### A Heuristics Approach to Genetic Counseling (MAIN TEXT)

 “But in this world, nothing can be said to be certain, except death and taxes.” – Benjamin FranklinIntroduction            Sometimes all the variables of a situation are known and it is possible to carefully and objectively choose the best option. If I am shopping for milk and have the opportunity to buy either a quart of milk for \$3 or a half gallon of milk for \$3, it seems obviously more rational to opt for more milk at the same price. This would be the rational decision as understood under a normative model of decision theory (Simon 1978), or the classical model of rational choice (Gilovich and Griffin 2003). I assume that the brand of milk is the same and that they are both available at the same store, kept in the same refrigerator, stored at the same temperature, and have the same expiration date. All the variables of the situation are known. I can systematically consider each one and arrive at the best decision. This perfect knowledge is implicit in classical models of rational choice and is consistent with what Savage, the founder of modern Bayesian decision theory, called small worlds (Gigerenzer and Gaissmaier 2011). These are decisions under certainty.       Any example of a decision made under this classical model, however, is necessarily simplified and decontextualized from the complexity of real life. Buying a half gallon rather than a quart of milk would also mean that I now have a heavier bag of groceries to carry home. I now also have a bigger carton of milk to fit into what may be a small and already crowded refrigerator. I may not be able to finish the milk before it spoils. Maximizing the purchasing power of my \$3 in buying milk is one goal. These other, complicating factors reveal other goals. What is optimal for one goal may be a poor choice for another goal (Simon 1956). When these additional variables and conflicting goals are considered, spending \$3 for a quart rather than a half gallon of milk may be seen as the more rational decision.            Under a classical model of rational choice, the probabilities and utilities of each possible outcome are carefully weighed to calculate the optimal decision (Gilovich and Griffin 2003). In some situations, these many variables may be known, but they may be too numerous or too complex for our brains to be able to perform a systematic calculation and come to a decision that would conform to a normative model of reasoning. Or the time and energy required to perform such a calculation may not be worth the effort, considering the \$3 that is at stake.Often, however, all of the information necessary to systematically reach an optimal decision is simply not available. I may not know how much space is left in the refrigerator. I may not yet know how many days of that week I am likely to eat at home and so may not be able to estimate how much milk I will likely consume before it spoils. I may not be able to anticipate whether or not I will have guests and whether or not they will take milk with their tea—or whether they might, quite unexpectedly, request hot chocolate, suddenly wiping out a significant portion of my milk supply. In this case, my decision is more properly conceived of as a gamble (Albert 1978; Palmer and Sainfort 1993). This imperfect knowledge is consistent with the concept of large worlds (Gigerenzer and Gaissmaier 2011). These are decisions under uncertainty.In these situations, when we find ourselves compelled to make a decision under conditions of uncertainty, how do we decide? When systematic reasoning is not feasible, we rely on heuristics: cognitive shortcuts that have been molded by prior knowledge and experience.Genetic counseling involves encounters between counselors and patients in a large world who often find themselves forced to make important decisions with limited information. The first part of this paper will provide an overview of the concept of heuristics and heuristic versus systematic reasoning. It will also present the three heuristics which have been most commonly cited within the genetic counseling literature. The second part of the paper will explore the relevance of these concepts to genetic counseling, reviewing the potential for heuristics to form an evidentiary basis for the utility of genetic counseling and presenting the potential for a knowledge of heuristics to inform the practice of genetic counseling as well as research within the field. The final part of this paper will introduce a second thread of heuristics research, one that has had a much smaller impact on the medical and genetic counseling literature to date. This alternative perspective will be proposed as essential to a balanced view of heuristic reasoning and uniquely aligned with the values and goals of genetic counseling. Part I: Heuristics, biases and their application in the genetic counseling literature What Are Heuristics?The general concept of heuristics is much more widely recognized and appreciated than is the term itself, which has mainly been used within the field of cognitive psychology. In recent years, many popular non-fiction books have been organized around the idea of heuristics, although the term itself is not often employed. A well-known example is Malcolm Gladwell’s Blink (2005), which gives some attention to errors in reasoning but is perhaps best known for presenting what is often the uncanny accuracy of intuitive judgments.The use of the term “heuristics” has changed over time, and many definitions of the term have been proposed and debated (Gigerenzer and Gaissmaier 2011). Heuristics are often briefly described as cognitive “rules of thumb” (Simon 1978; Peters et al. 2006).  More specifically, “heuristics” has been used to refer to quick, automatic, often unconscious processes that ignore part of the available information, as opposed to slower, more deliberate and conscious processes that take each variable into account and are labeled “systematic reasoning” (Gigerenzer and Gaissmaier 2011). More recent definitions have also asserted that heuristic processes need not be unconscious but may be used with awareness (Gigerenzer and Gaissmaier 2011). These two distinct cognitive processes are also often referred to as System 1 and System 2, respectively (Gilovich and Griffin 2003; Evans 2008; Weber and Johnson 2009; Croskery 2009).In 1953, Herbert A. Simon first introduced the concept that would come to be called bounded rationality. Simon postulated that, given the brain’s limited information processing abilities as well as the rarity of perfect knowledge of any situation with all of its variables, people are compelled to make simplified models of a complex world (Simon 1955). Simon further posited that since any given organism has more than one goal, the means to achieving one or more of these goals will necessarily diverge from optimal, as what is optimal for one goal may be a poor choice for another goal. With the ideal of optimization elusive and only limited cognitive resources, the strategy becomes more focused: “we are concerned only with finding a choice mechanism that will lead it [the organism] to pursue a ‘satisficing’ path, a path that will permit satisfaction at some specified level of all of its needs (Simon 1956).” Satisficing, considered the first “heuristic” to have been described, is a decision-making process whereby cognitive processing is engaged only until a first option meeting minimal standards is identified. This heuristic strategy contrasts sharply with the utility maximization standard of normative, or rational choice models, which assumes a more careful and systematic consideration of all available options and which had previously dominated psychological as well as economic theories (Simon 1956; Kuhberger 2002).The introduction of the concept of heuristics marked an important shift in the focus of cognitive psychology. The recognition of consistent, non-rational patterns of thinking also gave rise to the field of behavioral economics and has had a significant impact on other fields, from law and political science to public policy and medicine, inspiring hundreds of research studies (Gigerenzer and Gaissmaier 2011; Gilovich and Griffin 2002). Heuristics as a source of errors in judgment: Kahneman and Tversky’s Heuristics and Biases ProgramIn the early 1970s, a series of papers by Daniel Kahneman and Amos Tversky expanded on Simon’s ideas, focusing on systematic errors arising from various “heuristics,” or specific, simplified constructs used in making judgments. Kahneman and Tversky’s seminal article “Judgment under Uncertainty: Heuristics and Biases,” published in Science in 1974, has since been cited approximately 11,234 times (Google Scholar).  