SURVIVOR, SOCIAL CHOICE, AND THE IMPEDIMENTS TO POLITICAL RATIONALITY: REALITY TV AS SOCIAL SCIENCE EXPERIMENT

Ed Wingenbach

Department of Government

University of Redlands

Redlands, CA

Social choice theory occupies a privileged explanatory position within much of the social sciences, particularly economics, psychology, sociology, and political science. Social choice theory argues that the outcomes of collective action may be explained in terms of the calculation of costs and benefits in pursuit of individual preferences. Despite its near hegemonic position in the leading journals of political science and economics, most of the laboratory evidence gathered over the last 40 years mitigates against the social choice explanation of collective behavior. The absence of compelling empirical evidence may arise from our inability to create sufficient incentives or impose adequate controls to test the theory. In short, though subjects appear to be driven by social pressure, ethical norms, or failure to recognize rational strategy, the problem lies not in the results but the experimental conditions. Create high enough incentives or a sufficiently focused environment, and truly rational behavior would emerge. Social scientists have neither the funds nor authority to impose such conditions. Fortunately, CBS does.

This chapter analyzes the game structure set up by the rules and incentive structure of Survivor and elucidates the model of behavior predicted by social choice theory. It then analyzes the actual behavior of the contestants, showing only five of the sixteen follow the rational model to any significant degree. Next, the essay categorizes the non-rational factors which motivate the behavior of the majority of contestants; these factors include personal ethics, public image, duty, and emotions. I conclude that the Survivor experiment demonstrates the importance of acknowledging the limits of self-interest as a predictor of behavior, integrating moral/ethical beliefs into rational analysis, and recognizing the limits of individual rational capacity.

SOCIAL CHOICE AND THE SURVIVOR EXPERIMENT

Social choice theory should be viewed as a subset or application of rational choice theory. Though its roots might be traced back as far as Thomas Hobbes (1651, see particularly C. 13), modern rational choice emerged as a systematic and identifiable approach about mid-century, most prominently in the work of Downs (1957), Arrow (1963), and Buchanan and Tullock (1962); it has come to dominate contemporary economics and political science, and is important in a number of other fields such as psychology, law, history, and sociology. Rational choice encompasses a wide range of methods, all of which attempt to explain and predict human behavior by assuming the primacy of instrumental rationality, which merely means that individuals will always choose to pursue their self-interest by the most efficient strategy available to them. Efficient here indicates that individual actors will adopt the course of action maximizing their expected utility given the information available to them at the time of the decision. Rational choice methodologies attempt to systematize the calculations of interest and efficiency to produce predictive scientific models about large scale behavior.[1] If a theorist can identify or hypothesize the utility orderings at play in a given situation and accurately assess the strategic options open to the players, it should be possible to predict the outcome of any situation.

Social choice theory shifts the emphasis slightly, concerning itself with the formation and ordering of preferences by individuals and the problem of accurately aggregating these divergent preferences (for an overview, see Plott, 1976). Generally, social choice focuses on elections and decision making in groups. Riker's (1982) classic text reveals the profound implications of social choice for democratic politics: if more than two people are involved in a decision, and more than three alternatives exist (or individuals have the option of introducing other alternatives), then no method of aggregating the preferences of individuals produces a coherent outcome.[2] Thus, the theory both accurately explains large scale economic behavior and undermines the possibility of meaningful democratic governance.

The ascendance of social choice is not without critics. As Green and Shapiro (1994) have shown, the theory's predictive claims are vastly overstated, the empirical evidence sketchy, and the logic often so broad as to be tautological. The flaws in social choice explanations of political behavior fall into four categories. First, social choice theory tends toward circularity. When construed broadly, the claim that all actions are rational defies falsification. ^[3] Second, the social scientific evidence seems not to support the theory. As Green and Shapiro (1994, pp. 194-95) show, where good data exists it tends to refute the theoretical predictions, and where the theory seems to work best, little data exists.[4] Third, laboratory tests attempting to generate and observe rational behavior have been almost completely unsuccessful. In fact, according to Hauptmann and Clement=s (1999) excellent review of lab tests in economics and political science since the 1950's, the only experiment to yield genuinely rational outcomes was one using graduate students in a PhD program in economics. We might conclude that rational behavior is observed in the lab only among those already trained in rational choice theory. Fourth, it is difficult to determine motives with any certainty, and discerning instrumental rationality without a clear understanding of motive is all but hopeless. More technically, to predict rational behavior, the theorist must assume knowledge of the actors' maximum utility (perfect outcome) and their expected utility (best plausible outcome given current constraints), while assuming the actors have full information about their options. Unfortunately, transparency of motive and full information rarely characterize real human beings in complex situations.

