A response to “Affective Forecasting and Its Implication for Medical Ethics” by Rosamond Rhodes and James J. Strain (CQ 17(1))
In the article “Affective Forecasting and Its Implications for Medical Ethics,” Rhodes and StrainFootnote 1 present current research documenting a psychological bias that affects our ability to correctly predict future emotional states. The bias they discuss is the wrongful estimation of the emotional impact of significant events. Research on affective forecasting by T.D. Wilson and D.T. GilbertFootnote 2 indicates that people tend to overestimate both the positive and negative impact of events. Rhodes and Strain correctly note that the conclusions derived from the research—that bad events do not seem to affect us as much as previously assumed—has applications for medicine. In a field where doctors are often faced with delivering bad news and offering treatments that will significantly alter the lives of their patients, information that bad events are not as emotionality deleterious as once thought is good news.
I argue in agreement with the authors that the research on the impact bias could support an attitude change in terms of how doctors approach and communicate with their patients. It can contribute, also, to a change in attitude when doctors are faced with the grim task of telling patients that they have a chronic illness or that they are faced with making difficult decisions about their health.
I disagree, however, that the predicted impact of the biases in emotional forecasting is as all-encompassing as the authors claim. In the next section, I consider the effects of biases in affective forecasting on policymakers. I assert that, although the research on the impact bias reveals misconceptions about what is needed for happiness, it does not provide guidelines for endorsing policies. In the third section, I consider how the bias affects doctors and patients differently. I argue that doctors are not affected by the bias in the same way as patients. In the fourth section, I argue against Rhodes and Strain and claim that affective forecasting does not diminish decisional capacity. I argue further that because the bias is incorrigible, paternalism will not help in restoring patient autonomy.
Values and the Impact Bias
Wilson and Gilbert's research purports to capture a universal psychological mechanism. As such, the bias affects all people who make predictions based on future emotional states. Rhodes and Strain claim that “once we recognize that affective forecasting is part of normal human psychology, we are alerted to consider just who in the medical environment is susceptible to its effects. Patients are. The families of patients are. Clinicians are. And policymakers are.”Footnote 3
Let us first consider the influence of this psychological bias on policymakers. The research on the impact bias cannot offer any help when it comes to policy issues because the results of the research are explanatory and not predictive. The research shows that we are not good predictors of our future emotional states but it does not allow for predictions about what will make us happy. Policies, however, are an attempt to promote or protect universally held values. If the studies on affective forecasting were to have an impact on policymaking, they should help in the selection values to be endorsed. I claim that the affective forecasting research cannot be used to do that.
I take affective forecasting research to show that there is a disconnect between some of our values, which are set in terms of what we expect will make us happy, and what actually makes us happy. The findings indicate that there is not a clear connection between what we value and pursue—like health, employment, wealth, and prestige—and happiness. In other words, the research shows that there is a cleavage between some of our current commonly held values and what we require for happiness. The disconnection between happiness and values was manifested in the following way: People tend to overestimate their emotional reaction to negative events, but also to positive ones.Footnote 4 It is not just that losing a limb does not make somebody as unhappy as predicted, but that positive events, such as job promotions, are not likely to make one as happy as predicted. Significant events do not seem to make a large, lasting difference on our emotional condition overall. Momentous events might cause a spike in positive or negative feeling, but one ultimately settles back to one's previous emotional state.
The impact bias exposes our deep ignorance about the nature of happiness. Doctors, policymakers, and patients share this predicament, but I do not take that to indicate that particular policies are the result of the impact bias. Rhodes and Strain's claim is that affective forecasting biases influence policymaker's decisions. Consider policies regarding promotion of health. We can conclude based on the new information about the impact bias that not being healthy is not as devastating emotionally as we once thought. Plenty of people can be happy even if they have lived through or are still living with illness. We may not need health for happiness. But does this mean that policies promoting health are based on biased affective forecasting? I assume that most people would disagree with that conclusion. Even if we decide that health is no longer worth promoting, predictions about which values should be promoted cannot be based on affective forecasting.
If emotional states do not trail or depend on the events in one's life, one can only conclude that what we thought were necessary ingredients for happiness are not essential to it. Policies are meant to endorse generally accepted values that rest on ideas people have about what is good or worth promoting. If we do not require good events, or the absence of bad events, for happiness, one has no resources left for making predictions about what to promote. One can conclude that the impact bias research indicates that illness might not devastate your life but that research does not provide guidance in the selection of policies pertaining to healthcare.