Researchers in the 1980s struck a balance between the previously held view of a fundamentally rational mind and the new model of a mind prone to any number of errors. Critics of the Heuristics and Biases Program, such as psychologist and researcher Gerd Gigerenzer, have called attention to the complexity of the context in which real-life decisions are made, as opposed to the carefully defined parameters within which Kahneman and Tversky conducted their experiments. His Fast and Frugal Program emphasizes the adaptive nature of heuristics, given the boundedness of rationality (Gigerenzer and Brighton 2009). In medicine, however, the idea of heuristic thinking as a source of biases and errors in judgment has continued to dominate the literature (Gigerenzer and Gaissmaier 2011). A recent review of the literature conducted for this paper suggests that this emphasis on heuristics as a source of cognitive errors also holds true for genetic counseling. Sixty-one articles published between 1974 and 2010 were identified as relating to genetic counseling and including some discussion or even a brief mention of heuristics. Of these 61 articles, nearly three quarters presented heuristics as a source of bias leading to cognitive errors. (For a complete listing of the articles identified in the literature review, see Appendix A. For a description of search methods and parameters, see Appendix B. For categorization of the various articles, see Appendix C.) Kahneman and Tversky’s Heuristics in the Genetic Counseling literatureThe first three heuristics described by Kahneman and Tversky in their papers from the 1970s are those that have been most cited in the genetic counseling literature:   1. availability, 2. representativeness, and 3. anchoring and adjustment. This paper will focus on these three heuristics only. Later research by Kahneman and Tversky was focused on framing effects, or the way that the perception of information is affected by the way that information is presented, or framed (Kahneman 2003). The concept of framing as a heuristic has been discussed within the genetic counseling literature nearly as much as anchoring and adjustment. (See Appendix C.) Much more often, however, it has been discussed within the context of risk communication, absent an explicit discussion of heuristics.  Many of the findings related to framing and the provision of risk information in the healthcare setting are well documented. A review of the discussion of framing in the genetic counseling literature, however, is beyond the scope of this paper.The following sections will present each of the three heuristics originally described by Kahneman and Tversky, followed by a few examples of how these heuristics have been applied in the genetic counseling literature. For a complete listing of the articles referencing each of the following heuristics, refer to Appendix C. AvailabilityAvailability describes the tendency to assess the probability of an event based on the ease with which similar instances can be brought to mind (Tversky and Kahneman, 1973). This ease of recall of previous events was seen to be a product of the frequency of events but also other factors, such as the recentness and salience (Tversky and Kahneman 1974). To illustrate the effect of salience, Kahneman and Tversky provided the example of a clinician, presented with a patient who expresses suicidal thoughts. The clinician is most likely to recall not the actual frequency of previous patients expressing similar thoughts of suicide but the specific instances of previous patients who had expressed similar thoughts and subsequently attempted suicide (Tversky and Kahneman, 1973). Availability is mentioned in nearly three quarters of the articles identified in this literature search. Lippman-Hand and Fraser first suggested the role of the availability heuristic in genetic counseling, describing one woman’s perception of a high risk for having an affected child, given that she already had one affected child:The heuristics described by Tversky and Kahneman [1974] to explain how people process probabilistic information could, in these situations, influence what a parent imagined so that the consequences considered were those that were more familiar or more easily imagined rather than all the possibilities or even the ones that were “factually” most pertinent. (Lippman-Hand and Fraser 1974 b) A study by Wertz et al. also found that having an affected child in the home was associated with more pessimistic interpretations of the recurrence risk” (1986). Similarly, in an article that makes no explicit mention of heuristics, Fanos and Johnson reported on their findings from a study of the perception of carrier status among siblings of children affected with Cystic Fibrosis. Last-born siblings were more likely to believe they were not carriers. The six last-born siblings who did not believe themselves to be carriers had never lived at home with the affected sibling: either the affected child had died before the birth of the last-born or was no longer living at home by the time the youngest sibling was born (1995).Shiloh and Saxe have asserted that in genetic counseling, the availability heuristic implies that “the more familiar the disease, the higher will be the perceived risk” (1988). In a recent study, Katapodi et al. found individuals’ estimates of their personal risk for developing breast cancer to be more strongly related to a personal experience with cancer than to a factual understanding of family history combined with other of risk factors, including heredity (2010). These findings are consistent with those of many other studies which have shown an association between personal experience of cancer and increased personal risk estimates. (Refer to Appendix C for a complete list of articles identified as discussing both cancer and the availability heuristic.) Representativeness         Representativeness was originally described as the tendency to evaluate the probability of an entity or event belonging to a certain class by the extent to which it is representative of, or resembles, another event or entity. A classic example is judging the probability that someone is engaged in a particular occupation by the extent to which the description of that individual matches the idea of the “typical” person in that particular occupation. In their widely cited 1974 paper published in Science, Tversky and Kahneman describe Steve as “very shy and withdrawn, invariably helpful, but with little interest in people, or in the world of reality. A meek and tidy soul, he has a need for order and structure, and a passion for detail.” Tversky and Kahneman demonstrated in an experimental setting that, based on this description, people will use the representativeness heuristic to misjudge the probability that Steve is a librarian rather than a farmer. Such a judgment demonstrates neglect of the base-rate frequency, or prior probability of such an outcome. In this case, use of the representativeness heuristic ignores the fact that there are many more farmers in the population than there are librarians (1974).          