The View from Nowhere through a Distorted Lens:
The
Evolution of Cognitive Biases Favoring Belief in Free Will
By Kip Werking
[pdf version; htm version; doc version]
“In the distant future I see open fields for far more
important researches. Psychology will be based on a new foundation, that of the
necessary acquirement of each mental power and capacity by gradation. Light
will be thrown on the origin of man and his history.”
—Charles Darwin (1859: 449)
“One must view a wicked man, like a sickly one—We
cannot help loathing a diseased offensive object, so we view wickedness.—it would
however be more proper to pity than to hate & be disgusted.”
—Charles Darwin (1987: 608)
I. Introduction
Although philosophers have confronted the evidence for human irrationality in other contexts (Mele 1987; Stein 1996), there is one area in which this evidence may be of great relevance but has yet to be discussed—the free will problem. Instead, a vast ocean seems to separate the respective literatures on free will and human irrationality. One might say that there is a lacuna in the literature about the lacunas in our minds. The result is that philosophers who study free will work in a context which tends to presuppose that humans are rational.
The silence about human irrationality is related to another silence about the demanding concepts many have for free will. It is telling that compatibilists about determinism and free will defend concepts of that power which nobody would deny humans can satisfy. Their arguments against libertarian views are often limited to show that such concepts are inconsistent with actual human agency and not that the stronger notion of libertarian freedom does not govern the relevant terms in the debate. As against the non-realist, this option is unavailable to the compatibilist and so compatibilists often give non-realist views on free will little attention (Strawson 1994) or grant that they are defending something different than the traditional notion of free will (Dennett 2003). Indeed, one prominent compatibilist has characterized the stronger notion of free will as “a sort of metaphysical megalomania” (Fischer 2006b). Compatibilists are just now beginning to research the issue of whether their conception of free will is too weak and the initial results are both mixed and intriguing (Nahmias et al. 2004, forthcoming; Nichols 2004, forthcoming). Given the presumption of human rationality, compatibilists may assume that more demanding concepts of free will are irrational and so pass over them in silence.
In this article, I attempt to show, by bridging the ocean between these two literatures and rebutting the presumption of human rationality, that libertarian and non-realist concepts of free will are not so unreasonable. First, this introduction (I) will distinguish between two kinds of control and show how this distinction can frame the free will problem. The introduction will also describe the ways in which cognitive biases may evolve. In part two (II), I analyze the psychology of human decision making. In particular, I focus upon certain critical decisions and argue that a rational agent would look upon these options as one would look upon Thomas Nagel’s View From Nowhere. But certain cognitive biases allow people to view this View From Nowhere through a distorted lens—and therefore to avoid its paralyzing sting. I argue that these biases evolved because they are adaptive in various ways. In part three (III), I will show how certain cognitive biases lead people to feel they have more control and responsibility, with respect to their decisions, than they actually do. I speculate that these biases evolved because they falsely signal certain attractive traits to others. In part four (IV), I will show how one profound cognitive bias, the just world phenomenon, and several others related to it, would further incline one to believe in free will. I speculate that these biases may have evolved because they were adaptive or evolved despite being non-adaptive. I conclude, in part five (V), that a multitude of cognitive biases may infect human thinking about free will. The vulnerabilities in the human psyche suggest that libertarian and non-realist concepts of free will may capture, more than their compatibilist alternatives, the usage which governs the meaning of terms in the debate about free will.
First, it is important to distinguish, in the ancient controversy between those who defend the existence of free will and those who assail it, between two kinds of control.[i] The first kind of control is actual control. This is the sort of control human beings have over their lives. It is such that, given one’s ultimate ends and purposes, one can make choices according to those ends and purposes and act accordingly.[ii]
But there is another kind of control. To understand this second type of control, first consider what I will call novelist control. This is the sort of control that novelists have over their characters. According to novelist control, an author may choose not just how a character makes decisions in accordance with certain ultimate ends and purposes, but also choose what these ends and purposes will be. The ultimate ends and purposes are themselves chosen—not given. In this way, the scope of novelist control is much wider than that of actual control. A novelist has much more control over a character’s life trajectory than the character does. For this reason, some accuse actual control of being shallow (Smilansky 2003).
There is a superficial sense in which novelist control is buck-stopping (Dennett 2003: 99). In the ordinary case, to tell a story about some agent’s life, one would need to say something about how that agent’s ultimate ends and purposes arose. For example, one might want to tell a story about how the agent evolved, or inherited certain genetics, or lived in a certain childhood environment. Even this may not be entirely satisfying because one might always inquire: “why that evolutionary path, why those genetics, why that childhood environment?” At some point, simple logic shows that these questions must go unanswered. There is a temptation, however, to feel that, in the novelist scenario, the novelist provides a robust and sufficient explanation for agent’s life. The novelist is responsible for the meticulous creation of every detail in the character’s life. Surely this author can provide a sufficient explanation for why the agent’s life story followed the path that it did. One might ask why the character’s life unfolded as it did and the author could say “I wanted to create a character with property X and who lives a life with features Y and Z.” But the question remains: “why did you desire to create a character with that property and who lives a life with those features?” Again, such questions must eventually go unanswered.
When one considers novelist control in the context of this ancient controversy, however, one encounters a tantalizing possibility—to what extent might agents have novelist control over their own lives? I will call this sort of control novelist* control. If people have novelist* control over their own lives, then they make decisions not just in accordance with their ultimate ends and purposes, but they furthermore determine what these ultimate ends and purposes will be. Actual novelists already approach this level of control over their characters. To the extent that their stories lack in the richness of detail present in reality, one can imagine how God or an advanced being (perhaps a radically enhanced human being in a future civilization) might design the lives of actual humans and not just characters in novels. Greene and Cohen consider just such a scenario:
“It is very simple, really. I designed him. I carefully selected every gene in his body and carefully scripted every significant event in his life so that he would become precisely what he is today. I selected his mother knowing that she would let him cry for hours and hours before picking him up. I carefully selected each of his relatives, teachers, friends, enemies, etc. and told them exactly what to say to him and how to treat him. Things generally went as planned, but not always. For example, the angry letters written to his dead father were not supposed to appear until he was fourteen, but by the end of his thirteenth year he had already written four of them. In retrospect I think this was because of a handful of substitutions I made to his eighth chromosome. At any rate, my plans for him succeeded, as they have for 95% of the people I’ve designed. I assure you that the accused deserves none of the credit.” (Greene and Cohen 2004)
There is one crucial point to understand about this second kind of control. The point is that the apparent similarity between novelist control and novelist* control is deceptive. Of course, novelist* control is logically impossible. The act of choosing one’s character presupposes the existence of a character according to which one can make these choices. Character cannot create itself in a vacuum. But novelist control, unlike novelist control, is possible.
So the apparent similarity between novelist control and novelist* control is deceptive. To the extent that novelist control and novelist* control appear similar, however, one can understand how people might make this mistake. It is not obvious, for example, that one might have control over the ultimate ends and purposes of another but no such control over one’s own ultimate ends and purposes. As the rest of this article will explain, certain cognitive biases might enable this illusion of feeling that one has more than actual control over one’s lives—and that one approaches having novelist* control.
