(p. 19) Changing Behavior
I have identified reinforcement learning failures as central to situations in which behavior change problems are recognized (e.g., in public health, marketing etc.). Eliminating learning blockages is thus a crucial part of the behavior change task. The Behavior Centered Design (BCD) Behavior Challenge model (Figure 2.1) identifies a behavior change task for each step in the reinforcement learning process. These tasks are grouped, for BCD purposes, into three primary jobs that any behavior change intervention must achieve—Surprise, Revaluation, and Performance:
• Surprise is about the process of perceiving a stimulus via sensations and perceptions, with additional psychological responses, expressed as the problems of getting exposure, grabbing attention, and facilitating processing.
• Revaluation is about the parallel process of analyzing the stimulus for its value, via resources being acquired and valued in terms of rewards, which are the tasks of modifying value and altering rewards in behavior change terms.
• Performance is about the forward-chained process of establishing the best behavioral response to the stimulus via action selection and behavioral performance itself—by getting selected and generating opportunities for performance.
To reiterate, then, the challenge of behavior change consists of three basic problems (with two tasks for each):
• Create Surprise.
• Get exposure.
• Grab attention.
• Cause Revaluation.
• Alter rewards.
• Modify value.
• Enable Performance. (p. 20)
• Disrupt setting.
• Action selection.
Solving these problems is a central focus of BCD. Each is discussed in some detail in this chapter.
The first challenge in behavior change is to generate surprise. Most everyday behavior is settled into fixed patterns of response. To shake behavior out of these ruts, something new has to happen. So, if we are to persuade behavior into new directions, we first have to provide surprising stimuli. Let’s discuss how surprise can be created.
If the immediate environment doesn’t contain new stimuli, then no learning processes can take place and individuals will not respond to the intervention with changed behavior. Hence, the first job of any behavior change effort is to get new stimuli in front of people. Marketers tackle this problem by analyzing how and where their target population spends their time. They identify touchpoints—the places and times that they can exploit to connect people with products or to channels of communication . For example, mothers with new babies may be targeted by pharmacies, while people looking at property websites may be (p. 21) targeted with furniture ads. A premium is charged for placing products besides tills in supermarkets or to use billboards on the side of heavily used transport routes. A touchpoint analysis in a rural low-income setting may reveal that clinics, despite being an obvious choice for health messaging, may be a poor channel of communication, as mothers may, on average, make very few visits each year to a clinic and then may only interact with a health worker for a few short minutes. Behavior change campaigns that do not get exposed to the target audience fail at the first hurdle.
A second problem is that although people may be exposed to new stimuli, the stimulus may not be attended to, especially in message-inundated modern environments. Ensuring that the target stimulus gets processed is a design problem.
Luckily, we know what makes a stimulus stand out. We pay attention when we encounter a stimulus that confounds our predictive brain, when it seems that the world model in our heads may be wrong . This means that we could be behaving inappropriately and possibly dangerously and that to fix the problem we need learn. So, a stimulus that we attend to is one that confounds what we already know and expect. For example, no one pays attention to small stones by a path, but when the same stone is moving in a trajectory likely to collide with our bodies, we respond immediately. No one pays much attention to a crowd of people queuing at a ticket barrier, but when one individual behaves unexpectedly, by doing a handstand, for example, we pay attention. Phenomena that defy evolved expectations are surprising. For example, phenomena that defy our mental models of how physical objects or biological creatures act, such as physical objects moving by themselves or nonhuman creatures talking. Think of seeing an ad containing a picture of a tree with eyes: trees are plants, but only animals have eyes, so a tree with eyes confounds expectations and, therefore, is surprising. In terms of the predictive brain hypothesis, this is called prediction error: we predict how we expect the world to be and pay attention when we find the prediction to be wrong, as it may mean that we have to fix something .
Producing a surprising stimulus is therefore what interventions should do, as such stimuli will grab attention. Creative people are good at coming up with surprising ideas, because they can think out of the box, which is why engaging creative people to help design behavior change programs is a good idea.
Of course, a program’s aims won’t be achieved just by getting people to attend to a randomly surprising stimulus. Surprise must have consequences that result in desired changes to target behaviors.
(p. 22) So what sort of surprise will help us to achieve behavior change? We need to create the sort of new stimuli that will lead brains to think that the individual can get what they need by performing the target behavior. Brains are constantly scanning the environment for opportunities to employ behavior to get what the brain or body needs (or avoid what might harm it).