Representativeness is cited in about half of the articles identified by this recent literature search. Wertz and Fletcher defined representativeness as “the extent to which a particular example of a negative outcome is regarded as stereotypical of all negative outcomes” (1987). This heuristic may make it even more difficult for patients to grasp the concept of variable expressivity:In genetics, where many disorders have a wide range of severity, prospective parents may focus on one end of the scale—for example, viewing a Down syndrome child with close-to-normal IQ and capable of independent living as representative of all children with Down syndrome. They will interpret their risk as lower than will a couple who regard a profoundly retarded, institutionalized child as representative of those with Down syndrome. (Wertz and Fletcher 1987)          Kessler had described the family system phenomenon of patient pre-selection among families with Huntington’s, whereby an asymptomatic individual in the family is unconsciously singled out and assumed to be the one affected in the family (1988). Shoshanna Shiloh proposed that pre-selection can be seen as an example of the representativeness heuristic: “Preselection is often connected with family beliefs about similarity (representativeness) of the preselected individual to the sick parent” (1994).         In the case of breast cancer, Gerend et al. (2004) found that a woman’s perceived risk of developing breast cancer was influenced by how similar she believed herself to be to the “typical” woman who gets breast cancer. This effect was seen both with comparisons drawn between a woman and another family member and with a woman’s perception of herself compared to women in the general population.         The representativeness heuristic also encompasses neglect of the effects of a small sample size and the “law of small numbers,” whereby the outcome of a relatively small number of events is assigned statistical significance that is not merited. Kahneman and Tversky gave the example of the prediction that a coin tossed only six times would still produce results near 50/50 for heads and tails. This belief reflects the expectation that “a sequence of events generated by a random process will represent the essential characteristics of that process even when the sequence is short” (1974). In reality, a much longer sequence of events would be necessary to observe the 50/50 segregation of results generated by chance.         Shiloh and Saxe referenced neglect of small sample size in explaining patients’ tendency to judge their future risk of birth defects based on their own family history of birth defects: “This tendency persists despite the fact that any family’s size is too small to be a representative sample for the probability of recurrence and seems to rely on basic intuitions about uncertain events” (1988). Shiloh also applied this heuristic to parents’ perception of the recurrence risk for autosomal recessive conditions: “thus, if the chances of having an affected child within a family are 25%, and the couple has three children, all of them affected, they may believe that the recurrence risk for the next pregnancy is smaller than 25%” (1994).                          In their study comparing biases in genetic problem solving among undergraduates, genetic counseling students and genetic counselors, Dewhurst et al. (2007) included a birth order problem modeled after Kahneman and Tversky’s coin toss. Participants were presented with five hypothetical families, each with a different number of children and various boy/girl children combinations, and asked which birth order was most likely to occur. The correct answer is the birth order for the family with the fewest children. The smallest family, however, had three girls and no boys. Among undergraduates, only 21% of the sample (n=220) chose the correct response. Among genetic counseling students (n=89), 35% of the students selected the correct response, and among genetic counselors, 47% (n=156) selected the correct response. For all three groups, the most common incorrect responses were birth orders that had the most even distribution of boys and girls.         Klein and Stefanek cited neglect of small sample size in patients’ decisions regarding cancer treatments: “thus, when told that a certain treatment is 75% effective, they may be insensitive to whether this success rate is based on a small sample of patients (eg, 4) or a much larger sample (eg hundreds or thousands)” (2007).      Anchoring and AdjustmentAnchoring and adjustment describes the effect of an initial value on the estimate or the perception of a final value. Whether the initial value is suggested by an external or internal source, it is postulated that the heuristic mind treats this starting point as an anchor, from which adjustment to a final value then occurs. The resulting bias has its roots in a failure to make sufficient adjustments from the given anchor. In one experiment, participants were first primed with an arbitrary numerical value between 0 and 100 by spinning a wheel. They were then asked to estimate the percentage of African countries in the United Nations. Estimates given were found to be strongly influenced by the arbitrary numerical value generated by the wheel, which had served as an anchor. Groups primed with the number 10 gave a median estimate of 25%, whereas those groups primed with 65 gave a median estimate of 45%.Anchoring was referenced in a little over a quarter of the articles identified in the literature search. Two separate studies conducted in the 1980s found women’s perception of risk pre-counseling to be the best predictor of their perception of risk post-counseling, carrying more weight than the numerical probability that was given during genetic counseling (Wertz et al 1986; Shiloh and Saxe 1989). More recent articles have also supported the effect of anchoring. (See appendix C for a complete listing of articles discussing anchoring.) In a review of multiple studies on the communication of disease risk, Senay and Kephingst (2009) found that patients’ initial perception of their risk for breast cancer was a factor in their ultimate understanding of risk, in that their estimates were not sufficiently adjusted (either higher or lower) after being presented with their objective risk for breast cancer.  