The Impact Bias from Different Perspectives
Although research on the impact bias supports the claim that forecasting biases are part of human psychology, the way in which such biases affect reasoning is different for patients and those deciding for the patients.
Affective forecasting is a first-person phenomenon. The research on the bias supports the view that we tend to be wrong about our own reactions to events. From the first-person point of view the bias is incorrigible, because a person who has gone through bad events is likely to continue wrongful estimates about her future emotional states. In fact, we tend to be wrong about how we feel all the time. In a review of the literature on emotional memory, Christiansen and Safer concluded that “there are apparently no published studies in which a group of subjects has accurately recalled the intensity and or frequency of their previously recorded emotions.”Footnote 5 This conclusion can be explained in part by appealing to the workings of the “psychological-immune system,”Footnote 6 so termed by Gilbert and Wilson. A person recovers emotionally from a traumatic event by making sense of it.Footnote 7 The negative event becomes normalized and its emotional impact is less likely to be remembered accurately. Past good or bad events become less extraordinary because of the ameliorating effect of the psychological immune system. Thus, the disconfirmation of the prediction remains unnoticed because our memory of past emotional states is often inaccurate. If we cannot draw on past experience to generalize for the future, the impact bias continues to affect our predictions about emotional states.Footnote 8 The brunt of this research indicates that, when we are forecasting our own future emotional states, the prediction is plagued by errors. The studies, however, do not show that the errors occur when we make prediction about other peoples’ future emotional states.
The impact bias may not persist from the third-person perspective. In other words, it might be a problem for me but not for my doctor. At the very least the problem is not of the same nature from the two perspectives. My doctor might know that I will do much better dealing with my illness than I expect. She might be able to reason free from bias when it comes to my case. The research itself gives no indication that people deciding for others are susceptible to the influence of affective forecasting. Studies only show that people are affected with this bias when they are trying to predict how this or that event will make them feel. Conceivably the doctor might reason by adopting the patient's perspective and imagine what it would be like for her to be in the patient's condition. If the process of reasoning is simulating the patient's point of view, the effects of the impact bias could be transferred to the doctor. It strikes one as much more likely that when a doctor is thinking about future states of her patient, she is probably doing so based on probabilities about what happens on average for a particular population. The doctor's reasoning will be based on her experience with past patients with similar conditions. Moreover, if the doctor takes into account information about the impact bias, she may reach a different conclusion than the patient. The doctor's forecast, then, will be free from error.
In contrast to the patient, the doctor whose reasoning might be influenced by the impact bias is not in the same predicament as the patient. There is no reason to argue that the bias is incorrigible from the third-person perspective. Familiarizing the doctor with the bias is likely to influence her thinking about her past and future patients. If the doctor has not already concluded from experience that people deal with illness better than expected, the affective forecasting literature should successfully alter the doctor's point of view about delivering bad news and help improve communication with the patient. The problems that plague emotional memory from the first-person perspective do not apply to the doctor. Also, the psychological coping mechanisms that, in part, cause the bad memory of emotional events are specific to the first person and have no consequences on people reasoning for others.
Clinicians reasoning about patients’ future emotional states are less likely to be affected by some of the suggested causes of the bias. I number some of the sources of the bias and then illustrate how they differ in their effects on reasoning of doctors and patients. One of the sources of error is undercorrection for the passage of time. People forecasting their future emotional states take into account only how they will feel right after a negative event and fail to consider that first-blush emotional reactions will change after some time. A person reasoning about how she would feel after being diagnosed with a chronic illness is likely to make a false prediction because she does not take into account that her unhappiness will abate over time. A doctor attempting to make a prediction for her patient is more likely to take into account the passage of time because the prediction is based on different facts, such as her experiences with patients suffering from similar conditions. The doctor knows that her other patients’ negative emotional reaction abated after the initial diagnosis and can assume that the same will happen in the case of the current patient. The doctor's reasoning is much less likely to be biased by subjective factors that are endemic to personal decisions. When reasoning for others, people are more likely to make decisions objectively and in accordance with facts.