This distinction between actual control and novelist control helps frame the ancient dispute between those who defend the existence of free will and those who assail it. First, one might characterize the dispute as one just about the powers humans have. On this view, both parties agree about the relevant concept of free will and then examine human nature to see whether human being can, in fact, satisfy this concept. Indeed, the distinction between actual and novelist* control suggests that all parties might agree that the relevant notion of control being tested is quite strong—not unlike novelist* control. One happy consequence of this characterization is that it is conducive to empirical investigation. Scientists can examine human nature and see whether human beings have such powers. The dispute between libertarians and non-realists has this character (Pereboom 2001: 69-88; Nichols, forthcoming).
Second, one might characterize the dispute between these two parties as one about what the relevant concept of free will is. So according to this characterization, both parties would hold their understanding of human nature constant and then consider whether this notion of human nature satisfies the relevant concept of free will. The distinction between actual and novelist* control suggests that all parties agree that the account of human agency is relatively weak—not unlike actual control. One must then determine whether this account of human agency can satisfy the relevant concept of free will. Unfortunately, this view is much less conducive to empirical investigation. It is more difficult to study the exact meaning, if any, of the terms “free will” and “moral responsibility” than it is to study whether indeterministic processes in the brain enhance the control humans have over their decisions.[iii] The dispute between compatibilists and non-realists has this character.
The dispute, according to first characterization, is easier to solve than the dispute, according to the second characterization, because human beings are just more available to inspection than the terms “free will” and “moral responsibility” are. Similarly, the human genome is more exact than the concept of “love” and the Mona Lisa is more definite than the concept of God. That those who defend the existence of free will may engage in both tacit and overt acts of revisionism only exacerbates this difficulty. Whether indeterministic processes in the brain enhance the control we have over our decisions will never change; on at least some views, however, the terms of “free will” and “moral responsibility” might have meant one thing yesterday but should mean something different today (Vargas 2005).
Perhaps despairing at the prospects of winning the first battle, the majority of philosophers who defend the existence of free will have staked their claims in the murkier territory of the second battle. Considering the difficulty of resolving the dispute, according to this second characterization, the prospect of progress in the near future is grim. In contrast, I suggest a third way of approaching the dispute which promises imminent progress. Whereas the first and second characterizations of this dispute contrast the powers people have with the powers required by free will or moral responsibility, this third strategy contrasts both the powers and responsibilities that people have with the powers and responsibilities people think they have.
The answer to this third question may be relevant to the ancient dispute for two reasons. First, whereas libertarians can deny that people lack libertarian free will, they might not be able to deny, in the face of contrary evidence from the cognitive sciences, that people think they have more power than they actually do—and this illusory power bears a resemblance to libertarian free will. Second, whereas compatibilists can deny that the terms “free will” and “moral responsibility” refer to concepts demanding more than actual control, they might not be able to deny, in the face of contrary evidence from the cognitive sciences, that ordinary responsibility practices presuppose an inflated sense of control and responsibility. In particular, this third strategy avoids any danger of revisionism. Clever philosophers in ivory towers might settle for a weaker notion of freedom and responsibility than the ordinary use of these terms (if any) would demand; the folk or non-specialists in the labs of cognitive scientists do not have this luxury.
In this way, progress according to this third strategy might help to rebut the presumption of human rationality. Allegedly rational belief in free will has been the traditional default; the burden of proof was on those who would reject this orthodox view of human agency. For example, some suggest that the failure of the non-realist view to satisfy the standards of conservatism (whereby folk intuitions survive philosophical inspection) explains its unpopularity:
“But this view nowadays has few unabashed subscribers, and its proponents through the history of philosophy have been relatively few in number. Indeed, it is tempting to suppose that hard determinism is the exception that proves the rule, since a likely diagnoses of its unpopularity is its failing standards of conservativism; as Vargas (2005: 401) remarks, approaches commending substantial revision of folk morality have typically faced a ‘broad and ready skepticism.’” (Dorris et al., forthcoming)
Indeed, without a decisive argument or experiment, those who defend the existence of free will and moral responsibility might have difficulty even understanding what motivates the skeptic. If the defender of non-realism about free will can demonstrate, however, that people engage in widespread and systematic irrationality when making attributions of moral responsibility and estimations of the control they and others have over their lives, then those who defend orthodoxy can understand what motivates the non-realist’s view. The non-realist has just identified certain cognitive biases and revised certain beliefs in accordance with this new information. This would help rebut the presumption of human rationality and the burden might shift to the defender of orthodoxy to show how such seemingly irrational beliefs can survive the discovery of these biases.
There are two ways to proceed along this front. First, philosophers might investigate folk beliefs in the areas that interest them. Limited funds and scientific training may result in polls that are lacking in rigor and sample size. This seems to be the strategy of the experimental philosophy movement. That movement has its virtues but it is too small and too new to promise much progress in the near future.
An alternative strategy is to analyze data which cognitive scientists have already collected. One might suspect that the different motivations of the philosopher and the cognitive scientist would cause any such data to be of limited relevance. This article will argue, however, that the field of cognitive biases and heuristics provides an abundance of data which can help settle the dispute between those who defend the existence of free will and those who do not.
In exploring the dispute according
to the third characterization, I have followed this second strategy of
investigating the relevant cognitive science literature. To my surprise, I discovered at least fifteen
cognitive biases favoring belief in free will and none favoring
disbelief in free will. As one prominent
scholar in this area has noted, “[n]umerous psychological studies on blame and
responsibility, as well as the perplexing outcomes of high-profile criminal and
civil trials, demonstrate that everyday blamers are capable of violating
virtually every rational prescription that moral philosophers, legal scholars,
and rational decision theorists hold dear” (Alicke, forthcoming).
Although just showing how certain cognitive biases relate to the free will problem would be sufficient to create a noteworthy paper, this article will go further and explain how these biases might have evolved. These remarks will, like those found in most evolutionary psychology, be necessarily speculative. Indeed, this approach puts the conclusions of this article in double jeopardy because both the heuristics and biases literature (Gigerenzer 1996) and the evolutionary psychology literature (Rose & Rose 2000; Buller 2005) have their critics—including each other (Haselton & Buss 2003). Despite the lingering controversy over these subjects, however, this article will suppose that both literatures have their several virtues. Furthermore, the remarks on the evolutionary origins of these cognitive biases will provide a context for the philosophical content in this article and further the use of science to solve philosophical problems. Of course, not all of these speculations will be accurate. Furthermore, not all cognitive biases will invite just one evolutionary explanation; any one bias might suggest a variety of explanations and these may be either adaptive or non-adaptive. Nevertheless, exploring the evolutionary origins of belief in free will may strengthen the article’s larger argument.
Although there are several possible explanations for cognitive biases, this article will focus upon evolutionary explanations. Not all possible explanations are evolutionary: cognitive biases may represent, for example, systematic performance errors (Stein 1996: 8-9) or the failure of participants in an experiment to properly construe the given task (Gigerenzer 1996). Furthermore, not all evolutionary explanations are adaptive—natural selection is not omnipotent. In explaining any given trait, one must remember the importance of concepts such as pleiotropy (Williams 1957), exaptations (Gould & Lewontin 1979), and genetic drift (Wright 1932).
But
many evolutionary explanations are adaptive.