Part II: Implications of heuristics for Genetic CounselingThe preceding section laid out the most commonly discussed heuristics in the genetic counseling literature. These examples have been principally descriptive in nature, providing possible explanations for behavior and decisions, but without a prescriptive component. In the following section, a case is made for the relevance of heuristics to genetic counseling, not merely as a way of better understanding the genetic counseling process but 1) as a potential evidentiary basis for the value of genetic counseling; and 2) as a means of informing the practice of genetic counseling. Heuristics as a Potential Evidentiary Basis for Genetic CounselingCommon triggers of heuristic thinking         Since the initiation of Kahneman and Tversky’s Heuristics and Biases Program, numerous studies have identified specific situations in which heuristic rather than systematic forms of reasoning are more likely to be used and lead to errors of intuitive judgment: when under time pressure; when distracted and under stress; when multitasking; when something is a familiar, rather than a novel experience; when presented with a complex problem; when performing a task in the evening for “morning people" and when performing a task in the morning for “evening people;" when in a good mood; in emotionally charged situations; when there is ambiguity, when probabilities are involved; and when the situation is judged to be of limited personal importance (summarized in Kahneman 2003; Peters et al 2006; Steginga and Occhipinti 2004). Many of these conditions typify the situations of patients presenting for genetic counseling. The Interplay of Two Systems: The Dual Process ModelDual processing theories have been popular for over 20 years, both within and beyond the field of psychology. Multiple researchers, including Kahneman & Frederick and Chaiken, have proposed dual process models that encompass heuristic and systematic reasoning processes and attempt to describe how the two systems interact (Evans 2008). Debate has been ongoing over the question of how this occurs, and the question has been raised as to whether alternative models might help elucidate these cognitive mechanisms (Weber and Johnson 2009). But as Gigerenzer and Gaissmaier have written “All models are wrong. But some predict better than others and lead to novel questions” (2011). Dual processing models help to explain why heuristic biases are not consistently observed in experimental settings. Rather, normatively correct answers suggestive of systematic reasoning are also given (Evans 2008).Heuristics are often default responses. Significantly, however, studies have shown that a tendency toward heuristic reasoning can be suppressed and that it is possible to intervene to encourage systematic processing (Evans 2008). If both systematic and heuristic processes operate simultaneously and in conflict with one another, genetic counseling represents a unique opportunity to intervene in real time and encourage patients to engage in systematic rather than heuristic reasoning. Systematic reasoning as a pre-requisite for informed consentMany experiments have shown that in some situations, heuristic reasoning is associated with better choices. In one study, students were asked to taste and rate five jams. Some of the students were also asked to analyze their feelings about the jams in preparation for the evaluations. The students who were not asked to analyze the jams gave ratings more consistent with those of Consumer Reports. The authors concluded that prompting students to engage in analysis (systematic reasoning) in this situation caused the students to focus on “nonoptimal criteria” rather than simply choosing what they liked best. This focus on relatively insignificant criteria in turn formed the basis for a poorer judgment than would otherwise have been made (Wilson & Schooler 1991). In medicine, however, systematic reasoning is often implied as an essential part of an informed consent process (White 1997; Holmes-Rovner and Wills 2002; Van den Berg 2006; Etchegary and Pierre 2007). Research outside the field of genetic counseling has found that patients arriving at decisions by heuristic processes are less likely to be satisfied with their decisions long-term. Philosopher Abraham P. Schwab has suggested that “bounded cognition marks the limits of formal accounts of autonomy” (2006). As an alternative to formal autonomy, he proposes “effective autonomy,” predicated on informed consent arrived at by systematic reasoning, guided by debiasing techniques. (For discussion, see Schwab 2006.)            If informed consent is predicated on the engagement of systematic reasoning processes, documentation of those situations and psychological states that predispose patients to heuristic biases could amount to an evidence base for the utility of genetic counseling. If more data are collected indicating that patients faced with uncertainty and decisions regarding genetic testing are likely to engage in heuristic reasoning—leading to choices which they may not be satisfied with longer term—genetic counseling that specifically encourages systematic reasoning could be viewed as a pre-requisite to informed consent. Establishment of an evidentiary basis for the utility of genetic counseling            Ideally, such an evidentiary basis would be established by those outside the field of genetic counseling. A significant amount of research has established the confirmation bias: a tendency to selectively search, recall and assimilate information in a way that supports or confirms one’s own initially held belief, or favored hypothesis (Arkes 1991; Chapman and Elstein 2003). Any research conducted with the aim of assessing the utility of genetic counseling would therefore be more credible were it to be conducted by relatively disinterested experimenters. Psychologist and researcher Shoshanna Shiloh has introduced the field of genetic counseling to psychologists in multiple articles, commenting on the potential gains for both fields (Shiloh and Saxe 1989; Shiloh 1994; Shiloh 1996b). For the field of cognitive psychology, genetic counseling represents a unique context for the study of heuristics: a relatively controlled setting, often with an objective risk component that is quantifiable. In reporting on their study of the perception of risk recurrence in a prenatal setting in 1989, Shiloh and Saxe called attention to the potential for heuristics research in a genetic counseling setting:Almost all cognitive research on risk perception and decision-making is done in laboratory studies. In spite of the recognition of the artificiality of this kind of research (Ebbesen and Conenci, 1980), field studies are usually avoided because in them, probabilities and utilities cannot be adequately measured (Kahneman and Tversky, 1979). The present research presents a unique real-life situation in which exact probabilities are provided to individuals, and expected to be used in a real and crucial personal decision. Of the 61 articles identified in this literature search, however, most have only conjectured that patient’s risk perceptions or behaviors may have been attributable to heuristics. Less than a quarter were reporting on research involving heuristics. (See appendix C.)In great contrast to the study and application of heuristics and decision making science to public health policies and campaigns, genetic counseling represents an opportunity to engage in a dialogue, to intervene in real time and to note the effects of different counseling techniques on patients’ risk perception, attitudes, anxiety levels and other factors. Specifically, researchers could observe the effects of varying counseling approaches in helping patients shift from heuristic to systematic reasoning. Heuristics as a means of informing genetic counseling practice and researchIf systematic reasoning is necessary for informed consent, it follows that genetic counseling that is informed by a knowledge of heuristic and systematic processing may be more effective in achieving the goals of genetic counseling, which include informed and autonomous decision making. An understanding of cognitive processes, in both patients and practitioners, is vital to effective practice, whether that understanding is specifically theory-based or intuitive. An explicit recognition of heuristics and a discussion of issues in genetic counseling within this framework, however, can meaningfully inform both practice and research in genetic counseling.The vagueness of the term “heuristics” is a source of frustration and debate within the field of psychology. Shah and Oppenheimer have argued that