Related to: The Comedy of Behaviorism

Behaviorism's gotten a bad rap.

It's gone down in history as the school founded upon the idea that there's no such thing as mental phenomena or cognitive processing, and if there are we can't ever know anything about them, and if we can I don't want to know about it, and if you tell me I will put my fingers in my ears and whistle, and SHUT UP SHUT UP I CAN'T HEAR YOU.

Actually it was more subtle.

The movement did begin with a variation on that principle for historical reasons. John Watson began his work thirty years before the first computer. Information processing still looked like magic; most scientists didn't realize that reductionist accounts of information processing were even possible. Neurons were still "the thing that Spanish guy keeps talking about". Today we discuss the brain by analogy to computers; in Watson's day, they discussed the brain by analogy to their own most advanced technology, mechanical devices. Today we talk about looking for mental programs and subroutines; they sought its gears and levers instead. And just as today many philosophers dismiss consciousness as an epiphenomenon of information processing because computers don't seem to be conscious, so Watson dismissed all mental states as an epiphenomenon of mechanical processing because mechanical devices didn't have mental states.

As science advanced, and as it picked up glimpses of cognition from the Stroop effect and early priming experiments, behaviorism became more sophisticated. Maybe its pinnacle of subtlety came with B.F. Skinner's "radical behaviorism" movement, which accepted inner mental life (which Skinner called "mental behavior") and sought to explain it.

If Skinner was willing to acknowledge inner life, why do we still call his theory behaviorist? It's hard and not especially profitable to define "behaviorism", but if I had to try I'd say it is a methodology that doesn't consider mental phenomena useful as a fundamental level of explanation. So if we want to know why Wanda runs away from a wasp, saying "because her previous encounters wasps have been negatively reinforced" is more useful than "because she felt scared".

And if Wanda herself says "No, I ran away because I felt scared," we shouldn't be especially interested in her opinion: she has privileged access to a certain type of output of the process generating her behavior, but not to the process itself.

Imagine the better behaviorists, if you like, as playing a worldwide half-century long game of Rationalist Taboo, in which you're no longer allowed to use words like "want", "feel", "hope", or "decide". It's overwhelmingly tempting to fake-explain psychology using non-technical non-explanations like "Oh, she just acts that way because she has an overly emotional personality" and so the whole school just promised themselves to root out that way of thinking.

Although the witticism that behaviorism scrupulously avoids anthropomorphizing humans was intended as a jab at the theory, I think it touches on something pretty important. Just as normal anthropomorphism - "it only snows in winter because the snow prefers cold weather", acts as a curiosity-stopper and discourages technical explanation of the behavior, so using mental language to explain the human mind equally halts the discussion without further investigation.

This idea of Rationalist Taboo also explains B.F. Skinner's "mental behavior" loophole. When he discusses thoughts as mental behavior, he's not using them as explanations for other things - not taking the easy way out and saying "The reason I stayed in tonight is because, after thinking about it, I decided I didn't want to go to dinner". He's taking an extra burden upon himself, trying to come up with explanations for thoughts as well as actions.

Behaviorism became less popular in the 1950s after clever experimental protocols allowed more direct measurement of what happens inside the mind, making its taboo on mental occurrences unnecessary and restrictive. Although the philosophical commitments involved became obsolete, the scientific findings remain as valuable as ever. They have entered into the new paradigm as "reinforcement learning", a process widely believed to underlie many diverse mental subsystems all the way from motor coordination to social behavior.

Although reinforcement learning is almost universally known, Skinner's philosophical context for the process is not. He believed that the Darwinian evolution of organisms was just one instance of a wider principle called "selection by consequences", the most successful optimization process in the history of the universe. Evolution can successfully design permanent features of an organism like its skin, claws, and eyes. But it is too slow to fully optimize an organism's behavior, and too large-grained to produce complex behavior on its own. It is is especially too slow and large-grained to produce human-level behavior: citing my sources in MLA format is an important skill, and I don't want to have to wait until ten generations of my ancestors have perished for citing their sources incorrectly before I can do it right.

So evolution conjured up a mini-evolution to serve it. Reinforcement learning is evolution writ small; behaviors propagate or die out based on their consequences to reinforcement in a mind, just as mutations propagate or die out based on their consequences to reproduction in an organism. In the behaviorist model, our mind is not an agent, but a flourishing ecosystem of behaviors both physical and mental, all scrabbling for supremacy and mutating into more effective versions of themselves.

Just as evolving organisms are adaptation-executors and not fitness-maximizers, so minds are behavior-executors and not utility-maximizers. This returns us to the case of the blue-minimizing robot, which executed its program without any representation of a "goal". Behaviorism holds out the prospect of an explanation of human behavior based on similar lines.