Thus investigators must assume that all actors are: (1) utility maximizing agents, meaning they will pursue the strategy most conducive to achieving their desired end in the most profitable fashion; (2) aware of the expected utilities of the various options given the current conditions and actions (anticipated and observed) of other players; and (3) able to access and understand this information, or, in social choice parlance, possess Afull information.@ Eliminate any of these factors and prediction or explanation become increasingly difficult. Unfortunately, people often possess conflicting motives or contrary utilities, rendering their behavior difficult to predict in terms of utility maximization. Even individuals with clear and noncontradictory preferences may not be able to judge their expected utility as accurately at the time of decision as an analyst can, nor are they usually able to anticipate or observe the actions of their competitors. Lastly, full information is almost always absent, for a variety of obvious reasons.

SURVIVOR AS LABORATORY EXPERIMENT

Given these difficulties, a rigorous experiment testing social choice predictions in an environment both controlled and easily observed would be of great significance. The first season of Survivor provides exactly such a test. The Survivor game offers: (1) a clear incentive structure for the players; (2) unambiguous rules for conduct and victory; (3) complete transparency, both for the players, who can easily observe the strategic behavior of others, and the experimenters, who have all actions on tape for future review; (4) long term observation, allowing the predictions to be tested over time and multiple iterations; (5) opportunities for learning, since failure is openly punished and competition relatively constant (thus allowing correctives for mistaken irrationality); (6) in depth interviews with the subjects discussing their choices and some access to psychological profiles; and lastly, (7) a series of regular votes along with the voters= self-reports on their voting choices. In short, Survivor offers the perfect test of social choice theory.

The Model

I propose a relatively simple model to understand Survivor as an experiment, making behavioral predictions requiring only minimal rationality and obvious choices. Such a model is most likely to generate a favorable outcome for the theory, thus rendering the results difficult to dismiss should the theory fail. My underlying assumption is that all players share the same goal B they all wish to win the game. The reward for winning is ten times more lucrative than second place, and after third and fourth place the monetary rewards decline precipitously. Other than BB, every player expressed an unambiguous desire to win the first prize money.[5] Thus all the players present a clear and uncontradictory awareness of their own maximal utility.

One obvious objection to my premise must be addressed. It is conceivable that players might envision an alternative method to attain the monetary award, or even exceed it. That alternative is fame.[6] This powerful objection may be refuted in at least three ways. First, it is important to remember that the experiment occurred far before the series first aired. The subjects could not have anticipated the frenzy that built around the show. Second, even if the subjects assumed the show would be a success, winning the first prize remains the best way to maximize the return of fame. The winner would receive more publicity and thus have the best opportunity to cash in on fame, as Richard Hatch was well aware in his pregame planning (Lance, 2000, “Introduction”). Moreover, the longer one stays on the show, the greater the likelihood of fame. The early segments of the show are simply too crowded for any individual to make a serious impression on the audience, a factor intensified by the knowledge that the show would be edited, further reducing any player=s screen time. Thus, the pursuit of fame should, as an issue of rational behavior, be indistinguishable from the pursuit of victory. Third, the subjects all signed highly restrictive contracts consigning all story rights to CBS, constraining their ability to appear publicly for a year following the series, and giving CBS veto power over any books, articles, or interviews (Lance, 2000, pp. 16-27). Thus fame was unlikely to bring any player a return even approaching first prize, and the ancillary benefits of fame were largely dependent upon playing to win.[7]

The Team Segment

The Survivor game is divided into two distinct segments, each entailing a slightly different strategy. In the first segment, teams of eight compete with each other in competitions requiring cooperation and skill. The losing team must vote off a member. The primary goal of all players is to win the competitions and thus avoid exposure to exile. Assuming the improbability of victory in all team competitions, as all rational players would, the primary individual goal is to avoid being voted off should the team lose. As an issue of strategy, players can only influence the votes within their team (assuming all players compete to the best of their abilities in the intertribal competitions); thus, for social choice theory, players maximize their utility in the team portion by minimizing the likelihood that they will receive votes in tribal council. An efficient pursuit of this maximal utility dictates a rational actor should adopt three obvious strategies: alliance, culling the weak, and increasing personal value.