the term has come to be applied so broadly as to describe anything and everything (2008). Gigerenzer, one of the most outspoken critics of Kahneman & Tversky’s work, has criticized the original heuristics identified by Kahneman & Tversky as “largely undefined concepts” that “can post hoc be used to explain almost everything. After all, what is similar to what (representativeness), what comes to your mind (availability), and what comes first (anchoring) have long been known to be important principles of mind” (Gigerenzer 1991). Elsewhere, he has complained: “Not every proposed explanation of behavior is a heuristic [. . .] Strategies exist that are not fast and frugal, nor do they involve optimizing; endless committee meetings are one example” (Gigerenzer 2006). According to his own Fast and Frugal Program, each heuristic is a process model, describing a discrete underlying cognitive process (Gigerenzer 2006, Gigerenzer and Todd 2008).The heuristics proposed by Kahneman and Tversky that are described in this paper represent what Gigerenzer has called “a preponderance of vague labels” (Gigerenzer and Todd 2008). Applied to the setting of genetic counseling, however, this arguable lack of clarity does not necessarily mean these concepts are any less useful. Vague though they may be, these labels can provide at least a rough framework for assessing which reasoning processes patients may be using and therefore help to guide effective counseling. Application of heuristics in the clinical settingIn genetic counseling, it is not necessary to predict—and perhaps even important not to attempt to predict, or to believe that one can predict—how a patient will respond in a particular situation. If during the course of the session, however, it becomes evident that a patient is very anxious about a risk that seems relatively low or perhaps unconcerned by a risk that is quite high, it may be helpful for the counselor to consider the influence of heuristics and to broach this possibility with the patient. In certain situations, nondirective counseling may lead patients “to make decisions that are partially informed or poorly reasoned” (White 1997). An alternative approach is to engage in a dialogue with patients to help them reach a decision. (For additional discussion of a dialogical rather than nondirective approach to counseling, see White 1998). Although genetic counseling offers risk information and supports autonomous decision making, several researchers have noted that patients are often not satisfied with probabilistic answers. Rather, many are seeking certainty, or at least direction as to what they should do (Wertz et al. 1986; Shiloh 1996a; Shiloh 2006). “Willingly or not, most genetic counselors become decision-making counselors” (Shiloh 1996a).Of the articles identified in this literature search, nearly all of those mentioning heuristics as a source of cognitive errors advised genetic counselors to beware of heuristics as potential pitfalls in counseling. Only a minority, however, proposed practical suggestions for debiasing techniques in counseling.With a few patients, it might be appropriate to raise the notion of heuristic biases directly, thereby validating the patient’s feeling of a risk as very high while simultaneously providing some balance to the discussion through the suggestion of which cognitive tendencies might be contributing to this feeling. Suggesting an abstract and novel concept is likely to stimulate systematic reasoning processes, which may in turn lay the groundwork for an informed decision.  Moreover, for some patients, an intellectual understanding of their feelings may lead to a decrease in anxiety.With other patients, an overt discussion of heuristics could be overwhelming or perceived as tangential to the patient’s concerns. In these cases, a genetic counselor’s knowledge of heuristics could still inform the genetic counseling process. In counseling a patient who says, “I watched my mom die of cancer . . . I am sure it’s going to happen to me,” it is immaterial whether this feeling is conceived of as an example of representativeness (perhaps the woman identifies with her mother) or whether it is explained as a product of the availability heuristic (the salience of the memory of her mother’s death from cancer). In either case, a response of primary empathy from a genetic counselor might be “You feel like you can’t avoid getting breast cancer. It feels inevitable.” This type of reflective response invites the patient to explore her feelings. (Veach et al. 2003, pp. 52, 56-57) A response of advanced empathy (Veach et al. 2003, p. 52) but informed by heuristics might instead be, “You identify with your mother, and because you were so close, you feel like you are destined to die of breast cancer too.” This would carry an assumption of representativeness. It is still an empathic response, but it is perhaps presumptive in suggesting the possible cognitive processes that may underlie the patient’s feeling of risk. Taking another tack, one that assumes that the availability heuristic is at work, the counselor might instead respond, “The memory of watching your mother die of breast cancer is still so strong that it’s hard to imagine any other possibilities for yourself.”As with either primary or advanced empathy, either of these responses would give the patient the opportunity to correct the counselor if the counselor’s interpretation of the patient’s feelings were incorrect. Unlike many other empathic responses, however, this type of advanced empathy, informed by heuristics, might have the benefit not only of helping the patient to express her feelings but also of potentially mitigating the effects of heuristic processes at work by drawing them out of the patient’s subconscious and into the patient’s consciousness.This example of a potential approach to counseling is purely speculative, an illustration of one way in which a greater awareness of heuristic biases could help counselors to identify when patients might be particularly prone to making decisions that they may not be happy with long term and how this knowledge might inform their dialogue with patients. Psychiatrist and researcher Robert Klitzman has suggested that within the context of genetic testing, the anxiety associated with uncertainty may lead people to make decisions that decrease anxiety at the moment but possibly lead to greater future risks (2010). Identifying these heuristic biases could enable genetic counselors to intervene in a way that steers patients toward systematic reasoning, thereby increasing the likelihood of genuine informed consent and decision making. Helping genetic counselors to recognize and check their own biasesMuch of the practice of genetic counseling necessarily relies on intuition and previous experience. Decisions as to how to respond to patients in a session are made on the spot, and are therefore more likely to be heuristic. In some situations, especially in the case of experts, this intuitive reasoning leads to better judgments (Wilson and Schooler 1991). In other cases, however, it may lead to errors in judgment. In the literature review conducted for this paper, nearly three quarters of the articles presented heuristics as a source of bias leading to cognitive errors, yet less than a quarter of the articles mentioned the susceptibility of practitioners to these same errors in reasoning.Although the term “heuristic” may not often be used, most people are not surprised to hear that someone whose best friend died from breast cancer or had a miscarriage might perceive her own risk to be higher than someone who had not lived through this experience. This is a reflection of the availability heuristic. Attention to the psychosocial