Cognitive biases may represent vestigal traits that, although they were
adaptive in the earlier times, are no longer so (Haselton & Buss
2003). Furthermore, traits may be
adaptive in a variety of ways. First,
they may contribute to one or both of the individual’s longevity (natural
selection in the classic sense;
The cognitive biases and heuristics approach and EMT may not necessarily conflict. One can understand this potential harmony in the context of Buss and Haselton’s notion of a smoke alarm as an example of error management logic (2003: 31). Smoke alarms are intended to be hypersensitive because the dangers of a false negative are so much greater than the dangers of a false positive. The high ratio of false positives to false negatives is an intended feature of their design; by analogy, the human terror of harmless snakes may be a feature of our design. But fire alarms engage in this error management because they have limited computational resources. A fire alarm with a supercomputer and sophisticated spectrometer could afford to make less false positives by also avoiding false negatives—even if the ratio of safe errors to dangerous errors remains biased towards safe errors. In this way, the cognitive biases and heuristics approach and EMT may complement each other.
To
illustrate the EMT approach, consider what is, to my knowledge, the only
evolutionary explanation ever proposed for belief in free will: “The Illusion
of Free Will Evolves” by Tamler Sommers (forthcoming). Sommers’ approach is also unique to the
extent that it proposes that cognitive enhancement aggravated, rather than
ameliorated, the error management solution to adaptive problems. He describes how “cognitively sophisticated”
(CS) creatures might begin to doubt the rationality of their reactive
attitudes. For example, a CS creature
might doubt that enacting retribution against a rival is worth the risk of harm
to one’s self. Sommers’ insight is the
suggestion that humans evolved belief in free will and retribution to compensate
for the pacifying affect of this cognitive enhancement.
Like Sommers, I suggest a dual process explanation for belief in free will. Nichols has suggested that a dual process theory explains intuitions about free will (forthcoming) and Greene has defended a similar dual process explanation for deontology (forthcoming). Such theories are consistent with the seminal work by Haidt showing that people make quick judgments about morality and only later justify these judgments through a process of rationalization (2001). Dual process theories all suggest that our evolved psychologies contain a number of domain specific circuits. For example, Steven Pinker argues that:
“[T]he mind is
organized into cognitive systems specialized for reasoning about object, space,
numbers, living things, and other minds; that we are equipped with emotions
triggered by other people (sympathy, guilt, anger, gratitude) and by the
physical world (fear, disgust, awe); that we have different ways for thinking
and feeling about people in different kinds of relationships to us (parents,
siblings, other kin, friends, spouses, lovers, allies, rivals, enemies); and
several peripheral drivers for communicating with others (language, gesture,
facial expression).”[iv]
Such domain specific neural circuitry may provide fast and cheap calculation whereas more general processing in the cortex may provide slower and more expensive, but more thorough and accurate, computation. These circuits may account for both the reactive attitudes people feel towards others as well as the beliefs people maintain about human agency. Because they are faster and cheaper, but less accurate, than more general processing in the cortex, the two processes may come into conflict—reflection might make the illusion of free will disappear. This is consistent with the remarks of Watson (1986), Strawson (1986), and Nagel (1986) that, although reflection can incline one to doubt the existence of free will, this skepticism is difficult to maintain. As soon as one exits calm reflection, the faster and cheaper, but inaccurate, domain specific circuits begin to dominate again.
Unlike
Sommers’ view, however, this article will not argue for any single adaptation
explaining belief in free will. Instead,
this article will provide an indirect argument by considering a
multitude of cognitive biases which may combine to explain, in part, belief in
free will.
This introduction has noted a silence about human irrationality in the literature on free will. To address this silence, it first distinguished between two kinds of control and showed how these kinds of control frame the free will debate. Second, it suggested that the abundant literature on cognitive biases and heuristics will help the debate over free will progress. Finally, the introduction discussed two evolutionary explanations for cognitive biases (as heuristics and as error management) and suggested its own overarching explanation for belief in free will. In section two (II), the article will begin by considering several cognitive biases which may favor belief in free will in light of one philosopher’s notable work on the subject.
II. The View From
Nowhere Through a Distorted Lens
Less than two decades ago, Thomas Nagel (1986) and Galen Strawson (1986) suggested not just that free will does not exist but that it is also impossible.[v] In particular, Nagel distinguished between the subjective viewpoint, which houses our precious beliefs, and the objective viewpoint, which threatens to destroy them. The subjective viewpoint presents metaphysical alternatives before agents without inquiring into the forces that shape their characters. It presupposes, and demands, that people are the ultimate creators and sustainers of their own ends and purposes. But reflection shows that nothing can satisfy this conception of human agency.
Nagel writes that the objective view “takes in not only the circumstances of action as they present themselves to the agent, but also the conditions and influences lying behind the action, including the complete nature of the agent himself.” So, in Nagel’s view, the objective view is problematic because it presupposes that the agent has a given character and that this character constrains its future choices. But Nagel also describes the objective view as presupposing no such character. In this sense, the objective view requires one, in order to be free, “to act from a standpoint completely outside ourselves, choosing everything about ourselves, including all our principles of choice—creating ourselves from nothing, so to speak.” This description suggests that the agent, in uneasy tension between its presently blank and future selves, must look out over the space of possible characters and choose, ex nihilo, what kind of agent to be. This last image deserves Nagel’s lovely title: the “View from Nowhere.”
The View from Nowhere has a paralyzing sting. As Nietzsche wrote, one just cannot “pull oneself up into existence by the hair, out of the swamps of nothingness” (1886: §21). By analogy, although one might have novelist control over another person, one cannot have novelist* control over one’s self. Yet people regard themselves as free and responsible. Furthermore, they seem to accomplish this by embracing an illusory feeling of control, and not by weakening their demands for such control. In the same paragraph, Nietzsche noted that “[t]he desire for ‘freedom of the will’ in the superlative metaphysical sense, [] still holds sway, unfortunately, in the minds of the half-educated.” This article will further describe the abundant evidence showing that people regard themselves as possessing more freedom and responsibility, with respect to their lives, than they do in fact have. Somehow, people jump out of the View from Nowhere.
This article’s central thesis is that people accomplish this feat by looking at the View from Nowhere through a distorted lens. In particular, certain cognitive biases warp the View from Nowhere such that it presents agents with reasons for preferring, from amongst the space of possible characters, one particular character—their later selves. These biases are the endowment effect and the outcome effect. Similar biases, such as contamination and the mere exposure effect, further this phenomenon. Moreover, the fundamental attribution error, working in concert with anthropomorphic bias and asymmetric attributions of blame, provides a ready vehicle in which the endowment effect and outcome bias may infect human thinking about freedom and responsibility. These biases render The View from Nowhere into the View from Somewhere.
First, to understand how the endowment effect might allow people to jump out of the View from Nowhere, one must appreciate that the free will problem is about gifts. In particular, the free will problem involves the difficulty of embracing the gift of one’s character. For example, one prominent libertarian bases his position upon the existence of indeterministic “Self-Forming Actions” (Kane 1996). Similarly, one prominent defender of non-realism about free will has observed that:
“Compatibilists
claim that this is the right thing to say. They believe that to have free will,
to be a free agent, to be free in choice and action, is simply to be free from
constraints of certain sorts. Freedom is a matter of not being physically or
psychologically forced or compelled to do what one does. Your character,
personality, preferences, and general motivational set may be entirely
determined by events for which you are in no way responsible (by your genetic
inheritance, upbringing, subsequent experience, and so on). But you do not have
to be in control of any of these things in order to have compatibilist freedom.
They do not constrain or compel you, because compatibilist freedom is just a
matter of being able to choose and act in the way one prefers or thinks best
given how one is.” (Strawson 1998, 2004)
This character is constraining in the sense that, although it might later modify itself, it will only do so in accordance with its original makeup. In Saul Smilansky’s words, free will does not exist because people are just the “unfolding of the given.” We are not the unfolding of the chosen. The quality of human nature which remains disturbing is this given-ness. This given-ness threatens to alienate people from themselves because one might always question whether one is as one ought to be. Yet people somehow jump out of the View from Nowhere and embrace the gifts of their characters.