Despite its subsumption by the cognitive paradigm, behaviorism continues to hold a special place because of its association with reinforcement learning, as well as its uses in industrial psychology, applied psychology, and various successful therapies including the famous CBT. It's also one of the major inspirations for connectionism, a more modern and exciting eliminativist model which we'll return to later.

This sequence will continue by exploring some of the basics of reinforcement learning in the behaviorist paradigm, and then get into more controversial applications of the theory to explain previously mysterious human behaviors.

New to LessWrong?

New Comment
24 comments, sorted by Click to highlight new comments since: Today at 8:15 PM

If this is to be a sequence, I suggest giving it a distinct tag so that the Article Navigation widget works for it. Also, an explicit name.

[+][anonymous]13y-70

This is tangential, but it's worth noting that even without all this there's a good reason to avoid anthropomorphizing humans - anthropomorphizing something doesn't seem to mean so much assuming something acts like a human, but rather assuming it acts like our naive model of a human, and we know that's wrong!

minds are behavior-executors and not utility-maximizers

May I request that you put this in boldface?

Just as evolving organisms are adaptation-executors and not fitness-maximizers, so minds are behavior-executors and not utility-maximizers.

This web site apparently has "a thing" about bashing utility-maximizers. Biologists use the concept of fitness maximisation as the fundamental unifying principle in biology. It builds animal brains, as best it can - so that they maximise the same function. As such, the idea deserves a hat tip - or you risk losing the baby with the bathwater.

I have a hypothesis about what is going on here: If humans don't maximise utility but machine intelligences do - and if maximisihg utility sometimes leads to utilitronium shockwave scenarios - then that makes machines look bad and scary. Making machines look bad and scary is a major theme around here.

This is the hat-tip to that idea. Evolution/organisms are so good at maximizing fitness that it's really tempting to think they're perfect at it, or at least perfectly efficient at it, which is something we have to be reminded again and again isn't true.

If it wasn't so close to being universally true, we wouldn't have to keep reminding people that it's occasionally false.

This is the hat-tip to that idea.

Yay - this comment is better!

Evolution/organisms are so good at maximizing fitness that it's really tempting to think they're perfect at it, or at least perfectly efficient at it, which is something we have to be reminded again and again isn't true.

Hmm. I can't say I have ever encountered that idea. Organisms are imperfect. They are resource limited. They are cobbled together by natural selection and random mutations. I think most people usually "get" all of that - at least by the "Darwin 101" stage.

minds are behavior-executors and not utility-maximizers

I think it would be more accurate to say that minds are more accurately and simply modeled as behavior-executors than as utility-maximizers.

There are situations where the most accurate and simple model isn't the one you want to use. For example, if I'm wanting to cooperate with somebody, one approach is to model them as a utility-maximizer, and then to search for actions that improve everybody's utility. If I model them as a behavior-executors then I'll be perceived as manipulative if I don't get it exactly right.

Specifically, people don't like to hear an explanation of my behavior of the form "Yes, I reasonably guessed that you would have preferred to get to the grocery store, and I knew that the only grocery store was to the south, but you were driving north so I helped you drive north." Thus, if I model people as behavior-executors, I have a much more complicated game to play because I have to anticipate what they'll do when they discover that I helped them to make a mistake.

That's the difference between trying to save cute pandas and trying to rebalance a broken ecosystem.

To give somebody good advice, you need to understand what they actually want. Then you can introduce an idea into their mind for a behavior which will lead to them succeeding. The customer never knows what he really needs ; if he did, he wouldn't be a customer.

Another tangential: memes are often described as the mental equivalent of genes. if reinforcement-based behaviour are competing and evolving in an ever-changing environment, shouldn't the concept of meme be reduced behaviouristically too?

Doesn't the reduction happen automatically? We have a built-in imitation module (how much we use it is subjected to conditioning), like we have built-in vision and locomotion modules. From this module, we acquire new behaviours from the environment, which we then keep if they're reinforced or lose if they're punished. A meme is a behaviour picked up this way which 1) leads to others picking up the behaviour and 2) is reinforced.

ISTM that the hard part is how we pick up behaviours, but that this is completely beyond the scope of behaviourism - like it explains why we keep walking, but not why we start learning, let alone proprioception and motor control.

Doesn't the reduction happen automatically?

Yes, the theory doesn't need more explanatory power to account for the idea of memes, so it does happen automatically. However it didn't happen automatically in my brain... until you pointed to mirror neurons.

I'm still nagged though by the sensation that maybe "memes" can portray more complex features: think for example of a tune you keep singing to yourself to the point of irritation.