What are those things that we need? I will again turn to evolutionary theory to answer this question, because what we need, fundamentally, is to be able to survive and reproduce. So we have evolved to perceive value in doing things that present us with evolutionary benefits—that is, to engage in behaviors that produce resources that translate into improved chances of survival and reproduction. In essence, we are motivated (by expected value calculations) to pursue goals that will put us into these “good” situations. What is the nature of these good situations? The two basic evolutionary tasks can be further broken down into particular strategies, given the way in which humans live. For example, since we reproduce sexually, we must attract and bond with mates. Further, these bonds produce dependent offspring who must be nurtured. A full discussion of the different evolved human motives awaits in Chapter 3; for now, simply remember that there is a distinct set of motives that can be exploited in various ways by behavior change interventions.
So how can human motives be used to change behavior? One way is by emphasizing the rewards from performing the behavior or by highlighting the possible negative consequences. In both cases, the sensory salience, temporal proximity, or statistical likelihood of these rewards being experienced may not be obvious. For example, the spread of pathogens is imperceptible but nevertheless causes disease—a real threat that simply isn’t easy to understand or identify using the human senses. By artificially making this process visible through exaggeration (e.g., the supersized bugs with evil personalities that appear in toothpaste ads inhabiting the mouth), the benefits of the behavioral practice can be made more salient.
Similarly, people may have the perception that rewards will only appear in the relatively distant future and, therefore, discount them in a way that is maladaptive. For example, the benefits of education may only be associated with getting a better job in the long-term future. But education can be revalued by emphasizing the good feelings associated with being more knowledgeable and capable in everyday matters now or by focusing on common expectations (i.e., simply doing what everyone expects of a young person).
Finally, the likelihood of experiencing desired rewards can be increased by offering guarantees of various kinds. For example, marketers often offer “money back if not satisfied” deals on new products to make reward statistically more likely.1
(p. 23) In summary, occurrence of a given behavior can be made more likely if the rewards that naturally follow from performing it can be altered by making the reward appear more or less obvious, more or less immediate, or more or less likely. However, this tactic may not actually change the amount of reward that is experienced from carrying out the behavior—only some perceived quality of it (e.g., how fast it is experienced or its sensorial richness). Hence, any revaluation may only be temporary. Actions may change once and then revert to the old pattern. However, we can do more than this; we can also change the value of the behavior itself.
A key task for the would-be behavior changer is to modify the value of the target behavior. One way is via the classic marketing strategy of adding value to a product by adding a new association, for example, by suggesting the behavior is what high-status individuals (e.g., celebrities) do or by linking the behavior with a moral cause, such as suggesting that a brand is environmentally friendly or socially conscious.
All behaviors have a proper domain motive (i.e., that for which they evolved: eating assuages the Hunger motive, while the Hoard motive causes us to gather resources for a rainy day). However, other motivations can also be recruited to help instigate behavior change. For example, you can make eating an Affiliative activity or one that displays your Status. Indeed, through branding, values of all kinds can be added to a food: it can be portrayed as what good mothers feed their children (the Nurture motive) or as grown by local farmers (Affiliation motive). Obviously, the opposite can also be important: antismoking campaigns try to make smoking less valuable by suggesting that the practice isn’t really popular among peers (i.e., not a norm in the Affiliation group) or only practiced by those of low Status, thereby reducing the number of motives associated with the behavior. By varying the number of motives and kinds of rewards attached to the performance of the target behavior, the new behavior is revalued and, hence, will become more/less likely.
Despite getting attention, being processed, and engaging in Revaluation, the intervention must still get the target behavior to be enacted. The purpose of Performance is to interact with—and change—the environment in a way that minimizes Surprise as previously perceived—that is, to remove the perceptual error through behavior and thus make the look of the real world conform as closely as possible to the person’s internal world model . Performance of the target behavior should be the means by which this reconciliation between the (p. 24) individual’s world model and the world itself is achieved. This problem can be split in two:
• Setting disruption.
• Action selection.
Performance of the target behavior has to be perceived by members of the target population as a way to achieve the objective of minimizing the Surprise instigated by the intervention. This means that the context within which the target behavior occurs has to be modified in a way that causes the desired behavior to be part of the setting’s mission. The basic rule is, Disrupt the context in some way so that the target behavior is performed. There are multiple ways in which this can be achieved.