aspects of the counseling situation carriers an implicit recognition and appreciation of such heuristics. The idea that a patient’s prior beliefs and sense of risk will affect her perception of the information she receives during the genetic counseling session is, as Gigerenzer has said, not novel. This is an implicit recognition of anchoring.Much of a genetic counselor’s education consists of clinical experience. This same experiential learning that forms the basis for good counseling and informed intuition also leaves genetic counselors open to biases arising from the unique characteristics of cases they have personally seen, which may or may not be statistically representative of actual frequencies, given the law of large numbers. In relying on intuition and learning synthesized from previous counseling experiences, counselors are susceptible to the same biases as patients.Many practitioners are doubtless aware of how their own experience influences their perception of risk, whether or not they use the label “availability.” A study by Roggenbuck et al. (2000) showed that genetic counselors had an increased perception of genetic risk, relative to teachers and nurses. The authors of this study discussed the possibility that this increase in risk perception was due to representativeness and availability. Greater recognition of the source of these biases may or may not lead to a lowered risk perception among genetic counselors, but an increased awareness of the source of their biases and elevated sense of risk may help genetic counselors to check some of these biases when counseling patients.Additional support for the idea that genetic counselors may be able to suppress these heuristic biases is suggested by a small study of 13 participants that combined an experimental decision-making problem with neuroimaging. Participants had an average of 16 years of education. When asked a classic Kahneman and Tversky question used to illustrate the bias of representativeness, the region of the brain associated with conflict was activated. This was true even for participants who chose the incorrect answer, indicating that they had ignored the base rate (presumably because of the heuristic of representativeness). The authors concluded that inappropriate use of heuristics is reflective of a failure to suppress biases rather than an inability to systematically reason through probabilities (De Neys et al. 2008).  Heuristics as a framework for discussion and research in genetic counselingA heuristics perspective has the potential to facilitate the collection of diverse clinical experiences and identify common threads in a way that encourages a more systematic and academic discussion of issues in genetic counseling. While conversations in the field may begin from an anecdotal basis, a heuristics approach represents a recognized and tested framework that is still actively being researched and which could help guide discussions toward a more critical examination of clinical experience. By extension, the use of a heuristics perspective, along with other theories, models and findings from the field of cognitive psychology, can inform research methods and suggest directions for future research. While such a statement may seem obvious, in a 2008 review of the risk literature, Sivell et al. found that only four out of 59 studies used theoretical frameworks developed by psychologists to explore how patients construct their perceptions of risk. The authors recommended “greater exploration of the ways in which risk is constructed, along with the development of stronger theoretical models, to facilitate effective and patient-centered counseling strategies."Studies assessing patients’ perception of the risk for developing cancer routinely query participants as to their family history of cancer. Only a few studies, however, have also inquired about any close friends with cancer. Given the large number of studies supporting the use of the availability heuristic, this additional variable could give new insight into differences in risk perception. The Colored, Eco-Genetic Relationship Map (CEGRM) is a unique example of a tool that has been developed based on social systems theory and which gathers information from patients related not only to family members but also related to friends and social networks (Kenen and Peters 2001).            The emerging field of Cardiogenetics also offers a unique opportunity for heuristics research. In this new area of interdisciplinary medicine, many empirical risks are not yet well established. Patients with long QT syndrome, for instance, are often told that their risk is low, medium or high (T. McDonald, M.D., personal communication, March 9, 2011).  Practitioners find themselves in the position of making important judgment calls with limited data. The possibility of sudden death in someone who previously seemed completely healthy brings up psychological issues that are unique from those encountered in other areas of genetic counseling fields.None of the 61 published articles identified in the literature search for this paper addressed issues in cardiogenetics through a heuristics framework. There is a substantial amount of literature, however, on practitioner biases and cardiac risk assessment (see McKinlay 1996; Arndt et al 2008) as well as on patient perception of cardiac risk (see Dracup et al. 1995; Schwartz and Vaughn 2002; Horowitz et al. 2004; Bunde and Martin 2006; Griffith et al. 2009). The uncertainty and challenges of this field are significant, and heuristics concepts and research could play an important role in establishing best practices.  Part III: Taking a Page from GigerenzerNearly as old as the Heuristics and Biases Program is a debate in the field of psychology over Kahneman and Tversky’s approach to the question of heuristics and its focus on heuristics as a source of cognitive errors. Most notably, Gerd Gigerenzer has criticized the normative assumptions behind Kahneman and Tversky’s work and has instead emphasized the adaptive nature and strengths of heuristics in a complex world where all the variables cannot be clearly defined (Gigerenzer and Goldstein 1994).  Gigerenzer has also pointed out that the first heuristics (such as satisficing) focused not on inferences leading to judgments, but on preferences, where there was no external criteria for success, no correct answer (Gigerenzer 2006). This focus on preferences and appreciation of the complexity of each individual’s situation reflects the values of genetic counseling. Gigerenzer has only been occasionally and briefly cited in the genetic counseling literature (Steginga and Occhipinti 2004; Katapodi et al. 2005; West and Bramwell 2006; Sivell et al. 2008; Dixon and Kohnheim-Kalkstein et al. 2009, p. 76, 80-82), particularly as compared to the attention that Kahneman and Tversky have received. A Biased Framing of Heuristics in the Genetic Counseling Literature            This paper, like much of the literature in the field, has concentrated largely on heuristics as a source of cognitive errors. Of the 61 articles identified in the literature search, a majority discussed heuristic biases leading to cognitive errors, and only a quarter acknowledged the adaptive aspects of heuristics. Three quarters of the articles identified engaged in a discussion of heuristics without also explicitly mentioning systematic reasoning—the alternative to heuristic reasoning which, given that our rationality is bounded, is not always feasible. Addressing heuristic processing without also presenting the alternative of systematic processing, and presenting heuristics only as a source of cognitive errors, rather than also presenting heuristics as necessary adaptive tools, constitutes a framing bias in the way heuristics has been discussed in the genetic counseling literature.            