One subtle way in which people might accomplish this feat involves the endowment effect. This effect was discovered by Kahneman and Tversky (1979) and named by Thaler (1980). According to the endowment effect, ownership confers value upon goods and services. Although some have questioned the existence of the effect (Shogren et al. 1994; Plott & Zeiler 2005), a large number of experiments have confirmed it. Recently, Nayakankuppam and Mishra repeated the traditional experiment by situating eighty-eight students next to a coffee mug, telling them that they were either buyers or sellers, and then asking them for their price (2005). Nayakankuppam and Mishra reported that “The basic endowment effect emerged with sellers quoting significantly higher reservation prices ($5.59) than buyers, ($4.43), F(1, 82) = 5.64, p < .02.”
This endowment effect is similar to, and perhaps a subset of, the mere exposure effect. Psychologists first noticed this effect in the nineteenth century (Fechner 1876); in modern psychology, the most prominent psychologist to research it is Zajonc (1968). According to the mere exposure effect, just the fact that one is familiar with something will increase one’s liking it. A recent meta-analysis of research on the mere exposure effect concluded that “[t]he first 20 years of research on Zajonc's (1968) mere exposure effect leaves little doubt that the exposure–affect relationship is a robust, reliable phenomenon” (Bornstein 1989).
Several
scholars have speculated on the evolutionary origins of the endowment
effect. David Friedman suggests that an
endowment effect reflects territoriality and notes that such territoriality has
been observed in many species.[vi] Friedman argues that such territoriality was
adaptive in earlier times, even if it is no longer adaptive, because it
motivated humans to “fight very hard” for their possessions. Others have observed the precise endowment
effect in capuchin monkeys (Chen et al. 2006).
This led these researchers to conclude, as Friedman did, that the bias
is innate and not learned. From the
perspective of economics, Steffen Huck, Georg Kirchsteiger and Jörg Oechssler
have provided a precise mathematical model for how an endowment effect evolved
because it “improved one’s bargaining position in bilateral trades” (2005).
Similarly, Herbert Gintis has developed a game theoretic model to explain how
the endowment effect enabled the natural evolution of private property (Gintis,
forthcoming). In the context of the gift
of one’s character, including one’s genetic endowment, one must also note that
natural selection could not favor any tendency to doubt or disown this
character. Genes which make one
irrationally prefer them will thrive at the expense of genes which allow one to
approach them with skepticism.
Scholars
have also speculated on the origins of the mere exposure effect. Bornstein notes that it is adaptive “for
adults to prefer the familiar over the novel” because, even if there are
advantages to exploring new territory, “some risk is inherent in any venture
into the unknown” (1989). In contrast,
it is adaptive for infants to prefer the novel over the familiar because they
need to grow familiar with their environment and can rely upon their mothers to
protect them. Bornstein further notes
that his suggested explanation is consistent with the findings that delay
enhances, and boredom mitigates, the mere exposure effect. He proposes, therefore, that both the adult
preference for familiarity and the infant preference for novelty have “evolved
into the natural repertoire of human behavior.”
The endowment and mere exposure effects provide subtle but compelling explanations for how people jump out of the View from Nowhere. The View from Nowhere derives its sting from the vacuum of values existing before the agent’s birth. From the agent’s vantage point, it cannot yet choose which character to have because any such choice presupposes an existing character. So the agent remains paralyzed before this vista of character space because all possible characters are equally desired and equally abhorred. But the endowment effect and mere familiarity affect give people a reason for preferring one of these characters from amongst the others—their later selves. According to these effects, people will prefer what they own and that with which they are familiar. To the extent that people forget that, according to the View from Nowhere, their later selves can have no rational influence upon their choice, they will prefer the selves they have been given. Other cognitive biases show that people will tend to forget just this sort of fact.
One cognitive bias which would help people forget this sort of fact is outcome bias. According to this cognitive bias, people consider the eventual outcome of a decision as a relevant when evaluating the quality of the decision itself. In five different experiments, Baron and Hershey demonstrated the existence of such an outcome bias (1988). In later articles, they are quick to emphasize that outcomes can affect decisions in normative ways and that their original research met stringent requirements for demonstrating the nonnormative use of outcomes (Hawkins & Hastie 1990; Hershey & Baron 1992). Such ambiguities were present, for example, in research on the similar phenomenon of hindsight bias, according to which individuals counted good outcomes as one of the criteria for evaluating the decisions made by other people (Zakay 1984). Outcome bias and hindsight bias are perhaps subsets of the more general phenomenon of contaminating effects according to which almost any known information tends to influence decision making (Chapman & Johnson 2002).
I am not aware of any research on the evolution of such contaminating effects. As Baron and Hershey note, outcomes are only irrelevant in certain peculiar situations which are probably novel to our evolved psychology. This suggests that Kahneman and Tversky’s explanation of cognitive biases as computationally and metabolically efficient alternatives to more accurate decision processes may explain such contamination effects. Stanovich and West report a negative correlation between outcome bias expression and a cognitive ability composite and conclude that the bias reveals the computational limits of the human brain (1998). This finding lends further support to Kahneman and Tversky’s explanation. Finally, if one considers Buss and Haselton’s three criteria for EMT, then that theory does not seem to explain outcome bias. In contrast to situations involving uncertainty which are repeated throughout history (is that a snake in the grass?), the outcome bias experiments posit novel situations with perfect knowledge.
The contamination effects, including outcome bias, suggest an elegant solution to the problem of how people can forget that agents overlooking the View from Nowhere do not yet have any reason to prefer their later characters over any other. People have a natural tendency, in accordance with these biases, to take into account almost any known information—including the information of what characters they were eventually given and with which they have grown familiar. Given these contamination effects, the endowment effect and mere exposure effect become available to give people a subtle reason for preferring their later selves from any other possible character. In this way, people may jump out of the View from Nowhere and into the View from Somewhere. Certain other biases, however, provide further means by which people can escape the View from Nowhere’s paralyzing sting.
One
cognitive bias is so important to social psychology that researchers have
labeled it the fundamental attribution error (FAE) (Ross 1977). According to the FAE, people overemphasize
dispositional influences on observed behavior and underemphasize situational
influences on this same behavior. In the
classic experiment performed by Jones and Harris, observers listened to a
person read an essay that either supported or opposed
Without making a single reference to evolution, Gilbert and Malone speculate that the FAE exists because it is “easy” and “may have few unfavorable consequences and many favorable consequences” (1995). But other researchers have noted that the FAE, unlike outcome bias, lends itself to analysis according to EMT. In particular, Haselton and Nettle (forthcoming), Haselton and Buss (2003), and Andrews (2001) have approached from an evolutionary perspective the same “consequences” that Gilbert and Malone mentioned. Andrews suggests that, when our ancestors evaluated the trustworthiness of suspicious others, the costs of an erroneous inference of guilt were relatively low and the costs of an erroneous inference of innocence were often high. This cost differential could enable a bias to evolve such that humans err on the side of assuming that behavior reflects disposition. Furthermore, as Andrews notes, correspondent causes of behavior are few but potential noncorrespondent causes of behavior are potentially infinite in number. So I would add that correspondence is empowering in a way that noncorrespondence is not: one can manage problems with familiar agents but one cannot manage problems with one from an unknown number of potentially infinite causes. In this way, a finding of noncorrespondence, although reasonable, may be worthless. So the FEA, according to EMT, may reflect how relatively empowering and worthless findings of correspondence and noncorrespondence are. In accordance with Kahneman and Tversky’s theory, the FAE may also reflect our cognitive limitations when evaluating the behavior of others. So, given our cognitive limitations, the heuristic “the person’s behavior follows from her disposition” may be preferable to a less efficient and more comprehensive evaluation of potentially infinite number of noncorrespondent causes. Finally, in the face of so many potential noncorrespondent causes, the FAE might just reflect another cognitive bias: the need for closure, whereby individuals have a need to form and freeze a definite position on any given issue (Kruglanski & Webster 1996). Alternatively, the FAE may be a result of the ambiguity effect, whereby people prefer options for which a favorable outcome is known over options for which a favorable outcome is unknown (Ellsberg 1961).
Considering that the free will problem is about given-ness, the FAE becomes relevant to the free will problem when one observes another performing not just any action but the particular action of forming one’s character. According to the FAE, observers would overestimate the importance of the person’s character, and underemphasize the importance of situational constraints, when explaining the other’s character formation. In the limiting case, one would regard the characters of others as forming themselves ex nihilo. The FAE allows the agent, looking onto the View from Nowhere, to exist in uneasy tension between its currently blank and future selves.
In enabling people to jump out of the View from Nowhere, the FAE may work in concert with another cognitive bias: anthropomorphic bias. Anthropomorphic bias is the tendency of people to attribute humanlike qualities to non-human things. Although this bias has not received enough attention (Guthrie 1993: 57), classic experiments have demonstrated its existence (Dennis 1953), and it is often discussed in the literature on religion (Guthrie 1993; Barrett & Keil 1996; Caporael 1986). In the religious context, Guthrie notes that Bacon (1960: 51-52), Hume (1957: 29), Darwin (1871: 67), Nietzsche (2001: 78), and Freud (1989) explained religion, in part, as anthropomorphism. The tendency is so common that it has its own name in literary criticism: the pathetic fallacy (Abrams 1993). As Hume observed:
“There is an universal tendency among mankind to conceive all beings like themselves, and to transfer to every object, those qualities, with which they are familiarly acquainted, and of which they are intimately conscious. We find human faces in the moon, armies in the clouds; and by a natural propensity, if not corrected by experience and reflection, ascribe malice or good-will to every thing, that hurts of pleases us. Hence the frequency and beauty of the prosopopoeia in poetry; where trees, mountains and streams are personified, and the inanimate parts of nature acquire sentiment and passion. And though these poetical figures and expressions gain not on the belief, they may serve, at least, to prove a certain tendency in the imagination, without which they could neither be beautiful nor natural.”
Guthrie notes that many of these
thinkers cited the importance of familiarity and comfort in explaining religion
as anthropomorphism. According to these
theories, people distrust the unknown (this observation anticipates the
discovery of the mere exposure effect), and so they prefer to believe God is
like them. Although Guthrie claims that
such theories have “some small truth,” he prefers his own “cognitive,
evolutionary, and game-theoretical” one.
According to Guthrie, guessing that something is human is a “good bet.” In explaining his theory, Guthrie suggests
two elements of EMT: the uncertainty of the situation (“in a complex and
ambiguous world our knowledge always is uncertain”) and the relative costs of
errors (“if we are right, we gain much, while if we are wrong, we usually lose
a little”) (1996). His theory would also
satisfy the third element: distinguishing between correspondent and
noncorrespondent causes of behavior would have had “recurrent effects on
fitness over evolutionary history.”
Indeed, Haselton and Nettle observe that “Guthrie uses error management
logic” to explain anthropomorphic bias.
Both Guthrie (2002), as well as Atran and Norenzayan (2004), have later
elaborated on the use of EMT to explain anthropomorphism.
Anthropomorphic bias becomes relevant to the free will problem when one considers this long tradition of homunculus theories. In particular, homunculus theories are an expression of the anthropomorphic bias with respect, not to other objects such as clouds or animals, but to subdivisions of the human mind. Recall the agent, in uneasy tension between its currently blank and future selves, overlooking the View from Nowhere. A rational observer would regard this agent’s character as blank. Tension might arise, however, if an observer projects anthropomorphic features onto this agent. In particular, an observer might project anthropomorphic goals and values. So, from amongst the vast space of characters (the overwhelming majority of which will be non-human), the agent might prefer a human character. This homunculus theory enables and strengthens the FAE in the context of one’s character formation by positing a preexisting character which can form itself ex nihilo.
There is strong evidence that just such an anthropomorphic bias has been motivating some of the parties in the free will debate. Kane notes that “[s]ome philosophers reified the Will as a mysterious homunculus within the agent” (2001); Claxton writes that “[cognitive scientists] have been at pains to do away with the homunculus which is so often concealed within the folk theory of personhood of which free will is such an integral part” (2000: 104); and Walter observes that “[o]ur everyday explanations rest on the assumption that the essence of an agent is something at its core, conceived as a soul or homunculus” (2001). Dennett, a compatibilist, criticizes Kane, a libertarian, for committing this homunculus fallacy (2003: 123); similarly, I have criticized Fischer, a prominent compatibilist, for also thinking of human agency in this way (forthcoming).
Another cognitive bias which may aggravate the FAE is the asymmetrical attribution of blame. Research shows that observers of immoral actions are more likely to infer correspondence between the actions and the actor’s character (Reeder & Spores 1983). In contrast, observers of moral actions are more sensitive to the situational constraints which may have caused this behavior. These results are consistent with other research showing that people identify words implying social costs than words implying poor skill or positive qualities (Ybarra et al. 2001). The result is that the FAE may be especially insidious in the context of observing immoral, rather than moral, behavior.
Asymmetrical attributions of blame may reach their zenith in the process of demonization (Ellard et al. 2002). Demonization is an unwillingness to empathize with another such that the person regards this other as evil. Despite correcting for any potential differences in personality, by having the same person write both a story as a perpetrator and a story as a victim, Baumeister, Stillwell, and Wotman found significant differences between the two kinds of stories: perpetrators recalled much shorter time spans, victims regarded the actions of perpetrators as more inexplicable, and victims regarded the perpetrator’s actions as more harmful (1990). One might worry that these findings are only relevant to attributions of evil but Baumeister and Vohs are quick to suggest that demonization admits of degress and so may be a common feature of our responsibility practices:
“To be sure, our research sample consisted of many everyday conflicts and misdeeds, few of which were sufficiently important to qualify for the grandiose term evil. Our assumption, however, is that similar processes operate in everyday transgressions as in large-scale misdeeds, and that if anything, the gap between victim and perpetrator would probably be even larger in horrendously evil events than in petty, everyday conflicts.” (2005: 87)
In the context of Nagel’s The View from Nowhere, the second of these differences is the most relevant. The researchers reported that although “[b]oth victims and perpetrators distorted their stories—and to almost identical degrees,” “the weight of the evidence tends to be closer to the perpetrators’ accounts” (2005: 89-90). This is so because “[p]eople rarely attack for no reason” even though the “perpetrator’s motives are often opaque to the victim” and “victims cannot or will not see this perspective” (2005: 88-89). These results suggest how the process of demonization puts the other into the View from Nowhere—where mitigating circumstances disappear.
