Archive for August, 2011

h1

Negative Reinforcement vs. Punishment — They Don’t Mean the Same Thing!

August 31, 2011

Over at PBS, the folks at NOVA have started a new blog series called “Animal Acumen” where they post interesting stories about animal cognition. Always interesting! The latest article is about the incentive for animals to cooperate in society.

Group Weight Lifting and Tug-of-War Event
Photo by prashant maxsteel

Unfortunately, the first thing I noticed about this post was that in the second paragraph the author makes a mistake that so, so many people make in talking about how animals (including humans) learn. It’s a mistake that makes me wince like I’m hearing nails on a chalkboard: she mixes up “negative reinforcement” with “punishment.”

Since the reply I tried to leave ended up being really long and their commenting system seems to be down, I figured I could just turn the comment into a blog post of my own.

Below are the first couple of paragraphs from the Inside NOVA blog post:

Why do most of us work together for the common good, even when it might personally benefit us more to cheat? Is it because we fear we would be punished for not following the rules? Because we have been taught to treat others as we would want to be treated?

Because cooperation is vital for survival in a variety of contexts, it may be that fair play is an evolutionary adaptive behavior, scientists say. Of course, the flip side is that when an individual does behave selfishly, others often use some form of negative reinforcement to dissuade such behavior in the future. [emphasis added] It appears animals do the same. First, though, an animal must recognize unfair situations.

Here’s the gist of what I tried to write on the Inside Nova blog:

In animal behavior language, the phrase “negative reinforcement” means something specific — in fact, it means the opposite of what is meant in this context. “Negative” means “absence or removal” not “unfavorable”, and “reinforcement” means “something that causes the behavior to increase.” Therefore “negative reinforcement” is taking stimulus away in order to *encourage* behavior. What the paragraph is trying to express (i.e., other members of the group provide “feedback” that dissuades repetition of the undesirable behavior by an individual) is actually “positive punishment”.

I know in popular speech people say “negative reinforcement” because it sounds less harsh than “punishment”, but that’s just wrong; reinforcing something you’re trying to discourage will backfire and thus would be a social strategy that groups of animals would quickly stop doing.

The word “punishment” in the realm of animal behavior science doesn’t have a bad connotation (it simply means “something that causes a behavior to occur less frequently”). In fact, “negative” doesn’t have a unfavorable connotation either, funny enough. Negative isn’t bad, it’s simply absence/removal of stimulus (which could be desired or not desired by the subject). Positive punishment isn’t the oxymoron it sounds like either — remember, in behavior lingo “positive” does NOT mean “good.”

wild parrot Photo by artolog

Words Matter…  aka. Why You Should Care

So why is it such a big deal that everyone always gets this wrong?

I know I’ve been flamed on parrot blogs for getting my knickers in a bunch that other members used “negative reinforcement” and “punishment” interchangeably. Actually, I think it’s more fair to say I was flamed for getting my knickers in a bunch that members continued to willfully and obstinately continue to do so, on purpose, after it’s been explained what the difference is.

So, back to the question: Why am I such a stickler? Because when people are trying to solve behavior problems and they are given the steps and the tools for solving those problems, it matters that they understand the concepts and the words used to refer to them.

If an animal behavior export or trainer begins to tell their client “avoid using punishment” and the client responds “oh, I would NEVER punish my bird! Never! I only use positive reinforcement.” … should you trust the client understands what “punishment” means? Nope! In my experience, as soon as you say the word “punishment”, people go on the defensive and claim they don’t do it. But they don’t take the time to really understand that just because they don’t mean something to be punishment, or just because they have good intentions, that this doesn’t mean that they aren’t *discouraging* the animal an action. You don’t have to hit an animal to “punish”. Punishment can be something as simple and tame as a stern look (which works wonders on dogs, by the way!).

It bears repeating: just because YOU don’t think something is punishment, doesn’t mean it’s not. What matters is whether the stimulus decreases the likelihood of the behavior in the future.

For example, let’s say that you’re trying to tame a fearful bird. Every time the bird gets a bit courageous and comes closer, you try to reward it by petting it on the head. That’s positive reinforcement, right? Well, no. It’s only “positive” in the sense that you’re “adding” stimulus immediately following the behavior, but despite your good intentions, if the bird doesn’t like being touched, it’s not reinforcing at all.

In fact, if the bird learns that you’re going to try to touch it every time it comes close to you, and it doesn’t like that, then it’s going to stop coming close to you to avoid being touched. See? That is, in fact, the definition of punishment. It doesn’t mean you’re evil for wanting to show your pet affection, but if you don’t want to accept that you’ve been “punishing” your bird the desired behavior simply because you don’t like the word, it’s going to be difficult for a behaviorist or trainer to help you do a better job working with your bird.

Did I just say petting is punishment? Of course not!  It depends on the context. If your pet likes being touched and will do things in order to get petted more, then petting would be a reinforcer.

84..365 | [Explored] Angry Birds
Photo by KatKamin

So what about negative reinforcement?

But we still shouldn’t use negative reinforcement, right? No. Again, negative does not mean bad. There are of course legitimate reasons why a trainer might need to pull “negative reinforcement” out of his/her bag of training tricks. You use negative reinforcement by removing something (undesirable) from the animal’s environment in order to encourage a particular behavior.

But if you’re so wedded to the notion that “negative reinforcement” is just a nice, more PC way to say “punishment” than you’re going to have a very difficult time understanding conceptually when and how to use actual “negative reinforcement”.

It helps to remember that negative reinforcement is a type of reinforcement. Does adding or subtracting this stimulus encourage the behavior to happen more? Then it’s reinforcing.

If you don’t understand the terminology (or insist on using terms in the exact opposite of what they mean), how will you be able to truly understand and apply the principles that they describe?

Is it really critical to understand the scientific principles behind training in order to do some basic training? Probably not. But if you have an animal with serious behavior problems that really need to be solved, or you’re giving advice to someone who does, I  encourage you to learn about the terms that professional animal trainers and behaviorists use and how to apply them. I promise it’ll help you put together a better strategy for fixing problem behaviors.

Too long; didn’t read?

Negative reinforcement = encourages behavior
Punishment = discourages behavior
They don’t mean the same thing. Don’t use them interchangeably.

Edited Sept 2: Clarifying that you punish or reinforce behavior, not subjects.

Advertisements
h1

Animal Intelligence: Kea vs. Crows — Which is Smarter?

August 22, 2011

Which is smarter, a parrot or a corvid?

Actually, that’s a trick question.

As this excellent article in Discover Magazine points out, you can’t boil animal “intelligence” down to a simple measure of how an animal performs on an arbitrary, human-designed test.

Animal intelligence isn’t a single thing. There is no standard IQ test for them to sit, and no universal checklist of skills to score them against. Instead, animals have evolved mental abilities to cope with different lifestyles and environments. Many early studies into animal intelligence simply looked at whether animals could or couldn’t perform specific tasks. But it’s far more interesting to see why and how they do different things, and how their own particular brand of intelligence has evolved.

In nature, only a few animals have been observed to use tools, likely because their “natural” lives don’t really require it. But last week, one more animal joins the ranks of “tool user”: Kandula the Elephant uses a stool to solve a puzzle. (But how often do elephants in the wild need to reach treats suspended in trees while simultaneously having access to out-of-sight stools?) The interesting thing in this development isn’t really the use of tools, but the planning undertaken by the pachyderm.

What are some surprising actions you’ve witnessed from an animal that you thought showed “intelligence”? What kind of problem-solving chops do your pets demonstrate? What’s the most amazing feat of animal intelligence you’ve ever heard about?