This idea of Rationalist Taboo also explains B.F. Skinner's "mental behavior" loophole. When he discusses thoughts as mental behavior, he's not using them as explanations for other things - not taking the easy way out and saying "The reason I stayed in tonight is because, after thinking about it, I decided I didn't want to go to dinner". He's taking an extra burden upon himself, trying to come up with explanations for thoughts as well as actions.

Hmm. On the other hand, behaviourism is not particularly successful at prediciingi human behaviour. It's something of an easy way out to say "it has causes. but don't ask me what they are".

Why do you say that behaviorism has not been successful at predicting human behavior? Its most popular models of choice behavior consistently account for around 95% of the variance in experimental settings (e.g. the matching law, or the contingency discriminability model). Behaviorist accounts have disproved naive conceptualisations of the "rational agent", and have developed models of self control which not only accurately predict at what point an individual will choose the smaller-sooner reward over the larger-delayed reward, but they also predicted a previously unnoticed behavioral phenomenon (i.e. preference reversal).

I'm aware of no other area of psychology which has been as successful at predicting human (and animal) behavior as behaviorist theories. The success of behaviorist accounts to not only predict, but also to control, human behavior is one of the features why behaviorism is considered one of the most useful approaches to psychology.

Why do you say that behaviorism has not been successful at predicting human behavior?

No form of psychology has been successful at predicting human behaviour. Where are the predictions of election outcomes, or which product will be successful ion the market place?

Your example is a bit absurd - why would the prediction of human behavior necessarily entail the prediction of behavior in a completely uncontrolled environment, with near-to-zero information on any relevant variables?

Your question is comparable to asking: If physics was so good at predicting the movement of physical bodies, then why can't it predict earthquakes? If it can't predict when an earthquake will occur, then it is not successful at predicting the movement of physical bodies.

The point is that we know that humans are incredibly predictable. The reason why humans appear to be unpredictable is simply a product of the vast number of unknown variables acting upon us at any given time. When we remove these variables, and place a person in a controlled experimental environment, the result is highly predictable human behavior. As I pointed out above, we are very good at predicting how people will respond in choice situations, and self-control situations, etc.

The great thing about behaviorism is that not only does it point out that human behavior has causes, but it identifies these causes and quantifies them in simple laws that accurately predict the behavior of individuals. In other words, there's a reason why behavioral science (underpinned by behaviorism) is considered a natural science.

Your question is comparable to asking: If physics was so good at predicting the movement of physical bodies, then why can't it predict earthquakes? If it can't predict when an earthquake will occur, then it is not successful at predicting the movement of physical bodies.

It ipredicts where it predicts and doesn't where it doesn't. If you are going to ask whether something is predictable without adding any riders about to what extent,and under what circumstances , the quesiton would reasonably be taken to apply in the raw, to "free range" behaviour. I would not be taken to mean "under controlled circumsrtances",Of course controlled behaviour is predictable, that is what "controlled" means. And 11 fingered people have 11 fingers.

The reason why humans appear to be unpredictable is simply a product of the vast number of unknown variables acting upon us at any given time.

You don't know that. To know it. you would have to exclude basic physical indeterminism, and the jury is still out on that.

When we remove these variables, and place a person in a controlled experimental environment, the result is highly predictable human behavior.

You get what you are looking for. Ask them to write a story or paint a picture, you do not know what you are going to get..

but it identifies these causes and quantifies them in simple laws that accurately predict the behavior of individuals.

No it doesn't. It predicts the behaviour of controlled individuals. You have no idea what the people around you in the office or street are going to do next.

It ipredicts where it predicts and doesn't where it doesn't. If you are going to ask whether something is predictable without adding any riders about to what extent,and under what circumstances , the quesiton would reasonably be taken to apply in the raw, to "free range" behaviour. I would not be taken to mean "under controlled circumsrtances"

All science only makes predictions within controlled circumstances. If you booby-trapped a billiards table, so there were unknown valleys and crests on the surface of the table, you wouldn't disprove physics because suddenly a physicist can't predict the motion of a billiard ball.

In uncontrolled circumstances our predictions become less accurate because there are literally millions of unknown variables. But, of course, since humans are so predictable, we can make decent enough predictions - for example, we can predict how much time, and to whom, a person will dedicate talking to at a dinner table. We can also predict, with reasonable certainty, when a basketball player will take a 2-point shot over a 3-point shot, or what play a football coach will play at any given time. So whilst our predictions are imperfect, they are still accurate enough.