While the Behavior Challenge model is precise about the psychological steps necessarily associated with behavior change, it is not specific about the nature of environmental influences (environment is left as a single variable). However, a very useful model of the environment is available in the form of the behavior setting concept . A behavior setting is derived from the work of Roger Barker in ecological psychology . Mealtimes, business meetings, air travel, educational classes, religious services, and waiting at a bus stop all constitute types of behavior settings. Such settings can be thought of by analogy to a stage play where actors congregate in a particular place to perform oft-repeated roles with certain props and well-rehearsed scripts. Each setting thus has a purpose, a designated place, a set of objects, and a prescribed set of behaviors. Each person entering into a setting expects the others who simultaneously participate in it to perform their (implicitly) designated roles. Deviation from these roles is generally punished by those participating in the setting.
To understand the power of settings, imagine that you are speaking in front of an important audience of your peers and you forget what you meant to say. In effect, you have failed to play your role properly, having been let down by your memory, due to the stressful situation. Those in the audience will express displeasure and annoyance, and you will attempt to recover as quickly as possible—your embarrassment being an internal censuring system to get you to perform properly. (For more examples of setting regulation, see the many videos of social disruption experiments on the Internet, which show what happens when people violate simple, everyday norms such as talking to or sitting too close to strangers in public place or not using silverware during a meal.) Settings are a powerful means of understanding what directs people’s behavior. They suggest that we need to look to the supportive social conventions, physical objects, and infrastructure that regulate ordinary behavior. Behavior settings are the situations within which people have learned what to expect from the environment and from other people’s behavior.
(p. 25) To continue with the theater analogy, a behavior setting can be said to have the following elements:
• Stage: the place and things surrounding where the setting regularly occurs.
• Props: the objects manipulated to help accomplish behaviors (often called synomorphic because they fit the behavior; e.g., a hammer is designed both to fit the hand and to bang in nails efficiently).
• Infrastructure: relatively large physical structures necessary for performance but that are neither manipulated nor damaged through use.
• Roles: the interacting strategies used by the cast of actors which meet their needs separately and together.
• Motivation: the motive(s) driving role-playing behavior in a setting (i.e., the goal or benefit an individual hopes to gain from playing a role).
• Routine: a learned sequence of behaviors performed regularly, and typically in the same order, to fulfill a role.
• Script: an individual’s knowledge of a routine—that is, a set of mental instructions about how to behave (which may be implicit) in a particular behavior setting to play a role.
• Norms: the implicit rules governing role-play in a setting.
Changing settings is therefore a powerful and sustainable way of changing behavior . Indeed, behavior change can be said to be essentially about disrupting behavior settings so that appropriate learning can occur. In effect, I argue that we must reset people’s situations, so that they can naturally learn the appropriate thing to do.
Barker used cybernetic cycles to represent control mechanisms, operating not at the level of individual psychology, but at the level of the behavior setting itself . His eco-behavioral operating circuits counter deviancies introduced by human or physical participants in the setting to maintain its progress through the setting’s dynamic agenda. These are mechanisms for controlling the performance of settings, keeping people acting within their routines in conformity with their adopted roles, using objects in ways consistent with their synomorphic nature, and correcting deviations from the normal pursuit of the setting’s objective. He argued that these control mechanisms reside not just in psychological processes, but also in the causal linkages between the physical, social and biological components of settings, making a setting a superordinate, dynamic, self-governing entity . Thus, it can be argued that a setting is, essentially, a positive feedback system—like reinforcement learning itself—operating at a higher level of organization.2
One kind of setting disruption is role change—the strategic position played by some individual in a setting. If individuals pledge in public to adopt a new role, social control makes it harder for them to defect.
Attaching particular behaviors to the enacting of a role (e.g., via pledging) is one way to create new opportunities for the behavior to be performed. Pledging is making a commitment—preferably in public, to add weight to the promise—to (p. 26) perform the target behavior. It is also well-known that making the pledge specific as to the situation in which it should be performed—what psychologists call an implementation intention—makes it more likely to be performed [83, 84]. An implementation intention thus makes reference to particular behavior settings. Having people sign their pledge as well turns it into an implicit social contract, lending it further weight because failure to adhere to the contract can be punished by others.
Ensuring that the target behavior is performed may also require modification of the script for the setting. For example, people may currently think that getting their hands clean before eating simply requires rinsing them with water to remove any visible dirt. An intervention may suggest that this is insufficient because there are also dangerous but invisible agents that need removing from hands. To be effective, this intervention must ensure that individuals believe that the handwashing setting can only be successfully completed by including a new step in the routine: washing with soap as well as water. This means individuals must modify their script for this setting to include this extra step. Even if the step is omitted next time through habit, memory of the script may cause individuals to go back and insert the soap use before leaving the setting.
Another way of facilitating performance is to have people modify their environment by increasing the level of technological support for the target behavior. For example, installing an irrigation system makes watering the garden regularly much easier. Getting people to take their medicine is facilitated by buying a container with the time and day labeled on separate cells in the box, making it easy to determine whether this afternoon’s dose has been taken or not. Simply putting the medicine out on the counter, rather than behind a cabinet door, can cause the medicine bottle itself to serve as a reminder. Such simple tricks can often suffice to increase the likelihood of performance when the setting arises. Similarly, the Target Setting can be modified such that some barrier to performance of an undesirable behavior is introduced, as when people stop having snacks in the house.
The target setting can thus be modified in many ways to safeguard that the target behavior is performed.
The Setting Transfer Problem
There is a further problem, however: target individuals are typically exposed to an intervention in one setting (e.g., a community event) but perform the target behavior in another (e.g., sexual interactions in the case of an HIV program). For behavior change to occur, the individual must retain some mental novelty (e.g., an intention to perform the target behavior or a new associational value) during the period between exposure to the surprising stimulus and the time when opportunities to perform the target behavior occur. I call this the Setting Transfer Problem. It is another reason that behavior changers must ensure that the psychological effects of their intervention are long-lasting.
This problem is related to the issue in psychology called prospective memory, in which an individual must recall an intention to do something at a specific time and place in the future . The prospective memory literature shows that the (p. 27) Setting Transfer Problem can be overcome to some degree by making the intention to perform a behavior specific to a particular situation—that is, by couching the intention in the form of an implementation intention  or by placing reminders in the target setting that will help cue performance . (Use of the word performance also purposely implies acting out or playing a role on a stage, which emphasizes that it occurs in a particular setting.)
The Setting Transfer problem can also be solved simply by creating some new mental association that is remembered long-term—that is, new information that comes to form part of long-term memory but that isn’t a conscious intention. Just finding the target behavior more attractive because it is now linked (relatively permanently) to new, desirable goals (that satisfy other needs than the primary one) can work to cause the target behavior to be more highly valued.
This kind of processing—forming an intention or long-term memory—is more likely to happen if, at the time of exposure to the intervention, individuals
• have relatively few other stimuli to process (i.e., low cognitive load and low stress).
• have relatively more time before the next behavioral response must be produced.
• are in a social setting so that the experience can be shared.
• are in a teachable moment or life-change event (e.g., birth of child, moving), when they are seeking to learn how to behave differently.
These are all situational variables that can be modified by choice of touchpoint. For example, a doctor’s waiting room may be a good touchpoint because people in the waiting room may be in a teachable moment and with time between activities. On the other hand, they could be stressed and, hence, unreceptive to new stimuli.
There are also ways in which the stimulus itself can be presented that make processing easier. These include employing
• multiple sensory channels to facilitate immersion (e.g., TV with sound and vision vs. radio, which is only sound).
• simplification and exaggeration of stimuli (as in animations) .
• highly recognizable settings to encourage rapid perceptual identification.
• narrative (i.e., using our evolved ability to process the simulated experience of others).
Even if all the previous steps have been taken, the target behavior may still not be enacted when an appropriate situation arises. This is because before a behavior is performed, it must be selected from among many other potential behaviors [62, 89]. That process can be delegated to the environment, as when a cue triggers an automatic habitual response. Alternatively, a motivated (although implicit) (p. 28) calculation may suggest that in the present state of hunger, consuming food has more value than hoarding it, for example. However, there are executive control factors that can also affect whether a behavior gets performed:
• Cost: the more expensive the action in terms of energy, time, and mental resources, the less likely it is to be performed.
• Confirmation bias: the tendency to value choices consistent with existing beliefs and values.