Alliance: Each team consists of eight players, allowing a majority faction (five or more) voting with discipline to control the selection of exiles with absolute certainty. Forging a majority alliance represents the single most rational and desirable strategy during the team segment. While a five member alliance would be optimal, a four member alliance might suffice if only one person out of the eight is unable to see the proper strategy and chooses to Ago it alone.@ This strategy demands a degree of exclusivity, since an eight member alliance is of no value; moreover, the larger the alliance, the more rational it becomes to defect and create a sub-alliance.[8] The obvious rational strategy thus involves forging an alliance of five players.

The experiment provides fairly clear evidence that the rational strategy was not chosen initially by the majority of the players, that those who learned to pursue this strategy did so haphazardly, and that even the players aware of the importance of alliances were largely unable to behave properly. The clearest evidence of irrationality is seen by the contrasting behavior of the two teams. The Pagong tribe was remarkably unstrategic. The tribal dynamic was characterized by a members= desire to assert their individuality, to the point that they were unable to meet even their basic needs for food and shelter, much less think strategically about voting (Burnett, 2000, pp. 73, 103). When they did act in the manner anticipated by the theory, it was for decidedly non-strategic reasons B the temporary (one vote) Afeminist alliance@ when Colleen, Gretchen, and Jenna unified to vote off Joel because of his chauvinistic attitude (see appendix, vote 6).

The Tagi tribe demonstrates more of the expected behavior; with the exception of Sean and Dirk, every member of the tribe attempted to build a majority alliance at some point. Excluding the well known Afinal four@ alliance, all of the Tagi alliances were short lived and undermined for non-strategic reasons. The first attempted alliance was between Susan, Sonja, Stacey, and Kelly, which disintegrated almost immediately due to the other members= discomfort with Stacey and Susan=s concern about the weakness of Sonja (Burnett, 2000, p. 35; Lance, 2000, p. 34; appendix, vote 1). A second attempt to forge a female alliance failed because of personality conflicts (Burnett, 2000, p. 66). Thus the data demonstrate rational behavior by six of sixteen players at best, and that among even these six rationality was neither consistent nor strategic.[9]

Culling the weak: A second obvious strategy during the team portion of the game involves ridding your team of the least valuable members. The reason should be apparent; to the extent the team avoids losing the competitions, each individual is safe from expulsion. Weakness under these circumstances may be defined by a combination of physical strength and mental resilience. The players were all aware that the competitions would be largely physical in nature and quickly learned the importance of stubborn self-discipline. Thus the rational player will vote to eliminate those people who demonstrate either physical or mental weakness.

The evidence for rationality is ambiguous, though tending toward the model=s predictions. Of the six voted off during the team portion, three were among the weakest players: Sonja, Ramona, and Dirk. Each revealed an inability to function as an asset to the team in a variety of ways. The other three, however, were relatively strong players: BB was a tenacious competitor, and one of the few on his team able to provide outdoor survival skills, Stacey was not only in excellent physical condition but also proved herself mentally strong during the grub-eating challenge, and Joel was a physical standout. Nonetheless, the exile of these three can be explained based on rational reasons. BB had requested to leave and had already built the group=s shelter, while his age indicated a limited value in physical challenges. Joel was the last loser of the team votes, so it would be rational to eliminate him prior to the individual portion of the game where he would be a threat to everyone.[10] Stacey, then, seems the only irrational vote. But even her situation has a rational explanation. Richard had identified her as his main threat and used his influence in the Tagi alliance to have her voted off (Lance, 2000, pp. 50-51). In short, the players generally acted rationally according to the criteria of Aculling the weak.@ The fact that a strong argument could be made for the expulsion of Gervase, Colleen, Rudy, and Sean, all of whom were weak players, does not particularly undermine the rationality of the choices that were made.

Increasing Personal Value: Social choice theory presumes not mere rationality but strategic rationality. People are strategically rational when they assume others to be rational and modify their own choice of strategies to account for the likely rational strategies of others. In this case, the strategic thinking need not be complex. If each player understands that the weakest is likely to be voted off the island, rational strategy dictates making oneself as valuable as possible. The physically strong need to perform well in physical competitions, the less strong might provide survival benefits or construct housing, and so on. The more indispensable players become to the team, the less likely they are to receive votes. Richard=s fish spearing ability is the best example of this strategy; as the only reliable source of food, no one could afford to risk his loss.

