Behaviorism Redux

December 31, 2010

Its proponents were right all along – behavior drives psychology.

In the distant past, dinosaurs roamed the earth. Almost as long ago – or so it seems – the prevailing approach to psychology was behaviorism, which studied the conditioning process through which organisms learn to respond to stimuli in the environment. Because thought processes can’t be directly observed and measured like behavior can, consciousness and the mind were pretty much left out of the picture.

By the 1980s, researchers had figured out lots of ways to make mental processes accessible to scientific methods, reducing the opacity of the black box that cognition had been. Mentalistic terms like purpose, importance and mood, banned from the behaviorist lexicon, became legitimate constructs for psychological theories, allowing researchers to better understand what makes us tick.

We don’t talk much about stimulus and response any more, perhaps as a reaction against psychology’s long obsession with them. But we still live in a world of causes and effects in which our life outcomes depend to a large degree on how we choose to behave (or not to behave) in response to the environmental situations we encounter. The mind’s main job is to choose the best responses to stimuli.

Cognition is not an end in itself, but a means to the end of action. The most important questions confronting organisms are not about environmental states (what is that thing, what are its properties, what relationships does it have with other things), but what we should or shouldn’t be doing right now. Answers to the former kinds of questions are useful only to the extent that they inform the latter (excepting the case of effects that are caused directly by thought, and do not require behavior to cause them).

Cognition, in other words, is the handmaiden of behavior. We think (primarily) in order to do. The task for which our information processing system evolved is the derivation, from a set of environmental inputs, of the behavioral response that maximizes the probability of the occurrence of an effect that contributes to our survival (or minimizes the probability of one that threatens it).

It’s in our interest to be good at casual attribution – knowing our environment’s cause-and-effect relationships, particularly those with implications for our lives. When the causal relationship is simple and unambiguous, the mental processing required to map a stimulus to a response is minimal and quick, and the probability that the desired effect will result is high, as when we jerk our hand from a hot stove. (We have a term for such a low-level behavioral decision: reflex.)

But the world is a complex place, and causality is often messy, with many factors interacting in complicated ways to cause outcomes. Some problems are beyond our computational power, and we have limited access to the data needed to fully understand (i.e., mentally model accurately and in full detail) some causal systems, leaving many of our solutions underdetermined. We don’t have the option of passing on life problems that are too hard for us to solve with a high degree of accuracy, because failure to act is itself a decision. We have to go with our best hunch.

Starting before we’re even born, we move through a continuous stream of rapidly changing stimulus fields that require us to dynamically adjust our behavior to maximize the good and minimize the bad. To make those adjustments, we monitor the events and states that follow our actions and those of other people and things and make inferences about what caused them. Our lives are one long operant conditioning experiment.

Whether and in what direction behavior is reinforced by what follows it depends on our perception of effects and on the causal attribution made by cognition. Suppose, for example, two homeowners with lawns of exactly the same type and condition buy Turf Titan lawn thickener and apply it the same way with the same (objectively measured) result – their lawns get a little thicker. Homeowner A notices the difference and is positively reinforced by it: he becomes a regular customer. But Homeowner B has one of those annoying neighbors with an immaculate lawn, and against its perfection, he doesn’t notice his own lawn’s modest improvement. “Doesn’t work,” he mutters as he tosses the half-used bag in the trash.

Hold on, you might protest; it’s not a fair comparison – if this were an experiment, you would have controlled for the adjacent lawn to prevent a contrast effect from biasing one subject. But that’s the point – we behave in the natural world where parameters vary freely, their camouflage providing many places for causal relationships to hide from cognition.

If we misattribute or fail to detect many environmental effects, though, we must get some of them right. Our species wouldn’t have survived if we weren’t at least moderately good at detecting the signal of nature’s cause-and-effect relationships against the noise of competing confounds. Objectively existing forces – in both the natural world and the social one – impose effects on us whether we understand them or not. Those forces shape our behavior, but only to the degree that cognition’s causal attributions are accurate.

We might call this the “The truth will out” threshold (though it’s really more of a gradient, because our mental models correspond to their objective referents in degrees). If the signal is strong enough and/or the noise level is low enough, we learn the causal relationship, and behavior is positively or negatively reinforced appropriately. Example: feeling violently ill moments after eating poison berries, leading to the (probably rapid) extinction of that behavior.

Suppose, though, that the berries are only mildly toxic, and it takes many hours for symptoms to develop. Result: the berries still cause their negative health effects when consumed, but the truth of their toxicity doesn’t “out” – we don’t learn it, and don’t modify our behavior accordingly. So it is with many of the environmental effects that impact us. Their causes go un- or misattributed by cognition.

The degree to which the truth doesn’t out is the latitude that we have to construct social realities that are robust to correction by objective forces. As far as we’re concerned, the mildly poisonous berries are a valuable part of our diet. We might even be able to manipulate others’ mental models of the world’s causal relationships and convince them that not only do the berries lack negative effects, they cause a positive one: enhanced sexual function. Coincidentally, we may have some available for purchase.

I’ve been speaking of cognition modeling the world’s cause-and-effect relationships more or less literally, as if we carry a list of rules and rationales in a mental look-up table. E.g., “Don’t touch hot things because they hurt.” Some of our behavior selection might work something like that, but it’s important to note that what we typically think of as comprehension or understanding is not required to learn a behavior that causally contributes to a desired outcome, only that we produce an efficacious response when presented with a particular environmental situation. As far as objective causation is concerned, doing the right thing for the wrong reasons is just as good as doing it for the right ones, because it’s the behavior, and the behavior alone, that does the causing (or the contributing to a cause). How we chose it is irrelevant. All a neural network (and that may be all our brains are) “knows” is that Input Array X maps to Output Array Y – and that’s enough.

The behaviorists were right after all that behavior and reinforcement lie at the core of psychology. They just left the interpreter (cognition) out of the feedback loop that coordinates responses with stimuli.


Leave a Comment

Previous post:

Next post: