What matters most when choosing between two rewards, one imminent and the other distant? The critical factor is one’s evaluation, at the moment of choice , of their respective values: This is what is going to give the motivational force for obtaining one or the other. As the saying goes, a bird in the hand is worth two in the bush. It is because we have performed this calculation of probability that we may prefer a small reward that is near (all of a sudden it seems very important) rather than a larger, distant award (it seems to be of a lesser value). Let us refer to this evaluative factor as time preference.

If one finds herself removed from two rewards, the more distant reward might seem more important than the closer one, because the latter is still far away. But because of the hyperbolic , rather than exponential , nature of discounting , there is a moment when, as the reward nears, its motivational force abruptly rises and may become superior to that of the distant reward. And so the agent chooses it, even if beforehand its value appeared inferior to the one that the distant reward would have in the end. Figure 7.3 gives a graphic representation of this phenomenon.

Let us imagine an individual who is advancing toward two rewards: One is near while the other is more important, but further away, and she must choose: It is one or the other. She will schematically pass through three stages. In stage A, she evaluates the more distant reward as being more important than the more immediate reward: She prefers the distant reward. But as she approaches the more immediate award, she enters stage B: Because of the hyperbolic nature of the discounting curves (this would not be the case if the discounting function were exponential ), the discounting curves cross and as the value of the nearer reward acutely rises, and the nearer reward now appears to be more important than the distant reward. There is a real reversal of preferences; the individual now prefers the immediate reward: And so she chooses it rationally, even though it has an objectively inferior value compared to the distant reward—and in spite of her having once preferred the more distant reward.

The following metaphor, taken from Elster , illustrates the phenomenon of preference reversal: Imagine standing in front of a small building that has a taller building behind it. As you approach the first building, it will at some point obscure the taller building behind it, as if it were bigger. So it is with our preferences: Rewards which are nearer loom larger in our preference calculation.

Elster also helps us to understand preference reversal due simply to the passage of time (Elster 1999b, 430). When I enter a restaurant (period A) my desire to stay on a diet may be stronger than my desire to eat a dessert. Unfortunately, as the dinner progresses, my good intentions weaken, perhaps also due to the consumption of alcohol , and when the waiter arrives with the dessert cart, my desire for cake suddenly grows and becomes superior to my desire to stay thin (period B). We must stress, as Elster does, that here it is not the sight of the dessert that propels the force of desire, but simply the passage of time: One can say that the server and the cart arrive at the right time. We might note here in addition that it is not by chance that as you arrive at the restaurant, you are offered a cocktail before you can even look at the menu. Alcohol increases the discounting of values over time and the resulting ‘myopia ’ makes you see everything on the same plane (Steele and Josephs 1990).

For Damasio , in Descartes ’ Error,

this concept that has been proposed to explain the behavior of individuals under the influence of alcohol and other drugs . Inebriation does narrow the panorama of our future , so much so that almost nothing but the present is processed with clarity (Damasio 1994, 218).

Then, when you take the menu, the curves of your conflicting desires —one to eat cake and one to stay thin—will have already crossed and you will order the dessert without hesitating, postponing your diet and your good intentions until tomorrow.

Robert Nozick pertinently asks why we have a tendency to call stage B the temptation stage, and why we think it must be resisted (Nozick 1993, 16). Why do we consider that it is during stages A and C, when the more distant reward appears to be the more important one, that the agent should make her choice , even though, during stage B, it appears to be rational for the agent, taking into account her preferences at that moment, to choose the more immediate reward? Nozick offers the following explanation: Stage B is not an appropriate period for a decision because it is not representative of what the agent generally believes: It is very short, in fact a lot shorter than the sum of stages A and C. Let us suppose that the agent has consumed the immediate reward that keeps her from receiving in the future the more distant reward. Afterwards (at the end of stage B), she will evaluate the reward she has now given up, and she will again assign it a greater value because she is now at stage C (she again can see the taller building!), and she will regret her decision. But if she has resisted temptation at stage B, then when she arrives at stage C and looks back, she will still believe herself to have made the right decision.

An ex-smoker wrote me that he keeps a pack of his favorite cigarettes in his desk drawer and from time to time looks at it “like the toreador at the bull”. This extraordinary flirtation with risk is only possible, in the long run, if the former smoker is already in period C, where the contemplation of the unsmoked cigarettes confirms the validity of his choice .

So the reason we believe the agent should resist temptation during stage B is that it is relatively short, and the choices made in stage B are marked by regret , while those in A and C are not. Besides, during period B, the agent is at periods A or C for other similar choices, which may also influence her.

And sometimes one does resist temptation . Indeed, we saw that doing so requires a particularly human capability. The rewards expected at different times can be chosen for their importance, but on condition of making the choice sufficiently in advance, i.e., at the moment t1 in Fig. 7.3 where, because of the hyperbolic nature of discounting curves, the attraction of the distant choice appears to be even more important than that of the more immediate reward . It is enough for the agent to realize that there is a risk of the two curves crossing: After this moment (when it reaches t2), the attraction of the more immediate reward will be stronger and it will be too late. Before the curves cross, the agent will use her self-control , in the form of a ruse , to avoid giving into temptation: A system of warnings, or what Ainslie (1985, 144–145) and Elster (2000b) call a precommitment device or strategy. For instance, Ulysses faces two choices: The distant reward, returning to Penelope ; the immediate reward, a dalliance with the Sirens . Before the desire for the Sirens becomes too strong, he appeals to his self-control and uses the well known ruse: He commands his sailors to tie him to the mast. And knowing that his will is not all-powerful, he then instructs them:

If I beg you to release me, you must tighten and add to my bonds. (Homer 2003, Odyssey, Book 12, quoted in Elster 1977)

Implemented at stage A, Ulysses ’ ruse keeps him from giving into temptation during stage B.

If medical nonadherence depends on similar mechanisms , these considerations should be useful for developing ways to improve adherence.

According to Elster , one cannot voluntarily change one’s rate of discounting for future rewards if it’s elevated. Rather, wanting to be motivated by long-term consequences entails that one already is motivated by such rewards. Compare morality : If one wants to be immoral, then one is immoral (Elster 2000b, 28). Let us call the failure to be motivated by far-removed consequences impatience; then we may say that one either is or is not impatient . Elster and Skog note that desires , which are the driving force behind our actions,

cannot be classified as rational or irrational […] This argument applies not only to substantive preferences for one good over another […] Some like chocolate ice cream, whereas others have a taste for vanilla. […] Similarly, it is just a brute fact that some like the present, whereas others have a taste for the future . If a person discounts the future very heavily, consuming an addictive substance may, for him, be a form of rational behavior . (Elster 1999a, 17)

Inasmuch as this type of impatience leads to the possibility of ‘caving in’ associated with a reversal of preferences similar to the one described above, the first way to avoid it is to develop the techniques of precommitment described by Elster (2000a, 188–190). For example, one might not keep cigarettes —or sweets—in the house to render impulsive use impossible. Such a strategy may also force some delay between the urge and gratification; while driving to the store to buy some cigarettes, one has a moment to re-think the decision . Another precommitment strategy is to associate a punitive cost with the reward . For example, the drinker who takes Antabuse has, in effect, set it up so that alcohol consumption is inextricably linked with severely unpleasant physical sensations. Yet another strategy is to increase the value of the more distant reward: “If I can lose some weight , I will be able to wear my favorite suit again, something that’s much more attractive to me than the abstract prevention of cardiovascular diseases”. This increase in value can occur by cognitive reframing (as in the previous example), or by even attaching an additional reward to the original goal: “If I lose 30 lbs and can fit into my favorite swimsuit, I’ll go on a vacation to the beach.”

Nozick suggests that using a principle is a way to fight temptation (Nozick 1993, 17–20). In order to not start smoking again, I tell myself: I will never again smoke a pipe. The essence of a principle, for Nozick, is that it confers on the act of smoking this particular pipe the symbolic value of all future pipe smoking (Nozick 1993, 26). Speaking of this value of principles, we can mention here what are called New Year’s resolutions . Making them on this particular day makes them more effective, thanks to its symbolic strength. A study showed that the people who follow this tradition are more adherent to their commitments to lose weight , exercise or quit smoking than people who have the same objective at a different time of the year (Norcross et al. 2002), and a Canadian article linked this to the increase in cigarette advertisements in January and February, the goal being to counter these good resolutions, which are unfortunate for cigarette manufacturers (Basil et al. 2000).

This explains why using a principle allows one to resist temptation during stage B: This pipe that I run the risk of smoking becomes, symbolically, charged with all the future misdeeds of tobacco use, and thus its value cannot go up enough to make me cave in. And this is also a way to avoid a relapse when I have arrived at stage C for the intertemporal choice concerning that pipe. Violating my principle would symbolize the destruction of all the efforts I have made in refusing all of the preceding pipes. Hence, the sentiment “I’ve made it this far…” can serve to motivate us during difficult stage B temptations.

The use of a principle as described by Nozick is analogous to what George Ainslie calls a personal rule (but it is only analogous: We will see at the end of this chapter the fundamental difference between the two concepts). An individual bundles her temptations into a group, so that each choice involving a temptation becomes a precedent allowing her to predict all the future choices in this group. Thus, she becomes capable of emphasizing her desire to attain an important but distant goal each time she finds herself faced with a small step that would take her in the wrong direction (Ainslie 1985, 145). Ainslie suggests that this tactic is the same, at the moment of choice, as considering the sum of future choices. The respective curves of the two rewards, the shorter-smaller one (for example, the dessert), and the larger-later one (staying thin, for instance) are added together and end by taking on a quasi-exponential form that prevents them from crossing (Fig. 7.4).

On the left side of the figure (a), we can see someone headed towards the choice between a cake and the objective of staying in good health . The hyperbolical form of the curves describing the force of the respective desires makes them cross, and it becomes rational for the agent to order the dessert. On the right side of the figure (b), the agent believes that the choice she makes today commits her to future choices, and she adds the respective curves: They take on a form that does not allow them to cross.

In health maintenance, for example in the case of cardiovascular diseases, a major obstacle is that the efforts needed to maintain health are prolonged and (often) without visible benefit.

As a doctor , I believe we have here a key to improving patient adherence. Perhaps it is more effective to tell the person to whom you are prescribing a diet that, thanks to her efforts, she will simply feel better quickly or that she will be proud of committing herself to a health objective. To someone who wants to quit smoking , knowing that she will be able to go up the stairs without getting out of breath, or that she will no longer cough when she wakes up each morning, or seeing the relief on the faces of her spouse and children when they see her improved health are all more immediate and meaningful rewards. The doctor need not brandish the ghost of diseases the patient maybe has never heard of and that often remain just abstract scarecrows (perhaps measuring glycated hemoglobin—which reflects glycemic control in the diabetic patient over the preceding three months—also has an additional value: It acts as an intermediate reward ).

This approach calls for strategic use of emotions . Pierre Livet suggests that emotions may be another strategy for guiding and motivating behavior (Livet 2002, 80), as compared to precommitment strategies, which are not foolproof. Because rewards are in large measure emotional , it matters very much whether we experience an emotion now versus experience it in the future . In the example of the dinner, it is immediate pleasure of a piece of cake as weighed against the unpleasant emotions that I will feel tomorrow when I step on the scale and find out that I have ruined the dieting efforts of an entire week. Livet:

my emotions can lead me to the right path, because they oppose, maybe not the reality, but at least the impact of reality on my affects to the desires that don’t take them into account. Emotions are a reminding force for the desires […]. Nonetheless, emotions can have a reminding force, while the desires cannot. And Elster ’s objection that all that passion can do, reason will do just as well with less effort, does not apply here, because long-term reason does not have the force of reminder for short term reason, at the moment when the decision is made. The differential of emotions can, on the other hand, play this role.

We can suggest that a therapeutic technique to be developed could specifically consist of helping the patient identify the intermediate rewards that she imagines being able to obtain by accepting a treatment regimen, and to also explicitly address the emotions related to the alternatives she faces.

Olav Gjelsvik suggests that

a simple form of self-control of one’s mind in connection with quitting might be to concentrate on all the good things that accompany nonsmoking […] The result of the concentration might be a change in the present assessment of the value of nonsmoking compared with smoking . (Gjelsvik 1999, 57)

Obviously these ‘good things’ must be emotions . We saw that the concept of resolution implies that the agent is able to mentally repeat the value of the decision in the long run. This must be done in the most attractive way and Baumeister suggests that the vividness with which paradise and hell were represented in medieval imagery was not accidental (Baumeister and Voes 2003, 210].

In Alcoholics Anonymous , drinkers are advised to ‘think through the drink’, meaning to imagine the emotions they will experience if they pick up a drink, how they will feel about the reactions of others, etc.¹

Elster notes that it is debatable whether we should call the “caving in” during period B an incontinent action , because I chose to perform the action that I considered at that moment to be the best (point t2 in Fig. 7.3). It is only incontinent in relation to my general resolution of staying on a diet , which was still there when I entered the restaurant . In this case, it is strictly speaking an incontinent action only if the moment when the individual decides to perform the more immediate action, and the one that doesn’t seem the best to him, is situated at t1 in Fig. 7.3, i.e., before the reversal of preferences. And in fact, taking up the example of the restaurant, Elster also notes that this crossing is not necessarily related to the effects of alcohol or to seeing the dessert cart. It may have already occurred in the taxi on the way to the restaurant (Elster 1999a, 22). This last remark could lead one to question the very existence of incontinent actions (Elster 1999b, 425–445).

Here it is necessary to note the difference between two types of phenomena that are called patient nonadherence: On the one hand are the actions involving non-renunciation of pleasure , for instance staying on a diet (it is significant that Elster uses the example of giving up on a diet at the end of dinner) or refusing a cigarette . On the other hand are the various tasks that the patient must perform in the course of the treatment of a chronic disease: We have argued from the start of the book that to refuse a cigarette is not the same thing as to take a pill.

First, while there is a strong correlation between addictive behavior and an elevated discount rate, just as the Theory of Interpersonal Choice predicts (Bickel and Johnson 2003, 419–440; Perry and Carroll 2008, for a full discussion see Story et al. 2014), the correlation may be weaker in other health behaviors: We observed in a small number of type 2 diabetic patients an association between preference for a smaller-sooner monetary reward and adherence to medication and metabolic control (Reach et al. 2011) and in the study in obese diabetic patients, already quoted, where we demonstrated the association between seatbelt behavior and adherence to medication, we also observed that giving the priority to the future was an independent determinant of adherence to medication (Reach 2011). However, another study found only a weak relationship between the discount rate and such diverse health activities as exercise , weight loss, dental care , and seatbelt use (Chapman 2003).

Second, as Peter Herman and Janet Polivy show , it might be going a little too fast to take the so often used example of caving into eating dessert as the quintessential preference change described by the Theory of Intertemporal Choice . There are important differences between having to choose between $100 now and $200 next year and a dessert now and the possibility of staying thin: For one thing, it is not certain that I will in fact become thin if I do not eat this dessert (and previous unsuccessful attempts may reinforce this impression). In addition, for the monetary choice , you are offered a certain sum today and a larger sum tomorrow, while in the case of a diet , the future can be summed up as no cake today and still no cake tomorrow.

Nevertheless, the perception of time presumably plays a role in the decision to stay on a diet : Herman and Polivy consider that individuals who have a short life expectancy in front of them should be less inclined to stay on a diet (although they do mention the case of a man sentenced to death who for his last meal ordered a glass of Diet Coke! It is difficult to understand this except as the result of habit or of supreme irony). They admit therefore that experimental research is needed to clarify the relationship between perception of time and adhering to a long-term goal (Herman and Polivy 2003, 459–489).

Finally, Olav Gjelsvik discusses in detail these two conceptions of the weakness of will , Davidson ’s version that includes the phenomenon of akrasia , and that of Ainslie and Elster , which brings into play the reversal of preference (Gjelsvik 1999, 47–64). He shows that the two conceptions are not mutually exclusive; but he admits that there are at least certain cases where Davidson’s explanation is more apt, such as when an individual falls back into a bad habit . Consider, for example, consciously accepting a cigarette , even while you are well aware as you are lighting it that this cigarette might be the event that will make you relapse, with serious consequences for the future .

Inasmuch as I have described in the preceding chapter the importance of habit for the subject of our analysis, patient nonadherence, it seems pertinent to describe it, or at least some aspects of it, in terms of akrasia in the strict sense of the word. However, after what has been said, it is clear that a change of preference related to the way we discount the future can also explain these choices, which then would only seem to be irrational .

Here we have arrived at a key in our understanding of patient nonadherence: The perception of time intervenes in the choice between an adherent and a nonadherent action in all cases. In certain cases, particularly where addiction is involved, the Theory of Intertemporal Choice accounts adequately for the role of time perception. But in most other cases, it is necessary to find another explanation for the intervention of time in what appears to be an incontinent action. The question ‘why do we take care of ourselves?’ might now be reformulated as: How is it possible for us to not conform to our own principle of continent action? How do I not take my pills, or rather, how do I not make the decision to take my pills in the future ?

I propose that, in the particular case of akrasia represented by patient nonadherence to long-term therapies, there is such disequilibrium between the two types of actions, the continent and the incontinent , that it does not allow the principle of continence to express itself, or rather this principle becomes insufficient, or even inappropriate, if used alone. This leads me to propose a hypothesis where time plays a role in the partitioning of the mind proposed by Davidson in his explanation of irrationality ; this hypothesis introduces a second principle, which I will call the principle of foresight , which pushes us to give priority to the propositional contents related to the future , i.e., to accept taking care of ourselves.

In short, then, the conflict between our reasons for continence and those for incontinence is well-matched and understandable. We find our mind at war with itself. It is only by recourse to another guiding principle, that of foresight , that we can see our way out of this dilemma.

Let’s summarize. When we analyzed the taking of a pill as an action , it led us to consider patient nonadherence as a particular instance of incontinent action, or akrasia . As Davidson proposes in Paradoxes of Irrationality, incontinent action is caused by a motivation found in a part of the mind separated from the part containing the motivation that could have caused the continent action (which in our case would correspond to patient adherence). An akratic person’s irrationality lies in defying the principle of continence that tells her to take into account all the available reasons, in whatever region of the mind they may be, before acting, and to act on the reason that she finds, all things considered , to be the best. This evaluation brings into play beliefs and desires , but also, as we saw with Livet , emotions . Signaling to the agent a differential between the reality of the world and her assessment of it, certain emotions, such as fear , indicate to her that a revision of preferences is necessary, while others, such as revulsion, push her to resist and maintain her position.

In this investigation, I defined patient adherence as ‘to accept performing the repetitive actions recommended for a long-term health objective’. That is why patient nonadherence is most likely to occur in the case of a chronic illness , making it necessary to add the dimension of time to the analysis. Time may intervene negatively either by reducing the strength of the emotions that had pushed us at the beginning of the illness to take care of ourselves, or by leading to a change of preference at crucial turning points when nonadherence may occur, according to the mechanism described in the previous chapter. Boredom , the dull plodding of time, can deplete our ability to sustain adherence over the long haul. Energy for and interest in repetitive tasks wanes, and the resulting weariness can lead to giving up.

The desires and beliefs that drive the actions of adherence and nonadherence differ in their respective relationship to time : In a nutshell, the reward of adherence is far off, that of nonadherence is immediate. From this perspective, one might reasonably ask if the real question is not how is adherence, rather than nonadherence, possible, but rather: Why is anyone adherent? Why does one make the choice to take care of oneself?