Given the obviousness of this strategy, it would be odd indeed to find it rare. Yet an examination of the players= actions reveals that only five of the sixteen followed this course. In Tagi, Richard (fishing), Susan, and Stacey (daily hard work around the camp) attempted to prove their value, while the rest were indifferent at best. Dirk and Sean might have been expected to act this way, since they were the youngest and strongest and would be valuable in the competitions. But they quickly proved ineffective in that arena (or no more valuable than the others) and made no attempt to compensate. In Pagong, the numbers were even lower. Only BB and Gretchen worked hard on a consistent basis, making the lives of the others appreciably better.

The Individual Segment

Once the teams merge, rational strategy shifts dramatically. Each player has two related goals: survive to the final two and remain popular with the seven who don=t make it, since they will decide who receives the grand prize. A secondary goal is to win the immunity competitions, though relying upon winning all the competitions is obviously irrational given the variety of skills tested and intentionally random elements introduced into the competitions. A minimally rational actor would pursue three strategies: alliance, culling the strong, and controlling perception.

Alliance: As with the team portion of the game, membership in a majority alliance allows a player to gain relative safety, at least until the number of players becomes very small. The best case scenario (as seen in Survivor II) would be to create a five person alliance, allowing complete control of the voting. While no group managed such a feat, the final four alliance did manage to maintain itself for a short period. Even a four person alliance is a valuable tool, since the last four players win far more money and fame than any others. Though an alliance won=t guarantee a win, it does insure the best chance to win; therefore; it is rational to both form an alliance and stay in it as long as possible.

The evidence for rationality in this aspect of the game is stronger than the team segment, but still represents a minority position. Most of the players recognized the need for an alliance once they saw the final four group in action; however, they seemed unable or unwilling to follow the rational course, even when faced with obvious costs. Thus what we see in the team segment is not ignorance of rationality but its lack of primacy. Some examples illustrate the point vividly. The Pagong tribe knew even before the merge that an alliance existed among the Tagi players and that they would need to counter it.[11] Yet even after seeing the alliance in action they failed to act while they still might do so effectively. Eventually some members attempted a desperation alliance (Colleen and Jenna; see appendix, vote 9), but it was far too little, and far too late. Only Greg seemed to understand the need for an alliance, but he waited too long to work to build one, thus demonstrating only marginal rationality (Lance, 2000, p. 60). Even the final four alliance was subject to irrationality. After the first three votes Kelly began to feel guilt and left the alliance, endangering her own survival (in fact, only her improbable streak of victories in the last five immunity challenges kept her on the island). The experiment reveals that even under the best circumstances (high incentives for rationality, clear modeling of rational behavior, opportunities for direct observation of the strategic actions of others) only Rich, Rudy, and Sue acted as social choice theory predicts, and Rudy=s actions were, as I show below, only rational in appearance.

Culling the Strong: This strategy inverts the team segment. Since the immunity challenges become individual competitions, rational players will do their best to expel those players most likely to win them, increasing the remaining players= prospects of victory. The final four alliance did make their decisions this way, though Kelly found this behavior so objectionable that she defected from the alliance. The other players’ choices were non-rational at best, primarily a result of emotional entanglements and resentment.[12] The players knew the rational strategy and saw it operating to their detriment. Social choice theory dictates that rational behavior will be learned under such conditions. Yet in Survivor the opposite occurs B rational behavior decreases based on emotional and moral considerations.

Control Perception: The last aspect of a rational strategy concentrates on the final vote. Should a person make it to the final two, they need to have done everything possible to ingratiate themselves with the seven jury members. Given the obvious fact that the final vote is the aggrieved players only opportunity to take revenge, it is best to avoid offense. A rational player will want to be liked and respected, seen as a kind, fair, honest, and friendly person. This strategy is dependent on the a pursuit of the first two strategies, as it does a player no good to be popular without an alliance to protect them (Gretchen or Colleen are good examples; each would likely have won had they made it to the final two, which insured they would be expelled by the alliance).

In the Survivor experiment, none of the three rational players makes even a minimal attempt to follow this strategy. Rudy is cantankerous and lazy, Rich arrogant and petty, and Susan obnoxious and mean. Were the players ranked based on popularity, it is likely these three would finish last. The oddest aspect of this failure lies in the fact that pursuing a strategy of popularity costs the players nothing at all and its value is easily perceived: as Rich states at the start of the game, good players must control how they are seen by the others (Lance, 2000, p. 29). Despite knowing this, the three most rational actors still act irrationally in this area.

ACTUAL MOTIVES VERSUS RATIONAL PREDICTIONS: VARIATION EXPLAINED

As a test designed to induce rational behavior, Survivor fails miserably. Of the sixteen players only five act in a consistently rational manner, using the most generous criteria: Rich, Rudy, Susan, Stacey, and Sonja. Even these five are not consistently rational and do not act in the most optimal manner when they are acting rationally. As for the other eleven, their actions pose a puzzle for social choice theory. In this section, I divide the players into six categories of motive: ethics, image, duty, emotion, unclear, and rational. I explain the meaning of each term and explain which players fit in each category. The value of this exercise lies in its demonstration of the importance of non-rational motives in human behavior, motives generally dismissed by social choice theorists as either illusory, strategic, or responses to incomplete information. The Survivor experiment shows that these less predictable and often non-rational motives dictate human behavior even under conditions ideally constructed to produce rational actions.

ETHICS (9 Players): Ethics in this usage indicates not the presence of a coherently developed moral code derived from first principles but rather a minimally developed sense of one=s own identity and conscience, whether it be consistent and explicable or merely anecdotal and shallow. Despite my less rigorous usage, ethics as a category of motive usefully describes a dominant mode of behavior among the contestants. Many of them claimed to want to win, but only in the Aright way,@ expressing the desire to behave according to the dictates of their conscience or Amoral code.@ The best way to express the conception of ethics most prominent in Survivor involves a culturally common notion of Aindividualism.@ Contestants motivated by ethics conceive of themselves as defined by their own standards of behavior, and are unwilling to sacrifice their core identity in order to win the game. They tend to value consistency of action over strategic benefit and positive self-conception over material reward. Social choice theory generally dismisses such concerns as secondary to the pursuit of tangible interests. Thus, the importance of ethics in the decision making of the players undermines the authority of the theory=s predictive value.

Gretchen exhibits the most consistently ethical behavior. She refuses Joel=s offer of an alliance out of principle, refuses to vote strategically, and acts altruistically, even when doing so benefits her direct competitors. She sums up her position by saying, AI wouldn=t feel right about myself winning the million dollars if I had to change my behavior to do it@ (Burnett, 2000, p. 91).

BB also reveals a stringent ethical standard. He values hard work and merit over everything else. Once he realizes that his values and standards are not reflected by the other players and winning will require a change in his behavior, he requests his teammates vote him off (Burnett, 2000, p. 39).

Greg follows an ethics of radical individualism. He sees himself as a nonconformist unwilling to submit to manipulation or compromise. He associates idiosyncratic individuality with honor, and refuses to compromise, even in the interest of building allies. AI=ve acted honorably the whole time and have nothing to be ashamed of@ (Burnett, 2000, p. 144).

Dirk’s evangelical Christianity provides an inflexible moral code, and his self-conception as a fundamentally generous person leads him to judge the rational strategy Amean-spirited,@ and thus unworthy of his allegiance (Burnett, 2000, p. 97).

Gervase reflects a less consciously articulated version of Greg=s individualism. In his case, the impulse to ethical consistency comes from pride: ASo I didn=t win the million dollars. So what? The important thing is that I played the game Gervase=s way. I can go home and hold my head up, knowing I didn=t change one bit@ (Burnett, 2000, p. 181).

Sean is an interesting case. His ethical code demonstrates an attachment to procedural justice. He wants to do the right thing and seems to feel that the only way to avoid injustice is to follow an unbiased method of voting. Hence his attachment to the Aalphabet strategy,@ which allows him to claim that he has treated everyone fairly, despite the deep antipathy it generates among the very people to whom he wants to be fair.

Collen and Jenna both articulate similar approaches. Each seems to value their own integrity, demonstrating an unwillingness to violate their notions of proper treatment of others. Even their attempt to form a belated alliance is half-hearted; they recognize the rationality of this approach but are unable to pursue the strategy with any consistency. Collen=s comment captures their positions nicely: AI just want to leave with my integrity intact@ (Burnett, 2000, p. 147).[13]

Kelly might not seem an obvious candidate for this category, given her membership in the final four alliance. Nevertheless, her actions and words place her here. It is important to note that she was not a major player in the alliance and she was not required to vote against someone she liked until the merge. But Kelly leaves the alliance less than half way through the individual segment, and expressed concerns about its morality even earlier. Her choice to leave the alliance because of the toll it took on her conscience would have cost her her place in the final four had she not won the immunity challenges, and she knew it. Thus, at the most decisive stage, Kelly sacrificed rational strategy to the demands of ethics.