Framing, also studied by Kahneman and Tversky, is ironically perhaps the heuristic most often discussed in the genetic counseling literature, although it has most often been addressed without an explicit mention of “heuristics.” A little over a quarter of the articles identified in this literature search mentioned framing, but these articles represent only a small proportion of the articles discussing framing within the context of risk communication. An example of framing given by O’Doherty and Suthers (2007) is loss versus gain framing. A patient may be told that she has a 60% chance of developing cancer (loss framing) or that she has a 40% chance of not developing cancer (gain framing). The authors recommend framing risk information in a variety of ways, presenting the chance that she will not develop cancer as well as the chance that she will develop cancer (O’Doherty and Suthers 2007). This constitutes a more balanced presentation of information, with the probabilities equaling 1. O’Doherty and Suthers are not alone in suggesting this balanced presentation of risk information (Abramsky and Fletcher 2002; Etchegary and Perrier 2007). A discussion of heuristics only as a source of biases leading to cognitive errors represents a loss frame, or a negative frame. A more balanced discussion of heuristics should also present these same biases in a gain frame, or a positive frame: as part of an adaptive toolbox to deal with the limits of bounded rationality in a large world where perfect knowledge does not exist (Gigerenzer and Brighton 2009). Genetic Counseling in a Large World:  Ecological RationalityGigerenzer has lamented the fact that in medicine more generally, heuristics have come to be so closely associated with bias and error (Gigerenzer and Gaissmaier 2011). “The positive power of heuristics has been replaced by a bad reputation” (Gigerenzer and Todd 2008). As he points out, Simon’s original question leading him to the study of heuristics was ‘How do human beings reason when the conditions for rationality postulated by the model of neoclassical economics are not met?’ (cited in Gigerenzer and Gaissmaier 2011). This question relates to large worlds—complex situations in which not all the variables are known. Gigerenzer has criticized the Heuristics and Biases program as applicable to small worlds but not to large ones (2011), pointing to what he says is an assumption of omniscience, optimization and universality: “Heuristics run counter to these ideals, in that they assume limited knowledge rather than omniscience. Their goal is to find a good solution without the fiction of an optimal one” [emphasis mine] (Gigerenzer 2006).            The world of genetic counseling is decidedly a large world, one in which perfect knowledge is not possible. Even when genetic counseling involves recessive or dominant conditions rather than multifactorial traits or sporadic events, there are always many variables that cannot be predicted. Penetrance is one, variable expressivity is another, but the realities of a patient’s life situation, present and future, are even more complex and even less predictable. The rationality of a decision can only be understood in the context of its environment. Spending \$3 on a quart rather than a half gallon of milk may make sense after all. This is what Gigerenzer refers to as ecological rationality. Applying Gigerenzer’s approach to genetic counselingIn Gigerenzer’s rejection of normative assumptions in decision making and his program of ecological rationality, his approach to heuristics is consistent with the modern philosophy of genetic counseling. A heuristics approach to genetic counseling could benefit from Gigerenzer’s line of research for two major reasons.First of all, Gigerenzer’s insistence on the specificity of process models points to greater potential for developing useful debiasing techniques in the practice of genetic counseling. Other researchers within genetic counseling have already pointed out that an accurate characterization of heuristics could suggest debiasing techniques to help patients make more informed decisions (Chapman and Elstein 2003; Katapodi et al. 2005). In those instances where it may be appropriate to encourage patients to engage in systematic reasoning to ensure informed consent, the identification of discrete processing habits represents a potentially more targeted basis for intervention in a genetic counseling session than those heuristics proposed by Kahneman and Tversky which have been discussed in this article. Second of all, Gigerenzer’s perspective represents an approach to risk communication and counseling that begins with the assumption of heuristics as intelligent and adaptive and includes an appreciation of ecological rationality. Gigerenzer has consistently argued that in large worlds, where variables and consequences cannot be known, heuristic reasoning can often lead to better decisions than systematic reasoning might. Gigerenzer’s program of ecological rationality studies heuristics as discrete process models but also investigates which environmental structures will allow these heuristic process models to be successful (Todd and Gigerenzer 2007):Kahneman and Tversky centered on norms and were anxious to prove that judgment often deviates from those norms. I am concerned with understanding the processes and do not believe that counting studies in which people do or do not conform to norms leads to much. If one knows the process, one can design any number of studies wherein people will or will not do well.  (Gigerenzer 1996) In response to Kahneman and Tversky’s experiments, Gigerenzer has conducted experiments with questions presented in alternate ways with the result of fewer errors (Gigerenzer 1991; Gigerenzer and Hoffrage 1995; Gigerenzer and Hertwick 1999). Gigerenzer has written extensively on communication between doctors and patients (Gigerenzer et al 2007) and also called for more transparent communication of risk through public health messages (Gigerenzer et al 2009; Wegwarth and Gigerenzer 2010). Gigerenzer’s emphasis on the ecological rationality of heuristics represents the possibility of going beyond a negative frame of heuristics and lays a scientific and theoretical basis for respecting the patient’s point of view. Rather than merely intervening to try to stave off the use of heuristics, efforts could be focused on designing environments and counseling practices that will promote the most intelligent use of adaptive heuristics. The ecological rationality of the patient’s perspectiveIn early risk literature, patients were considered to be “confused” if they took a line of action, often a reproductive decision, inconsistent with the researcher’s perception of what the rational course should be (Pilarski 2009). At the time of its publication, Lippman-Hand and Fraser’s work was unique both for its qualitative methodology and research question: Most of those studying genetic counseling have focused on what reproductive choices are made by parents at different risks for having abnormal children. In contrast, we were interested in how these decisions were made. (Lippman-Hand and Fraser, 1979d). Lippman-Hand and Fraser’s approach to genetic counseling marked a shift from an objective to a subjective characterization of risk within the literature, reflecting an evolution in the goals of genetic counseling (Palmer and Sainfort 2003). Today, the legitimacy of a subjective interpretation of risk is well established (Pilarski 2009) and the goal of risk communication is to ensure that patients understand the objective risk component, often