In explaining the cause of asymmetrical attributions of blame and demonization, researchers have noted that these biases lend themselves to EMT analysis (Haselton & Nettle, forthcoming). Such analysis requires that the respective costs of false positives and false negatives differed over evolutionary history. But asymmetric attributions of blame reflect, by distinguishing between moral and immoral actions, these differing costs between false positives and false negatives. For example, the cost of mistakenly assuming that a murderer is a friend may have been more costly than mistakenly assuming that a friend is a murderer.
The
FAE, together with anthropomorphic bias and asymmetrical attributions of blame,
provides a more plausible context in which the endowment effect and outcome
bias give people an irrational reason to have preferred their later characters
from amongst the space of possible characters.
These four cognitive biases form a potent mixture which inclines one to
regard others as possessing more control over their lives than actual control;
we regard others as possessing control over their lives which approaches
novelist* control. Together, these
biases inoculates people against the View from Nowhere’s paralying
sting—allowing them to jump into the View from Somewhere.
This section has focused upon how
people attribute responsibility to others. In Nagel’s famous essay, he calls this the
problem of responsibility. But
Nagel noted another problem—which is perhaps more disturbing—involving how we
attribute freedom to ourselves.
Nagel called this the problem of autonomy. Although the two problems sometimes
intersect, distinct cognitive biases may account for inflated senses of
responsibility and autonomy. The
following section (III) will focus upon this second problem of
autonomy. In particular, it will explore
the “positive illusions” which inflate one’s sense of control and suggest a
novel explanation for how these illusions evolved.
III. False Advertising
and the Positive Illusions
Humans are vulnerable to a variety of positive illusions about themselves. In a seminal article, Taylor and Brown reviewed and integrated the research on such positive illusions and concluded that they be a sign, not of mental illness, but of mental health (1988). Taylor and Brown later defended (1994) their work against, and their conclusions seem to survive, a critique by Colvin and Block (1994).
Taylor and Brown divide the positive
illusions into “unrealistically positive self-evaluations, exaggerated
perceptions of control or mastery, and unrealistic optimism.” First, consider the tendency of humans to
evaluate themselves in irrationally positive ways (Greenwald 1980). This tendency is related to the
These positively biased self-assessments may complement the endowment effect in explaining the illusion of free will. The endowment effect would allow one, given no consensus on values, to value what one has been given. In contrast, positively biased self-assessments would allow one, given a consensus on values, to feel that one has maximized the traits that would further those values. But it is not clear that the two biases are so distinct or, if they are, that the former has no role to play in the free will problem. Consider the measure, not of traits such as warmth or integrity whose value is uncontroversial, but of having a good character in general. Unlike being more or less warm, or having more or less integrity, there seems to be little reason to prefer being the kind of person one is to being someone else. For example, there is little reason to prefer, within a certain healthy range, being more or less sensual or more or less introverted—to regretting the accidents of birth which have made us one kind of person rather than another. Positively biased self-assessments may, however, in concert with the endowment effect, incline people to feel that they have better characters, even in this controversial sense, than others—allowing people to jump out of the View from Nowhere into the View from Somewhere.
The relevance of another positive illusion to the free will problem is more obvious—the illusion of control. Taylor and Brown trace the literature on this illusion to Langer’s classic studies in the context of gambling (Langer 1975; Langer & Roth 1975). In those experiments, Taylor and Brown showed how people feel they have more control over situations than actually have. For example, people with a light switch overestimate the amount of control they have over whether a semi-randomly flashing light turns flashes. Similarly, a wide body of literature demonstrates the same positive illusion in situations that are heavily determined by chance (Crocker 1982).
Taylor and Brown do not mention another cognitive bias which may also belong to the positive illusions—the trait ascription bias. According to the trait ascription bias, people regards themselves as variable in terms of personality, behavior and mood while regarding others as being less varied and more predictable. In the original study on this bias, Kammer had 56 undergraduates rate the behavioral consistency of themselves and their friends (1982). Kammer discovered the undergraduates rated their friends’ behavior as more consistent than their own. To the extent that predictability is undesirable, the trait ascription bias may represent another positive illusion.
Taylor and Brown speculate that the positive illusions are adaptive because they promote mental health by: increasing happiness or contentment, enabling social bonding, and increasing the capacity for creative or productive work. In particular, they argue that the positive illusions can stimulate creative or productive work by facilitating intellectual functioning and increasing motivation or persistence. In support of this claim, Taylor and Brown note that “[p]ositive conceptions of the self are associated with working harder and longer on tasks (Felson, 1984); perseverance, in turn, produces more effective performance and a greater likelihood of goal attainment (Bandura, 1977; Baumeister, Hamilton, & Tice, 1985; see also Feather, 1966, 1968, 1969)” (1988). Although Taylor and Brown discuss how the positive illusions may be adaptive they do not speculate as to how such illusions evolved.
Haselton
and Nettle note that Taylor and Brown’s theory, like Guthrie’s theory of
religion as anthropomorphism, “tacitly contains an error management argument”
(Haselton & Nettle, forthcoming). In
particular, Haselton and Nettle seize upon Taylor and Brown’s emphasis upon
motivation and persistence. Situations
in which individuals must decide whether to stop or persist in their efforts
may satisfy the three elements of EMT: they involve uncertainty (if I continue,
will I eventually succeed?), the decisions had repeated effects upon fitness
throughout history (if I succeed in wooing a mate, or killing a rival, my genes
will propagate at the expense of those who fail), and the relative costs of
errors or correct inferences were asymmetrical (irrational persistence often
costs less than irrational quitting).
I would suggest an alternative—and
perhaps more disturbing—explanation for the prevalence of the positive
illusions. In my estimation, Haselton
and Nettle further the tendency of psychologists after
It is uncontroversial that people often lie in the mating context. For example, “people report being deceived by friends about mating rivalry more often than they themselves report engaging in deceit about rivalry, and women more than men deceive each other about how sexually experienced and promiscuous they are” (Bleske & Shackelford 2001). Similarly, Buss cites research (Keenan et al. 1997) showing that “men motivated to seek casual sex frequently attempt to deceive women about their commitment, social status, and even fondness for children—domains of deception about which women are well aware” (Buss 2003: 283). These results are consistent with widespread deception in the plant and animal kingdoms. Searcy and Nowicki note that, although some have expressed doubt about deception in the animal kingdom, these doubts have been invalidated by adopting a functional definition of deception and by focusing upon individual selection instead of group selection (2005: 219). Buss notes two vivid examples: some male insects give gifts to female insects but take them back after mating (Thornhill & Alcock 1983) and some orchids mimic female wasps such that male wasps attempt to copulate with them; the male wasps fail to compliment but succeed in carrying pollen (Trivers 1985). These results are also consistent with the principle, from communications theory, that signalers have been under selection pressure to manipulate, even through deception, receivers (Krebs & Dawkins 1984). In this context, Trivers obverses that deception is “a parasitism of the preexisting system for communicating correct information.” (Buss 2003: 105).
One difference between the positive illusions and deception in other contexts immediately presents itself: the usual form of deception involves knowledge of one’s deceit and the intent to deceive whereas humans experiencing positive illusions believe their lie. So the positive illusions involve not just deception but self-deception. One must therefore explore the evolutionary origin of self-deception in order to explain the evolutionary origin of the positive illusions.