IMAGE (4 Players): The players in this category tended to value control of their public perception among the other players or the television viewers over their desire to win. It is important to distinguish this motive from the rational strategy of controlling perception. These players were not motivated to control how they were seen in order to win; rather, they wished to be seen as a certain kind of person for reasons extraneous to the competition.

Joel seems most motivated by his need to be the dominant male figure in his tribe. He consistently challenges the authority and competence of the other men and belittles the women. In discussions of tribal life or competition strategy he always asserts leadership, even when others were clearly better informed. And he was aware (or should have been!) of the effects of this behavior on the others. As Burnett points out in his analysis of the show, this was not at all a rational strategy: ABut Joel was so wrapped up in achieving alpha male status that he wasn=t paying attentionY Being alpha male wasn=t necessarily a good thing@ (Burnett, 2000, p.41).

Gretchen and Jenna both demonstrate a deep concern about the example they set for their children. Both appear to place this concern over any considerations of strategy, explicitly invoking their children in defense of their ineffective game strategy (Burnett, 2000, p. 113).

Gervase fits here as well as the ethical category simply because his ethic seemed to be one of image. He was deeply concerned with being seen as Aauthentic@ and Areal,@ thus refusing to act any differently than he might at home.

DUTY (3 Players): The players in this category see themselves as obligated to fulfill a higher purpose and restricted by their moral duties. Though they want to win the game, the parameters of duty restrict their choices. They will not take any action that is inconsistent with their higher purpose.

Rudy’s absolute loyalty to his teammates, once declared, and absolute adherence to his word, once given, make his behavior easy to explain. In fact, he was so committed to this unfashionable approach to life that he expected to be voted off the island early, and chose not to pursue any serious strategy (he didn=t even bother to learn the names of many of the players) (Burnett, 2000, p. 75). Though Rudy made it the final four and was part of the alliance, this choice cannot be attributed to a rational strategy. Rudy simply decided that Rich was the best leader,[14] declared his loyalty to him, and did what he was told. The fact that he chose the right person to follow does not indicate rationality, merely good judgment of leadership skills honed by experience in the military.

BB appears driven by a different form of duty, h commitment to the Protestant work ethic and merit based reward. His commitment to merit and work clearly hurt him among his teammates, annoying them to no end. Were his approach merely personal, he might have worked hard and shut up. But he also felt a duty to Awhip the others into shape.@ It was their resistance to his idea of duty that led to his expulsion.

Dirk also annoyed his fellow players with his consistent witness for his Christian beliefs. Though there were other Christians on the island, only Dirk saw his religion as the primary guide for his and others= behavior. As a result, he was reluctant to engage in the strategy and deception needed to win and unwilling to tone down his evangelism, which caused considerable anger.

EMOTION (2 Players): The two players in this category were motivated primarily by a desire for friendship and interpersonal satisfaction (negative or positive), to a degree that undermined their prospects for victory.

Susan was a member of the alliance, but her actions were governed by her desire to be friends with Kelly, and then her desire for vengeance once she felt betrayed by Kelly. Initially, Susan had planned to take Kelly with her to the final two, even though she recognized that would hurt her self-interest (Kelly was more popular, and thus likely to win more votes from the jury): ASo, yeah, if I don=t vote Kelly off I=m making a $900,000 mistake. But I don=t mind. I=d sacrifice that money in the name of friendship@ (Burnett, 2000, p. 177). Later, once hurt and betrayed, she actively tried to expel Kelly, ending any chance Susan had to make it to the final two since she knew Rudy and Rich would stick together. In both cases, her behavior was less rational than it at first appeared.

Kelly also seems driven by emotional need, at least if her psychological profile is accurate. According to the producers, Kelly=s decision to turn on Sue seemed motivated by a desire for Amaternal rebellion@ (Burnett, 2000, p. 217). This makes sense, given Kelly=s apparent need to publicly affirm her independence. Only anger could explain her action, since voting against Susan both reduced her chances of survival to the final two and cost her a vote if she did.

