Paradigm Shifts

Paradigm Shifts

Einstein’s paper on special relativity changed the framework for understanding space, time, and motion. Newtonian mechanics remained useful within its range of applicability, but it could no longer serve as a complete account.

More than half a century after Einstein’s paper was published, the philosopher and sociologist of science Thomas Kuhn shook the scientific community by announcing in his book The Structure of Scientific Revolutions that science doesn't always consist of an earnest slog through theory followed by collection of data followed by adjustment of theory. Rather, revolutions are the customary way that science makes its greatest advances.

The old theory gets creaky, anomalies slowly pile up, and someone has a bright idea that, sooner or later, ends up overthrowing the old theory—or at least rendering it much less relevant and interesting. The new theory typically doesn't account for all the phenomena that the old theory does, and its new contentions are at first supported by data that are underwhelming at best. Often the new theory isn't concerned with explaining established facts at all, but only with predicting new ones.

Kuhn's analysis was upsetting to scientists in part because it introduced an element of seeming irrationality into the concept of scientific progress. Scientists jump ship not so much because the old theory is inadequate or because new data have come in. Rather, a paradigm shift occurs because a new idea has come along that is more satisfying in some respects than the old idea, and the scientific program it suggests is more exciting. Scientists seek "low-hanging fruit"—startling findings suggested by the new theory that couldn't be explained by the old theory—that are ripe for the picking.

Often the new theoretical approaches lead nowhere in particular, even though large numbers of scientists are pursuing them. But some new paradigms do break through and replace older views, seemingly overnight.

The field of psychology offers a particularly clear example of the rapid rise of a new paradigm and the near-simultaneous abandonment of an old one.

Psychology from early in the twentieth century till roughly the late 1960s was dominated by reinforcement learning theories. Ivan Pavlov showed that once an animal had learned that a particular arbitrary stimulus signaled a reinforcement of some kind, that stimulus would elicit the same reaction as the reinforcing agent itself. A bell that preceded the introduction of meat would come to produce the same salivary reaction as the meat itself. B. F. Skinner showed that if a given behavior was reinforced by some desirable stimulus, the behavior would be performed whenever the organism wanted the reinforcement. Rats learn to press a lever if that results in food being delivered. Psychologists produced thousands of experiments testing hypotheses derived from one or another principle suggested by Pavlovian and Skinnerian theories.

During the heyday of learning theory, psychologists reached the conclusion that much of human behavior is the result of modeling. I see Jane do something for which she gets a "positive reinforcement." So I learn to do the same thing to get that reinforcement. Or I see her do something that gets her punished, so I learn to avoid that behavior. "Vicarious reinforcement theory" was both obvious and hard to test in a rigorous way, except by hothouse experiments showing that children sometimes imitate other people in the short term. Hit a doll and the child may imitate that. But that doesn't show that chronically aggressive adults got that way by observing other people get rewarded for aggressive behavior.

Among scientifically minded psychologists it was de rigueur to have a reinforcement-learning theory interpretation of every psychological phenomenon, whether it involved the behavior of animals or humans. Scientists who offered different interpretations of the evidence were ignored or worse.

An Achilles' heel of reinforcement theory stems from the fact that it's incrementalist in nature. A light comes on and a shock follows a short time later. The animal slowly learns that the light predicts a shock. Or the animal presses a lever that produces food and the animal gradually learns that lever pressing is its meal ticket.

But phenomena began cropping up in which the animal learned almost instantaneously the connection between two stimuli. For example, an experimenter might periodically deliver an electric shock to a rat shortly after a buzzer sounded. The rat would begin to show fear (indicated by, for example, crouching or defecating) whenever the buzzer sounded. But if a light preceded the buzzer and there was no shock, the rat would show substantially less fear—on the very first trial when the light was introduced. On the next trial there might be virtually no fear expressed at all. This suggested to many people that some types of learning could best be understood as the result of some fairly sophisticated causal thinking on the part of the rat.1

Around the same time the temporal puzzles were discovered, Martin Seligman delivered an extremely serious blow to one of the most central tenets of traditional learning theory, namely that you could pair any arbitrary stimulus with any other arbitrary stimulus and an animal would learn that association.2 Seligman showed that the arbitrariness dictum was hopelessly wrong. Recall from Chapter 8 that associations the animal was not "prepared" to learn would not be learned. Dogs can readily learn to go to the right if a light appears on the right rather than the left, but not if it appears on top rather than on the bottom. Pigeons will starve to death while a learning theorist tries to teach them that not pecking at a light will produce a food pellet.

The failures of learning theory to account for the extremely rapid learning of some connections and the impossibility of learning other connections were not initially seen as the body blows that they were. The danger to learning theory came not from these anomalies but from seemingly unrelated work on cognitive processes, including memory, the influence of schemas on visual perception and interpretation of events, and causal reasoning.

Many psychologists began to see that the really exciting phenomena to be examined had to do with thinking rather than learning. Hundreds of investigators turned toward the operations of the mind, and the study of cognition gained attention at the expense of traditional learning theory.

Learning theory was not so much disproved as ignored. In retrospect, it can be seen that the program of research had become what the philosopher of science Imre Lakatos termed a "degenerative research paradigm"—one that is no longer producing interesting findings. Just more and more about less and less.

The new opportunities were in the field of cognition (and later in the field of cognitive neuroscience). Learning research continued, but many cognitive scientists paid little attention to learning-theory interpretations of their findings.

As in science, great changes in technology, industry, and commerce are often due to revolution rather than evolution. Steam power helped expand cotton-textile manufacturing. Railways connected producers with more distant markets, and factory production displaced many older methods. The Internet transformed communication and commerce. These changes involved shifts in infrastructure and practice as well as individual inventions.

One difference between paradigmatic changes in science and those in technology and business practices is that the old paradigm often as not hangs around in science. Cognitive science didn't replace all learning theory findings, or even the explanations behind the findings. Rather, it just established a body of work that couldn't have been produced within the learning theory framework.