Trivers first suggested how self-deception might evolve in his classic forward to Richard Dawkin’s The Selfish Gene:
“The arguments themselves extend in many directions. For example, if (as Dawkins argues) deceit is fundamental in animal communication, then there must be strong selection to spot deception and this ought, in turn, to select for a degree of self-deception, rendering some facts and motivates unconscious so as not to betray—by the subtle signs of self-knowledge—the deception being practiced. Thus, the conventional view that natural selection favors nervous systems which produce ever more accurate images of the world must be a very naïve view of mental evolution” (1976).
Trivers’ fundamental insight is that the best liars believe their own lies, because they cannot betray their knowledge of the deceit, and so there would be selection pressures upon deceivers to deceive themselves. He later explored the evolutionary origins of self-deception in more detail (2000). First, Trivers sympathizes with the Taylor and Brown’s hypothesis that the positive illusions bring intrinsic benefit to the individual. Second, Trivers argues that self-deception can express itself in five ways: denial of ongoing deception, unconscious modules involving deception, self-deception as self-promotion, the construction of biased social theory, and fictitious narratives of intention. There seems to be some overlap between these kinds of situations; the first three may combine to form, amongst other things, the positive illusions. People may believe that are better than they actually are, or have more control over their lives than they actually do, to better convince others of these falsehoods.
Indeed, Trivers anticipates just this explanation for the positive illusions. He observes that one source of self-deception “has to do with self-promotion, self-exaggeration on the positive side, denial on the negative, all in the name of producing an image that we are ‘beneffective,’ to use Anthony Greenwald’s apt term, toward others” (2000: 117). But one does not need the idea of being “beneffective,” and inviting reciprocal altruism, in order to explain the positive illusions. Trivers does not mention a more fundamental form of cooperation amongst the animal and plant kingdoms, including homo sapiens—sex.
Being
above average on any given positive trait is sexy. This is especially true considering that
people may adapt their preferences to the conditions of the mating
marketplace. In the class study on the contrast
effect, Kenrick and Gutierres found that men rated an average woman as less
attractive after watching Charlie’s Angels (1980). Buss discusses several findings that followed
this classic experiment and shows how they are consistent with an evolutionary
theory of happiness (2000). For example,
people disregard objective measures of well being humans are on a “hedonic treadmill” (Diener et al. 1999) and because “differential
reproductive success is the engine of the evolutionary process” (Buss
2000). Buss describes how an
evolutionary theory of happiness reveals that we are on a treadmill such that
supposed increases in well-being, like winning the lottery, or the increase in
standard of living throughout the twentieth century, do not cause long term
increases in happiness. Being a rich
supermodel is less attractive if everyone is a rich supermodel. In the mating marketplace we are all graded
on a curve.
Similarly, control over one’s life is sexy. From the perspective of female choice, one’s good intentions are worthless to a woman if one is powerless to effect them. Buss details how women find economic capacity, social status, ambition and industriousness, dependability and stability, intelligence, size and strength, and good health to be attractive (Buss 2003)—all indicators of control over one’s life. As Henry Kissinger observed, “power is the ultimate aphrodisiac.”
There is evidence that such power is important to libertarian conceptions of free will. For example, Kane defends a theory of event-causal libertarianism according to which free will involves “ultimate control” (1999) or “ultimate dominion” (1985) over one’s life. Other theories of agent-causal libertarianism characterize humans as godlike unmoved movers. (O’Connor 2005). If power is the ultimate aphrodisiac, then libertarian free will would render one extremely sexy.
The case for the attractiveness of creative variability is more subtle than the case for the attractiveness of control over one’s life. This case is largely based upon Miller’s treatment of the subject (2001). Miller begins by observing that the behavior of simple organisms is often predictable; he further notes that, to the extent that such organisms are prey, this predictability is problematic. Miller describes how Von Neumann discovered that the best strategy in certain games from game theory involve randomness. Later, Fisher and others discovered examples of just such randomized strategies in the animal kingdom. These considerations lead Miller to conclude that this overall strategy, which Driver and Humphries called protean behavior, is a fundamental adaptation found in almost all mobile prey. Miller speculates that protean behavior expresses itself in humans through creativity. Such creativity is strongly correlated with reproductive success (Nettle & Keenoo 2005).
These last observations become relevant to the free will debate when we consider just how important creativity and unpredictability are in that context. One traditional definition of free will is the ability or power “to do otherwise”; a large literature explores just what this power entails and whether moral responsibility requires it. Moreover, one virtue of libertarianism is that it renders human decisions unpredictable by definition. Amongst compatibilists, Fischer has characterized the value of free action as the value of creative self-expression (2005; 2006a).
It is important to note that, although the concept of free will implies both control and unpredictability, perhaps expressed in the illusion of control and trait ascription bias respectively, these may represent two entirely different—and mutually exclusive—features of free will with distinct evolutionary origins. For example, Richard Double claims that no libertarian account of free will can jointly satisfy the following three requirements: control, rationality, and the ability to choose otherwise (1991, chapter 8). Putting aside the worry about rationality for a moment, one can note the tension between the desire for control and the desire for unpredictability (which would correlate with the ability to choose otherwise). Each seems to do violence to the other: greater control over one’s life ensures that one’s behavior will follow from one’s character—enhancing predictability—and greater unpredictability ensures that one’s behavior does not follow from one’s character—undermining control. This tension may reflect the distinct evolutionary origins of these desires which have been coupled together within the concept of free will.
Those
who study the free will problem must remember that throughout history there
have been tremendous selection pressures upon humans to become better than
average, to obtain control over their lives, and to express creative
variability. Furthermore, to the extent
that not everyone has been able to attain these lofty ideals, there have been
tremendous selection pressures for humans to fake them. In the same way that a man might lie about
being married in a singles bar, so too might a man lie about having more
control over his life than he actually does.
One can consider this phenomenon using
By analogy, suppose that a corporation such as Nike engages in behavior that is unattractive to the consumer. For example, suppose that some accusations are correct and Nike employs people in sweat shop conditions for unreasonable pay. Consider Nike employees responsible for marketing the company’s brand. Of course, the marketing employees would not advertise the sweatshop conditions. Indeed, they might even lie about them. It is not unreasonable to suppose that such employees would be so committed to their jobs that start to believe their own lies—and this makes them better liars. Finally, suppose that a group of marketing employee meets a person in a wheelchair. These marketing employees might tell the handicapped person that Nike does not employ workers in sweatshop condition even though this person is unlikely to ever buy athletic wear. The employees might do this because they are protecting a brand name. Handicapped persons are not islands; they may spread word about Nike to potential customers even if they themselves will never purchase Nike products. Marketing employees are promoting a brand and cannot afford to make subtle and dangerous distinctions between who may and may not learn of information damaging to that brand. By analogy, the positive illusions may have evolved because they allow people to promote themselves, not only to potential mates, but to society in general—and the best liars believe their own lies.
These first two sections (II and III) have explored how certain cognitive biases might have evolved because they were adaptive. These adaptive biases would give one direct reasons to regard others (I) and the self (II) as having more control over their lives than is actual—having control over their lives that approaches novelist* control. Such biases might have been adaptive in accordance with one or more of the following: a heuristics and biases theory, error management theory, or false-advertising through self-deception. The next section (IV) will argue that evidence of one last group of cognitive biases would only provide indirect reasons for regarding others and the self as possessing more control over their lives than is actual; nevertheless, these biases lend strong support to a great philosopher’s error theory of belief in free will. The section will further argue that such biases evolved as adaptations or in other non-adaptive ways.