UNCLEAR (2 Players):

Sonja was not on the island long enough to permit any certain claims, though it is possible she would fall into the rational camp (she was part of the first brief alliance of Tagi women).

Ramona lasted longer than Sonja, but was sick for the early part of the game and never seemed to act on her own. The best that can be said of her is that she made no attempt to build any alliances nor did she attempt to avoid conflict; thus, she does not fall into the rational camp. Without more data it is impossible to place her more precisely.

RATIONAL (4 Players): These players pursued a generally rational course, despite some errors, and were motivated in both action and word by a desire for victory within the game. Some of them are also in other categories, but they meet a minimal criteria for rational behavior. These players are: Stacey, Susan, Rich, Rudy.

If the players are forced into a single dominant category, assigning each a primary motive, the various motives may be ranked and evaluated. Under this admittedly simplified but more parsimonious approach, we can see that the players break down a follows:

1. Ethics (6): Gretchen, BB, Greg, Sean, Collen, Jenna

2. Duty (2): Rudy, Dirk

3. Image (2): Gervase, Joel

4. Emotional (2): Kelly, Susan

5. Rational (2): Stacey, Rich

6. Unclear (2): Sonja, Ramona

CONCLUSIONS

The Survivor experiment proves, with unusual clarity for social science, the paucity of meaningful strategic rationality in human behavior. Under the most generous interpretation, only five players acted rationally. Standard social choice rationalizations cannot sufficiently explain this failure, since the conditions of the experiment were ideal and opportunities for social learning expansive. The players were obviously intelligent, the incentive structure clear, strategic behavior observable, and consequences for irrationality high. Despite this, the social and political actions of the players demonstrate the priority of non-rational factors in human decision making. Thus the Survivor experiment may be considered a conclusive demonstration of the weakness of social choice theory for predictions of collective behavior.

Beyond the evident refutation of social choice, two other conclusions emerge. First, the players= actions revealed the priority of moral beliefs over rational self-interest. This inference is of striking importance to the social choice debate. The rejection of moral beliefs as a primary motive for behavior is a central assumption allowing social choice to predict outcomes on a large scale. Behavioral predictions within the theory work only if individuals consistently place their own evident self-interest ahead of moral motives, since only rational self-interest may be operationalized unambiguously for large population. Moral beliefs, on the other hand, are less predictable, both because the codes themselves are often inconsistent or vague and because a person=s particular moral orientation might differ between word and deed. If complex moral concerns drive even a significant minority of people, social choice theory cannot predict behavior accurately. Social choice theorists are not, of course, oblivious to the presence of moral doctrines in peoples= lives. They merely claim that these codes either give way in the face of the clear dictates of self interest or, more commonly, that moral codes simply represent ex post facto justifications of strategically rational behavior. The Survivor experiment refutes this claim. Many of the players articulate moral doctrines clearly at odds with their self-identified material interest. That the players follow their moral codes to the detriment of their interest shows that they are not engaging in after the fact rationalization. That the players sabotage their prospects for victory, and do so with a full understanding of the consequences, proves the independence of morality from self-interest. This is not to say that the players are irrational. They do choose to follow the most efficient means to achieve their ends; however, for those whose ends are moral or ethical, social choice can not make predictive claims. The Survivor experiment shows that a substantial portion of the sample, the six motivated primarily by ethics and the two motivated primarily by duty, follow moral rather than self-interested strategies.

Second, social choice theory tends to identify either emotional confusion, lack of intelligence, or lack of opportunity for social learning as the explanations for the failure of the model to predict accurately. The Survivor experiment refutes these assertions, showing that normative concerns are the largest barrier to strategic rationality. Only two of the players seem motivated by emotional concerns, though these two did show how emotional confusion undermines strategic rationality. The point is that emotions were only a small barrier to strategic rationality. Moreover, many of the players least likely to follow the dictates of the rational model were also the most conventionally intelligent (Sean, Colleen, BB, Gretchen). In fact, many of the players who acted least strategically understood the Aproper@ strategy quite well, and articulated that understanding in interviews. Lastly, the duration of the experiment offered multiple opportunities for social learning. It is evident that the barriers to strategic rationality are not intellectual, emotional, or iterative, but normative.