And this isn't even taking into account applied settings, like cognitive behavioral therapy, and applied behavior analysis - both of which are accurate enough in "free range" settings in order to treat and cure a number of conditions like anxiety, depression, autism, etc.

Of course controlled behaviour is predictable, that is what "controlled" means. And 11 fingered people have 11 fingers.

If I had said "controlled behavior" then you'd have a point. But I didn't, so your comment here is redundant.

You don't know that. To know it. you would have to exclude basic physical indeterminism, and the jury is still out on that.

Physical indeterminism is irrelevant since we're talking about the macro world. If we have to wonder whether human behavior is unpredictable due to some quantum mechanic weirdness, then we have to equally wonder whether billiard balls are unpredictable due to some quantum mechanic weirdness as well. The point is that we know human behavior is perfectly predictable in controlled experimental conditions, and less predictable in situations where some variables are unknown - this necessarily means that the myth of people being unpredictable is a result of ignorance of variables.

You get what you are looking for. Ask them to write a story or paint a picture, you do not know what you are going to get..

Once we can control the variables, of course we do. We can make them write a story or paint a picture of whatever we like.

No it doesn't. It predicts the behaviour of controlled individuals. You have no idea what the people around you in the office or street are going to do next.

Not "controlled individuals", you mean "controlled environments". And yes, of course people in the street are less predictable because we don't have access to relevant variables.

Out of interest, I take it you apply your criticisms of behavioral science to all other sciences? Physics is not successful at predicting the movement of objects because it cannot give me the exact time that a rock balanced on top of some mountain in the Northern Hemisphere will topple over.

All science only makes predictions within controlled circumstances.

Then "human behaviour is predictable" doesn't apply to life in general.

If you booby-trapped a billiards table, so there were unknown valleys and crests on the surface of the table, you wouldn't disprove physics because suddenly a physicist can't predict the motion of a billiard ball.

Predictability in controlled experiments isn't taken by physicists to prove a sweeping statement like "the universe is predictable". Some physical system are well known to be unpredictable.

In uncontrolled circumstances our predictions become less accurate because there are literally millions of unknown variables.

Or some other reason.

But, of course, since humans are so predictable, we can make decent enough predictions - for example, we can predict how much time, and to whom, a person will dedicate talking to at a dinner table. We can also predict, with reasonable certainty, when a basketball player will take a 2-point shot over a 3-point shot, or what play a football coach will play at any given time. So whilst our predictions are imperfect, they are still accurate enough.

Which is to say that some real situation have straightforward rules and rewards, allowing predictability.

Physical indeterminism is irrelevant since we're talking about the macro world. If we have to wonder whether human behavior is unpredictable due to some quantum mechanic weirdness, then we have to equally wonder whether billiard balls are unpredictable due to some quantum mechanic weirdness as well.

Oh we know that, They are.

The point is that we know human behavior is perfectly predictable in controlled experimental conditions, and less predictable in situations where some variables are unknown - this necessarily means that the myth of people being unpredictable is a result of ignorance of variables.

No, it doesn't necessarily mean that.

You get what you are looking for. Ask them to write a story or paint a picture, you do not know what you are going to get..

Once we can control the variables, of course we do. We can make them write a story or paint a picture of whatever we like.

That, again, is getting what you are looking for.

Physics is not successful at predicting the movement of objects because it cannot give me the exact time that a rock balanced on top of some mountain in the Northern Hemisphere will topple over.

No physicist would say physics is successful in predicting without specifying a system. What does "human behaviour is predictable" mean? We already knew you could predict behaviour in some situations, so that isnt a discovery. And we don't know that it is predictable by and large, because it isn't.

Then "human behaviour is predictable" doesn't apply to life in general.

Yes it does since we know that human behavior is predictable. Scientists don't need to add "in controlled environments" because that's necessarily the case. Nothing, absolutely nothing, is predictable in uncontrolled environments. So if you're saying that human behavior is unpredictable because it can't be predicted in uncontrolled environments, then you've simply defined human behavior (and everything) as being unpredictable. How did you describe this kind of argument? "And 11 fingered people have 11 fingers"?

Predictability in controlled experiments isn't taken by physicists to prove a sweeping statement like "the universe is predictable". Some physical system are well known to be unpredictable.

Your analogy is a misrepresentation. Behavioral scientists saying human behavior is predictable is akin to physicists saying that the movement of a falling object is predictable. The fact that physicists can only predict the motion of falling bodies when they know the mass of the object, the force of gravity, etc, does not mean that they have to say, "Falling bodies are predictable in controlled environments, but they are unpredictable in the real world".

Or some other reason.

There cannot be any other reason.

Which is to say that some real situation have straightforward rules and rewards, allowing predictability.