Another source of error in affective forecasting is framing effects: People tend to focus on the difference between two alternatives rather than on similarities. For example, the patient might focus on the difference between her life before and after being diagnosed with an illness rather than on the similarities. So, the consequences of illness and treatment stand out, and the impact of the event is exaggerated. The patient fails to take into account that much about her life is likely to remain unchanged.
A doctor thinking about her patient might, however, be able to normalize the negative event by reasoning based on previous experiences with her patients. Clinicians are able to track the progress of their patients after the initial diagnoses and notice improvement. The doctor, unburdened by the first-person errors in emotional memory, is able to holistically appraise the situation, take into account all the factors that will contribute to the person's future emotional states, and avoid the framing affects.
A further source of the impact bias is the projection bias where people fail to adjust their predictions to accommodate for change. Life at the time when the patient is faced with the decision to accept a particular treatment could be different from life after the treatment. Positive events like the birth of a grandchild or a promotion and negative events like death in the family or financial troubles could improve or aggravate our emotional states, respectively. All of those things seem to be ignored when predictions about future emotional states are made. The projection bias, like the other sources of the impact bias, is made from the first-person point of view. There are no indications that doctors are affected by those same errors when predicting their patient's future emotional states. Similarly as with the framing affects, a doctor familiar with the progress of her past patients will be able to apply that experience to her current and future patients.
The remaining issue is how to reconcile reasoning from the first-person point of view with what we know about the impact bias. Rhodes and Strain note: “When affective forecasting persists it can have serious consequences for the affected patient… . Some patients may avoid seeing a doctor, … others may avoid disclosing problems or symptoms.”Footnote 9 In addition, the patient might refuse treatment because he fears not being able to cope with the consequences. Inasmuch as affective forecasting is largely to blame for the listed avoidance behaviors, it is important to consider how the doctor might communicate with the patient.
Reasoning Free from Error
Becoming aware of the impact bias and noting its potential effects on decisionmaking in medicine raises issues of whether one should interfere in order to prevent such distortions in judgment. The impact bias might put a dent in our regard for patient's autonomy. Respect for autonomy calls for the doctor to respect the patient's decisions in all cases in which the patient has decisional capacity. To have decisional capacityFootnote 10 patients must understand the information presented to them (e.g., the severity of the diagnoses and possible treatment options), appreciate the information given to them, and be able to apply the information to their own case. In addition, Gert et al. propose that in estimating capacity the doctor is also judging the rationality of a patient's decision, which is evidenced in the patient's ability to articulate reasons for refusal. This is supposed to cover cases where patients are able to pass the Understand and Appreciate (U+A) criterion, but the decision to refuse treatment is deemed irrational by most people. The definition of capacity is task specific; so when a doctor is evaluating capacity, he or she is estimating the patient's ability to make a decision about a particular medical treatment. Rational thought can be impaired by mood disorders like depression or a phobia. Rhodes and Strain assert that affective forecasting is comparable to a “psychological distortion”Footnote 11 that disables decisional capacity.
I argue that the impact bias dos not impair capacity. As stated previously, the only case in which we need to worry about the damaging influences of the impact bias is when the person is reasoning from the first-person perspective and making decisions for herself. In considering whether the patient deciding for or against a particular treatment is making her decision voluntarily, we must first gauge whether the patient understands and appreciates the information at hand. Affective forecasting does not seem to interfere with a person's ability to understand the diagnosis, prognosis, and likelihood of risks and benefits and treatment alternatives. The patient should also be able to apply all that information to himself in order to appreciate the information presented about diagnosis and prognosis. Biased forecasting about future emotional states seems irrelevant in both of those steps necessary for decisional capacity, because I limit the U+A criterion to medical information about treatment alternatives. The impact bias also seems not to interfere with a patient's ability to make and communicate a decision that he has made. The last component of capacity pertaining to rationality rests on the patient being able to articulate the reasons behind his decision to refuse treatment, for example. The reasons cited by the patient should be in accordance with the patient's values. At this juncture problems with affective forecasting could arise. Each person has her idea of what constitutes a good life, which rests on convictions about what is necessary for happiness. The person deciding about undergoing medical intervention will attempt to forecast her emotional states in accordance with the outcome of treatment.