IV. Irrational
Optimism
The three categories of positive illusions discussed by Taylor and Brown involve biased evaluations of the self. The previous section considered two of these three categories: irrational self-esteem and the illusion of control. This section will focus upon the third positive illusion: irrational optimism. Irrational optimism is the central focus on this section but it is itself just a subset of the remaining cognitive biases which favor belief in free will. The compliment of the positive outcome biases within this larger set is composed of biases which cause belief inertia given the popularity of belief in free will: the bandwagon effect and status quo bias. From this perspective, the popularity of free will presents another obstacle to rational belief in that power.
According to the bandwagon effect, people regard popular ideas as more accurate than unpopular one. The effect is discussed in a classic article (Lee & Lee 1939), in the voting context (Allport 1940; Simon 1954: 245-253; Carter 1959), and has been recently repeated in that context (Mehrabian 1998; Marsh & O’Brien 1989). For example, Mehrabian reports two experiments which showed how bogus polls significantly affected voting, “supporting the bandwagon effect.”
One implication of the bandwagon effect is the “spiral of silence.” According to Noelle-Neumann, individuals may value belonging with the majority more than they value holding their own opinion (1974). The result is a “spiral of silence” which magnifies the prominence of majority views and suppresses minority ones. A recent meta-analysis concluded that the “spiral of silence” effect is small but significant (Glynn et al. 1997).
Related
to the bandwagon effect is another cognitive bias which would favor belief in
free will: status quo bias. Status quo
bias is the irrational preference people express for things staying the same
(Samuelson & Zeckhauser 1988). For
example, Kahneman, Knetch, and Thaler cite the following example of status quo
bias (1991). Consumers of a California
power grid were divided into two groups such that one had more reliable service
than the other. The consumers were then
surveyed about their preferences. Both
groups selected their status quo as the most desirable option. Kahneman, Knetch, and Thaler report that
subjects in another experiment expressed similar preferences for the status quo
when considering whether to adopt two kinds of automobile insurance (Johnson et
al. 1993).
I know of no literature on the evolution of the bandwagon effect and status quo bias. But one might speculate as to how these biases evolved. First, one might subject them to EMT analysis. The question of whether to join the majority or not would be a recurrent event throughout history that affects an individual’s fitness. Furthermore, the relative costs of false positives and false negatives, or correct positives and correct negatives, might vary. For example, the cost of mistakenly introducing an unpopular idea to the group, and incurring its wrath or disapproval, might outweigh the cost of mistakenly suppressing the idea and preserving the status quo. This is especially true considering that the cost associated with announcing new ideas depends, not just on how accurate the new ideas are, but how much people have invested in old ideas. Nobody likes to admit they were wrong and people will inflict fitness costs on others who threaten to expose their errors. The analysis here would be similar to that of the mere exposure effect, according to which selective pressures might favor a conservative attitude in an uncertain world. This variance would put selective pressure upon the species to evolve a cognitive bias tuned to avoid the greater costs—at the expense of accuracy. In considering the evolution of these biases, one might also note that Kahneman, Knetch, and Thaler regard status quo bias as an implication of the endowment effect and loss aversion (whereby people express undue dislike for losing their possessions). To the extent that their theory is correct, the analysis of how those other biases evolved would apply to status quo bias as well. Furthermore, the bandwagon effect and status quo bias bear a resemblance to the contrast effect and to that extent the discussion of how the contrast effect evolved would also apply here.
These belief inertia biases only show that people find novel ideas unpalatable but not that they find them unlikely. To complete this section’s case for the remaining cognitive biases favoring belief in free will, one must return to the positive illusions which relate, not to self-evaluations or the illusion of control, but to the likelihood of good things happening. These biases include belief bias, the confirmation bias, and the choice-supportive bias. Each of these belongs to the set of positive outcome biases because they incline people to regard certain positive outcomes as more likely than negative ones.
First, consider the third positive illusion: the valence effect. According to the valence effect, people overestimate the likelihood of good things happening to them and underestimate the likelihood of bad things happening to them. This effect seems to be quite general. For example, Rosenhan and Messick reported that, all else being equal, subjects considered turning over a card with a smiling face on it to be much more likely than turning over a card with an angry face on it (1966). They found no effect during an experiment that used the neutral images of big and little kangaroos. Similarly, in their review of the positive illusions, Taylor and Brown cite the following studies in support of the valence effect:
“People estimate the likelihood that they will experience a wide variety of pleasant events, such as liking their first job, getting a good salary, or having a gifted child, as higher than those of their peers (Weinstein, 1980). Conversely, when asked their chances of experiencing a wide variety of negative events, including having an automobile accident (Robertson, 1977), being a crime victim (Perloff & Fetzer, 1986), having trouble finding a job (Weinstein, 1980), or becoming ill (Perloff & Fetzer, 1986) or depressed (Kuiper, MacDonald, & Derry, 1983), most people believe that they are less likely than their peers to experience such negative events.” (1988)
The valence effect may account, in subtle ways, for a variety of cognitive biases related to belief inertia. First, belief bias is the phenomenon whereby subjects reject valid arguments with unbelievable conclusions and accept invalid arguments with believable conclusions (Evans et al. 1983). Belief bias is an error of deductive reasoning and may be an example of the larger phenomenon of belief perseverance. Belief perseverance is the cognitive bias which expresses itself as the irrational maintaining of beliefs in the presence of contrary evidence. For example, Anderson, Lepper, and Ross found that after subjects formed beliefs they would discredit contrary evidence and their beliefs would remain resistant to change (1980).
A
subset of the valence affect may be confirmation bias. Confirmation bias enables belief perseverance
by inclining people to seek confirmation of their preconceptions at the expense
of alternative hypotheses. In the
classic experiment, Wason presented subjects with three numbers (1960). The subjects were asked to determine the
appropriate rule governing the sequence by generating their own sequences and
submitting them to the researcher for feedback.
Wason observed that subjects tended to submit only positive examples
which would confirm their suggested rule and did not submit negative examples
which would disprove it. He called this
tendency the confirmation bias. A recent
review of research on the confirmation bias concluded that “[i]n the aggregate,
the evidence seems to me fairly compelling that people do not naturally adopt a
falsifying strategy of hypothesis testing” (Nickerson 1998).
A third positive illusion related to
belief is choice-supportive bias.
Choice-supportive bias expresses itself in the tendency of humans to
remember their chosen options as having more positive attributes than they do
in fact. For example, Mather, Shafir,
and Johnson gave subjects a choice between two options and later had these
subjects evaluate their choices (2000).
The subjects reported choice-supportive bias by tending to attribute,
both correctly and incorrectly, more positive attributes to their chosen
options than to their alternatives.
Earlier, this article suggested that the belief inertia biases may evolved through EMT in a way analogous to the evolution of one or more of the mere exposure effect, the endowment effect, loss aversion, and the contrast effect. In contrast, the positive illusions with respect to belief may have evolved in a way analogous to the other positive illusions. As discussed in the previous section, the positive illusions may have evolved because they are sexy. Similarly, the valence effect, belief bias, confirmation bias, and choice-supportive bias may represent false-advertising which benefits the individual at the expense of others. Furthermore, as Trivers argued, one of the best ways to lie is to believe one’s own lie.
These cognitive biases about both the palatability and likelihood of altering beliefs in the face of contrary evidence favor belief in free will in three distinct ways. First, to the extent that people regard free will as desirable, the valence effect would incline them to inflate the likelihood that free will exists. There is no denying that people tend to regard free will as a good thing. For example, considering the aspect of free will involving control, it is important to must note that feeling in control is an essential ingredient to people’s happiness (Larson 1989).