So it's just a coincidence that every "free range" behavior which has been tested happens to be predictable?

Oh we know that, They are.

You think the movement of billiard balls are unpredictable because of "quantum mechanics"?... You do understand that indeterminism and probabilism do not preclude predictability, right?

No, it doesn't necessarily mean that.

There is no other option.

That, again, is getting what you are looking for.

No. Knowing the variables in an environment does not mean you get what you're looking for. There is an almost infinite number of ways in which the variables could be combined to reach different conclusions and predictions. If we come up with a law that uses specific parameters that give us an accurate prediction, then that means the phenomenon we're observing is predictable.

I don't understand why you think controlled environments (i.e. environments where the values of parameters are known) automatically produces some behavior or outcome. That's nonsensical.

No physicist would say physics is successful in predicting without specifying a system. What does "human behaviour is predictable" mean? We already knew you could predict behaviour in some situations, so that isnt a discovery. And we don't know that it is predictable by and large, because it isn't.

"Human behavior is predictable" means that human behavior can be predicted. Informally we might have known that behavior is predictable in some situations, but that's no better than saying "We already knew that the effects of gravity are predictable because when we drop stuff it always goes down". The point is that as long as we have information on the values of parameters, then human behavior is predictable. Pointing out situations where parameters are unknown does not mean human behavior is unpredictable.

Then "human behaviour is predictable" doesn't apply to life in general. Yes it does since we know that human behavior is predictable.

Scientists don't need to add "in controlled environments" because that's necessarily the case.

No it isn't necessarily the case We can imagine Foundation style wide range prediction. (Un)predictability due to large numbers of variable is a contingent issue: it depends on how much computation you throw at it, as in weather forecasting.

Nothing, absolutely nothing, is predictable in uncontrolled environments.

Not even the examples of real life prediction of human behaviour you mentioned? Not even the positions of the planets in the solar system?

So if you're saying that human behavior is unpredictable because it can't be predicted in uncontrolled environments, then you've simply defined human behavior (and everything) as being unpredictable. How did you describe this kind of argument? "And 11 fingered people have 11 fingers"?

Not analogous: an uncontrolled environment is not a special environment that is designed to force unpredictable behaviour. It is a general environment that is not designed for anything.

Your analogy is a misrepresentation. Behavioral scientists saying human behavior is predictable is akin to physicists saying that the movement of a falling object is predictable.

Behavioral scientists saying human behavior is predictable is akin to physicists saying physics is predictable. Physicsts saying that the movement of a falling object is predictable.is akin to behavioural scientists saying the behaviour of road users or game players is predictable.

The fact that physicists can only predict the motion of falling bodies when they know the mass of the object, the force of gravity, etc, does not mean that they have to say, "Falling bodies are predictable in controlled e environments, but they are unpredictable in the real world".

"falling objects" are predictable because they are falling--to fall is to be under the control of one force.

Or some other reason.

There cannot be any other reason.

Yes there can. Physical indeterminism that effects humans is logically possible.

So it's just a coincidence that every "free range" behavior which has been tested happens to be predictable?

If it's not predictable in the free range, that doesn't mean much. Or, rather, it doens't mean what it seems to mean.

You think the movement of billiard balls are unpredictable because of "quantum mechanics"?... You do understand that indeterminism and probabilism do not preclude predictability, right?

For some value of "predictability". Weaker claims are easier to defend, but they mean less.

There is no other option. [than lack of knowledge of variables]

Indeterminism means even Laplace's Demon can't predict. That's definitional

No. Knowing the variables in an environment does not mean you get what you're looking for.

That you don't look for the unpredictable means you get what you look for.

There is an almost infinite number of ways in which the variables could be combined to reach different conclusions and predictions. If we come up with a law that uses specific parameters that give us an accurate prediction, then that means the phenomenon we're observing is predictable.

I don't understand why you think controlled environments (i.e. environments where the values of parameters are known) automatically produces some behavior or outcome. That's nonsensical.

I don't think that. I think that if you asked someone to write a story with rewards for originality. you would get unpredictable results. What I object to is the sweeping, uncontextualised nature of "behaviour is predictable"

"Human behavior is predictable" means that human behavior can be predicted.

That's ambiguous too. Some of the time?All the time?

The point is that as long as we have information on the values of parameters, then human behavior is predictable.

And all the other variable are being held constant. Which they never are in "free range" situations. In a sense, there are no causes in free range situations, as there are in controlled environments, because the "other variables held constant" clause doesn't apply. It is a mistake to think that you can sum one bit of controlled-environment causality against another and get even more causality. .

It looks like there are two definitions of controlled environment here. Maybe taboo it?