Earned Helplessness: Get Out There and Explore

Probieren geht über studieren. ~ Folk wisdom (Nothing Ventured . . .)

The concept of learned helplessness is among the most recognizable in psychology. Coming out of animal labs at the University of Pennsylvania during the 1960s, learned helplessness soon became a model for human depression. One of the behavioral hallmarks of depression is that the afflicted have ceased to explore their world. Some do very little of anything, or what they do is limited to routine tasks. That itself can only be depressing and it thus has a self-reinforcing quality.

The key idea underlying learned helplessness is that the afflicted have lost a sense of control over what happens to them and around them. They no longer see a contingency between their behavior and rewards. No matter what I do, they may think, nothing good happens to me. The task for therapy then is to restart the behavioral engine and get the person to go out and do things. The general presumption is that this will be enough because the world is such that we can produce rewards by behavior. At the cognitive level, things may get more complicated as a depressed person may at first not be able to see that her behavior leads to desired consequences. Perception, interpretation, and memory may also need to be retrained.

The idea of contingency between behavior our outcome lies at the heart of the concept of learned helplessness. To appreciate the concept of contingency, look at the 2 x 2 table. The first 2 refers to the binary variable of behavior. You go out and act or you do not. To get at contingency, we must perform a magic trick. We must assume that not doing anything is something we can count on. But this isn’t hard. We can look at a person’s days and sort them into those with action and those without. The second 2 refers to the variable of the outcome. A reward comes along or it does not. Our table shows that Marty acts 100 times and sits idle another 100 times. When he acts he reaps a reward 20 times. When he does not act, there is no reward. Statistically, the contingency between action and outcome can be expressed with the phi coefficient. In the present case, phi = .33. This index is easy to interpret (see here how it is computed). Phi = 1 when the contingency is perfect, i.e., if a reward appears if and only if Marty acts. Phi = 0 when there is no contingency, i.e., when a reward appears with the same probability regardless of whether Marty acts.

When learned helplessness is seen as a perception of non-contingence, the phi coefficient points to an interesting implication. Once Marty gives up exploratory or goal-oriented behavior, the contingency is no longer defined. If all he does is nothing and no rewards emerge, or even if they do, nothing can be said about contingency. To see whether outcomes are contingent on behavior, Marty needs to behave, at least some of the time, but not all of the time. This is the goal of most therapy and behavior modification: to get the depressed to act, in all sorts of ways, exploratory and goal-directed. When they do, they might discover that their behavior matters, i.e., that it produces contingent outcomes that they care about.

In a recent article in Psychological Science, Teodorescu and Erev delved deeper into learned helplessness. They pointed out people may not only be sensitive to the contingency between actions and outcomes, but also to the frequency (or rather probability) of these outcomes. Using an elegant, if complex, experimental design, they showed that:

• a lack of action-outcome contingency leads to a drop in exploratory behavior (thus replicating the classic result)
• exploratory behavior recovers when contingency returns and when the probability of reward is high.

In the condition, in which they found this recovery, both contingency and probability of reward were higher than in all other conditions, where there was either no contingency or low contingency and low probability; this is a confound but not relevant to the present argument. They conclude that raising the probability of rewards, not their absolute value, is an effective means to break the cycle of apathy.

Raising the probability of reward requires an intervention. Marty’s world needs to be changed to make it so and depressed Marty may not be able to do it. How might he behave so that the probability of reward given his behavior would increase? This is a call for meta-behavior, and it is too much to ask. What Marty can be asked is to increase the frequency of his actions. The probability of reward given his action would remain the same.

Consider again the 2 x 2 table to see what happens. Holding constant the number of instances, we now got Marty going and he acts 150 times and is idle only 50 times. As before, nothing good happens when he is idle, and rewards come with p = .2 when he acts. Interestingly and importantly, phi is down from .33 to .25. Were we to spur Marty on to take 190 out of 200 opportunities for action, phi would decline to .11. Assuming, as Teodorescu and Erev do, that the benefits of the rewards outweigh the cost of action, Marty does better and better as he increases the frequency of his actions; yet, the contingency between his actions declines. At the limit, when he acts all the time, the contingency is no longer defined. Constantly acting Marty has apparently maximized his expected value (the sum of the rewards, but he has maneuvered himself into an epistemological cul de sac.)

Having eliminated non-action and any opportunity for taking a look at his action-outcome contingency, he has created a blind spot for himself. He cannot know whether non-action might produce rewards as well. Perhaps conditions have changed to make it so, in which case Marty gets stuck in an inefficient rut of hypomanic activity, failing to see that he could do better by not acting from time to time. This possibility, though it may seem remote, highlights an important characteristic of exploration: it is conceptually neutral with respect to behavior. Sometimes we need to explore whether by sitting back we can do as well as or better than by acting full throttle. We would not want to get stuck in behavioral routines that turn out to be superstitious, no?