Based on the research about affective forecasting, any person reasoning in this way is likely to be wrong. The cause of the problem is not, however, that the person has inconsistent reasons for her dissent; the patient is, in fact, reasoning in accordance with her values. The root of the problem is that what most of us value turns out not to be what is required for a happy life. I see this as the outcome of the affective forecasting research. If my rendition of the problem is correct, the patient making decisions under the influence of the impact bias has capacity. The issue is not that the patient in question cannot relate her decision to previously endorsed values; rather none of us are able to correctly cite the values needed for a good life.
Rhodes and Strain suggest that “some degree of paternalism may, therefore, be justified to prevent people from making decisions based on distorted estimates of their future responses.”Footnote 12 According to Gerald Dworkin, paternalism not justified by diminished capacity is “hard paternalism.”Footnote 13 Hard paternalism is not restricted to patient refusal only in cases where the patient is deemed not to have decisional capacity. I do not, in this paper, argue about whether such a form of paternalism is a good idea. Let us consider the claim that some degree of paternalism would be needed to restore the autonomy of the patient and help her in making a decision free from error or distortion.
Before I continue with my argument I wish to note that if it is indeed true that the impact bias is a universal psychological mechanism that affects all in the same way, there are problems with defining “reasoning free from error.” In case we all reason similarly about future emotional states, one would have to posit an idealized criterion for correct reasoning in order to qualify reasoning affected by this type of bias as erroneous or skewed. In most cases what is considered rational or normal would be defined in terms of an average, and in this case the average person's reasoning is influenced by the bias.
According to Rhodes and Strain, paternalism is needed to restore autonomy. Instances where attempts are made to restore autonomy to patients are cases where the patients are in the grip of a temporary or treatable mental ailment. The patient who is depressed and refuses treatment will most likely make a different choice if her depression is treated. Persons suffering from bipolar disorder might think that they are impervious to illness when in a manic phase. For patients in those scenarios, there is a way of rectifying the problem. Once the person is on medication her reasoning will be free from error and distortion.
The impact bias is not analogous to those instances. The bias is incorrigible; it is not a temporary ailment impeding the reasoning of a patient. People do not reap the benefits of past emotional recovery because they do not accurately remember them. Consequently, even people who have overcome grave circumstances in the past will continue making the same errors in affective forecasting. In that sense, one cannot restore autonomy, because for most the baseline is biased reasoning. In the case of the depressed or the manic patient, the deviation from normal reasoning is more obvious. Further still, there is an established way of treating the patient. Given that the impact bias is a more ubiquitous and permanent feature of human psychology, it becomes more difficult to identify it as a “psychological distortion.” In conclusion, I assert that the impact bias is not a sign of abnormal reasoning and thus not an indication that a person does not have autonomy.
Conclusion
The discovery of affective forecasting biases reveals something about human psychology that some of us grasped intuitively. People tend to be misguided about what constitutes a good life and often find that what they assumed they would need for a happy life they can do without. We realize—mostly when thinking about other's lives—that people overestimate both the importance of desirable goals, such as jobs, attractive spouses, or wins in sports, and the devastation of negative events, such as dissolution of marriages or death of close relatives. For others, we are often able to notice errors in reasoning but still continue making the same mistakes when thinking about our own future.
The interesting work by Gilbert and Wilson points to a problem that one often encounters in the medical setting: People's ideas about illness do not always approximate the reality of living with a health condition. One might even conclude that most people cannot accurately predict what it will be like for them to live with or face an illness. This insight might help doctors differently than patients. It might help doctors deal with the onerous psychological burden of having to tell the truth to their patients when the truth is not anything anybody would wish to hear. For patients, the benefits of our psychological coping mechanisms are implicit in the process of recovery. When faced with illness, one might not take comfort in the information that things will get better with time or that one is more likely to feel better than one expected. At the very least, knowing about the literature on the impact bias might not have such a large effect, because some of us already knew that one can recover even from the worst things in life.
The disparities between what we know and how we reason are part of what restricts the benefits of affective forecasting research. In medical ethics, the impact of the bias is not as extensive as claimed by Rhodes and Strain. The information about the bias cannot be used to draw conclusions about policy because it does not help identify any new values and goods to be promoted. Furthermore, the impact bias influences doctors and patients in different ways, which prevents the far-reaching consequences cited by the authors. More generally, the biases do not affect people deciding for others as they do people making decisions for themselves. Finally, the biases do not affect capacity in a way that would justify paternalism because their ubiquity and permanence seem to indicate that they are part of the baseline for human reasoning.