• Availability in memory of relevant information (e.g., one can more easily remember the name of a celebrity than a friend due to the former being heard in so many different contexts).
• Endowment effects: the tendency to devalue choices that require one to give up something that one has already acquired (e.g., to pay tax on income).
• Locus of control bias: the tendency to believe that exogenous environmental factors can be influenced through action.
• Mere exposure effect: the tendency to value things simply because they are familiar, not due to intrinsic qualities of the thing.
• Fundamental attribution error: the tendency to believe that others willfully choose to engage in behaviors rather than recognizing the power of the situation on them.
These biases and heuristics (or rules of thumb) for decision making have been investigated by behavioral economists, who see these as “predictably irrational” forces behind behavior . Yet, from an evolutionary standpoint, all of these biases are rational, in the sense that they help people to make adaptive choices. For example, it makes sense to ascribe greater value to something one has possession of now than to a promise of the same thing later, since the future is always unpredictable and promises are not always kept. Similarly, the locus of control bias is often adaptive because it is better to assume that one is a forceful agent in the world than powerless, in which case no attempt at goal achievement would be made. For example, children with an internal bias (i.e., belief that events are due to their own, planned actions) function more positively and efficaciously in achievement situations than those with an external locus of control (i.e., those who believe events are primarily caused by factor external to themselves) . The fundamental attribution error is simply the locus of control bias applied to other people, rather than oneself, and has a similar explanation: assuming others are responsible for their actions is a natural consequence of trying to predict others’ behavior by imagining their mental states . Further, overestimating the probability of others having aggressive intentions (and preparing a defensive response as a consequence) is a better error to make than assuming they are benign and risk exploitation or attack [92, 93]. The error can be explained as an appropriate bias when it reduces the costs of potentially dangerous inferences . Thus, like behavioral economists, I believe these cognitive biases are operational in human decision-making, but unlike the economists, I argue that they are (p. 29) adaptive mechanisms. People should exhibit behavior that favors getting a reward now rather than later (i.e., temporal discounting) and make efforts to achieve goals in the face of obstacles (i.e., an internal locus of control bias) to be effective agents in the world.
The final valuation of a behavior is thus a function of the intrinsic value of the benefits received, as modified by the relevant action selection factors. The name of the game, then, is to make sure conditions are such that the target behavior wins in comparison to all other options after these factors are taken into account.
In Theory of Change terms, Performance can be considered the outcome of the intervention. As a consequence of having gone through this process of Surprise, Revaluation, and Performance, the desired outcomes should occur. These processes can be summarized in a central principle of BCD: Disrupt settings with Surprise to force Revaluation and thus cause Performance. This principle requires that the causal mechanisms linking the program implementation to the Target Setting in Theory of Change all work through as expected. In other words, the program design problem is to create an intervention that works like a guided missile, flying through the environment, ripping into bodies, and burying itself in brains, where it detonates, releasing repercussions that disseminate through the body and, via behavior into the environment, completing the learning cycle. As a result of the behavioral changes, the world should look more like the brain expects it to. If not, another round of behavior might be required to ensure that the world is a safe, well-known place, without any surprises in store.
This concludes my discussion of the basic theory behind the BCD approach to behavior change. However, the modeling thus far remains somewhat abstract in terms of the actual factors that might influence behaviors in the moment of performance. We still need to elucidate in greater detail what these factors might be, because it will be important when we actually come to the practice of identifying what aspect of the setting we need to change. Hence, we turn to the problem of behavior determination in the next chapter. (p. 30)
1. For public health practitioners, it is important to recognize that hyping up health benefits is unlikely to make people experience (or imagine) any immediate, tangible reward for behavior. Indeed, health is not one of the 15 human motives; we did not evolve a reward system for healthy behavior, as such. Hence, messaging about health consequences is unlikely to be an effective strategy for public health behavior change efforts .
2. It is important to note that this level of organization occurs between the individual and the usual social scientific concept of an organization, such as a business or school. In particular, multiple settings can occur within the operation of a social organization (e.g., “Mrs. Smith’s music class” within a school or the “weekly staff meeting” within a government bureau). This makes Barker’s notion of a setting different from the setting concept used in health promotion, or the World Health Organization’s “healthy setting” concept, which equate a setting with a hospital, village, or other social organization [81, 82]. When necessary, the phrase behavior setting will be utilized to ensure the ecological psychological notion is distinguished from this other use.