, than the type of reward. These schedules can get very complicated, but for our purposes we can keep it pretty simple.

If you’ve ever had a job for pay, you have been on a fixed reinforcement schedule. If you were paid every week, two weeks, or once a month you were on a fixed interval schedule. After a set period of time, you were paid.

The pitfall with these fixed schedules is pretty simple. What would have happened (or did happen) when it came time to get paid and you were not? Chances are, you stop working. Immediately.

So, if you give your dog a treat every time she sits for a few weeks, and then stop because “she knows it now and should just do it.” What happens?

Now think about the slot machines at Las Vegas or Atlantic City. We all know that there is no way that those machines can give out more money than is put in. But people stand there for hours, emptying their pockets into them because the machines operate on a variable reinforcement schedule. The players never know when the next reward is going to come. This is very compelling — compelling enough that some people literally cannot stop.

Most trainers will tell you, in simpler terms, to work on a fixed reinforcement schedule when you introduce a new behavior, usually rewarding your dog every time she performs the behavior (technically a “continuous” schedule). When the behavior is “reliable” (the dog does it whenever we ask) move to variable reinforcement. Many trainers will tell you “Become a slot machine!”

Many owners never get out of the fixed reinforcement stage or wait too long, and end up with what I call a “Show me the money!” dog. The dog will not perform unless she can see the treat.

Next week: how to use positive reinforcement to stop a behavior. (Think that doesn’t make sense? Good! You’re paying attention.)