measureable to some degree, regardless of their personal perception of the subjective risk component. Palmer and Sainfort have pointed out that the notion of risk is often conflated with uncertainty but is better understood as uncertainty + adversity, where uncertainty refers to the objective risk component and adversity recognizes the patient’s own interpretation of that event (1993). Defending the binary, outcomist view of risk            One patient from Lippman-Hand and Fraser’s study is often quoted regarding her perception of her risk, which was focused on the numerator: “In fact, the risk is always too high. There is this one chance we’ll always have” (Lippman-Hand and Fraser 1979a).The authors commented on this risk perception, common among participants:By concentrating on the ‘one,’ the undesirable outcome, she and others converted their risk of having an affected child into binary form, either it will or will not happen. This is what matters to them, and it had an important influence on their subsequent reproductive decision-making [Lippman-Hand and Fraser, in press]. (Lippamn-Hand and Fraser 1979a) Shiloh (1994, 1996) has remarked upon patients’ binary, or outcomist perspective in several articles: “For genetic counselees, the knowledge that a defect can happen is important, not the probability that it will happen” (Shiloh 1994).Numerous studies in risk communication have reported on the difficulties that patients have interpreting risks and probabilities. Hard-line frequentists, however, argue that because a probability is defined over a specific reference class, a single event cannot have a probability (Cosmides and Tooby 1996). For the patient, the notion of a 1/10 or 1/500 chance carries a certain aspect of absurdity when applied to a single event. Within the genetic counseling literature, West and Bramwell have also supported this idea: “Overall it is highly predictable what proportion [of women] will have affected babies; however, for each woman in isolation it is largely unpredictable whether her baby will have Down Syndrome or not” (2010).From a frequentist perspective, it is not reasonable for an individual patient to evaluate her own risk for a single event based on a probability that applies to a series of events. The outcomist view, on the other hand, may be highly adaptive. Shiloh has appreciated the possible ecological rationality of this outlook:Once parents formulated their problem in binary fashion and focused on outcomes, they began to consider the implications of being at risk and the potential impact of what might or might not occur—hence, moving forward in their decision-making process. This approach, which is a clear violation of normative decision rules, may not be irrational in their case." (Shiloh 1996a)Interestingly, viewing patients’ risks in terms of frequencies and empirical probabilities does make sense from the perspective of the practitioner, for whom the patient’s outcome is merely one event in a longer series. The patient is likely represented in the denominator and possibly represented in the numerator. The cases that an individual counselor sees may represent only a series of a modest length, but the cases that counselors see collectively should, in principle, finally conform to the law of large numbers. From the counselor’s perspective, therefore, frequencies do approximate a prediction of reality. For the patient, however, the law of small numbers applies. Yes, the outcome will be determined by chance, likely weighted more toward one outcome than another, but the coin will only be flipped once, twice, maybe a few times. Predictions cannot be made; uncertainty can only be quantified (West and Bramwell 2010).In genetic counseling, a probability or a frequency is usually the best and the only estimate of risk that can be provided to patients. Knowledge of or direct experience with what may sometimes seem to be patients’ inability to understand probabilities, however, should be balanced by an appreciation of the arguable absurdity of trying to apply a frequency to one’s own individual situation. Conclusions              Many varying theoretical frameworks and models have been applied to genetic counseling, all of which are necessarily limited. None will adequately explain any individual patient’s unique situation, or decision-making process. Genetic counseling will always rely on counselors’ intuitive judgments and experiential knowledge. Nevertheless, a more systematic application of theoretical frameworks and known phenomena from cognitive psychology represents an opportunity to improve standards of practice while contributing to an evidentiary basis for the utility of genetic counseling.              As the very large world of health care is becoming increasingly complex, the role of the genetic counselor continues to be refined and perhaps redefined. Incorporating a stronger psychological basis into genetic counseling research and practice could lead to greater recognition of this interdisciplinary field by other disciplines and lay the groundwork for more collaborative research. The possibility of heuristics to inform genetic counseling should be looked at alongside the numerous other theories and models that have been proposed, with the goal of moving toward a practice of genetic counseling that is increasingly evidence-based.  The literature search conducted for this paper was limited to those articles which explicitly discuss heuristics and represents an attempt to bring together the work that has been done so far in this area. The index of articles (Appendix C) was designed in the hope that it will serve as a starting point and a resource for other researchers, counselors and students investigating this topic. This compilation of articles does not amount to a definitive review of the literature, and a closer evaluation of these articles could represent an interesting direction for research that is beyond the scope of this paper.              Numerous other articles within the genetic counseling literature arguably discuss heuristics implicitly, without employing the label. Many articles have addressed how important it may be for a patient to have some sense of control. Some of these articles discuss this concept within an alternative theoretical framework. Berkenstadt et al., for instance, articulates the concept of Perceived Personal Control (PCC) and propose it as an outcome measure for genetic counseling (1999). A similar concept, “illusion of control” has also been considered as a heuristic. A review of the genetic counseling literature that implicitly discusses heuristics and one that synthesizes findings from various models would be a useful step in identifying new directions for research.              Research goals should be first and foremost focused on improving clinical practice and patient care, with an open mind to the possibility that some findings may be counterintuitive. Any research that sets out specifically to prove the utility of genetic counseling will be doing the field few favors. Devoted fans setting out to prove a pet hypothesis is not firm grounding for the establishment of an evidentiary basis. Rather than directing efforts toward finding proof for a firmly held belief in the value of genetic counseling, research should be used to help guide the evolution of the field and increase the utility of genetic counseling relative to whatever utility it may have at present. Only through a willingness to engage in critical analysis will it be possible to improve standard of care, creating the environments that are most conducive to informed decision making in order to provide genetic counseling services that are of even greater value to patients.
Comments