One last anecdote confirms this claim. At one point during the team segment, Sean, Greg, and Gervase did form a voting alliance. Their purpose, however, was not strategic or self-interested. Their goal was to act cooperatively in order to Achange the moral tone@ of the game and help the players Aregain their ethical senses@ (Burnett, 2000, p. 139). This incident reveals the unpredictability of rational behavior where motives are more complex than simple utility maximization, defined as material interest. The position that human behavior is motivated at least substantially by moral rather then simple self-interested concerns is one social choice theorists have been at pains to disprove, since it renders their theory useful for only a small range of social phenomena. Survivor provides a substantial service if it helps chasten social choice theory and return it to its proper arena.

Appendix: Voting History

Each person receiving votes in a given week is listed in bold type, followed by the names of those voting for the person in parentheses. The person voted off is listed first and in italics.

VOTE#

1: Sonja (Rudy, Susan, Sean, Dirk); Rudy (Sonja, Stacey, Kelly); Stacey (Richard)

2: BB (Colleen, Gervase, Jenna, Gretchen, Joel, Ramona); Ramona (BB, Greg)

3: Stacey (Rudy, Susan, Sean, Dirk, Kelly, Richard); Rudy (Stacey)

4: Ramona (Colleen, Jenna, Gretchen, Joel); Colleen (Gervase, Ramona); Jenna Greg)

5: Dirk (Richard, Kelly, Rudy, Susan); Rudy (Sean): Susan (Dirk)

6: Joel (Colleen, Jenna, Gretchen, Greg); Jenna (Gervase, Joel)

7: Gretchen (Richard, Kelly, Rudy, Susan); Colleen (Sean); Richard (Colleen);

Susan (Gervase); Jenna (Greg); Gervase (Jenna); Rudy (Gretchen)

8: Greg (Richard, Kelly, Rudy, Susan, Sean, Jenna); Jenna (Colleen, Gervase, Greg)

9: Jenna (Richard, Rudy, Susan, Sean); Richard (Colleen, Gervase, Jenna); Kelly (Sean)

10: Gervase (Richard, Kelly, Rudy, Susan, Sean); Sean (Colleen, Gervase)

11: Colleen (Richard, Rudy, Susan, Sean); Sean (Colleen, Kelly)

12: Sean (Richard, Kelly, Rudy, Susan); Susan (Sean)

13: Susan (Richard, Rudy); Richard (Kelly, Susan)

Goes to tiebreaker, Susan and Richard cannot vote: Susan (Rudy, Kelly)

14: Rudy (Kelly, Richard); Richard (Rudy)

Only Kelly cast an actual vote, since she was immune.

Final Vote: Richard (Rudy, Susan, Sean, Greg); Kelly (Colleen, Gervase, Jenna)

Bibliography

Arrow, Kenneth (1963). Social choice and individual values. New York: Wiley.

Axelrod, Robert (1984). The evolution of cooperation. New York: Basic Books.

Becker, Gary S. (1976). The economic approach to human behavior. Chicago: University of Chicago Press.

Bohman, James (1989). Participating in enlightenment: Habermas's cognitivist interpretation of democracy. In M. Dascal and O. Gruengard (Eds.), Knowledge and politics (pp. 264-289). Boulder, CO: Westview Press.

Brams, Steven (1976). Paradoxes in politics: An introduction to the non-obvious in political science. New York: Free Press.

Buchanan, James and Tullock, Gordon (1982). The calculus of consent. Ann Arbor: University of Michigan Press.

Burnett, Mark with Dugard, Martin (2000). Survivor: The ultimate game. New York: TV Books.

Downs, Anthony (1957). An economic theory of democracy. New York: Harper and Row.

Dryzek, John (1992). How far is it from virginia and rochester to frankfurt? Public choice as critical theory. British Journal of Political Science, 22, 397-417.

Elster, Jon (1986). The market and the forum: three varieties of political theory. In J. Elster and A. Hylland (Eds.), Foundations of social choice theory (pp 103-132). Cambridge: Cambridge University Press.

Elster, Jon (1983). Sour grapes. Cambridge: Cambridge University Press.

Frohlich, Norman and Oppenheimer, Joseph (1978). Modern political economy. Englewood Cliffs, NJ: Prentice-Hall.

Goodin, Robert (1986). Laundering preferences. In J. Elster and A. Hylland (Eds.), Foundations of social choice theory (pp 75-102). Cambridge: Cambridge University Press.

Green, Donald and Shapiro, Ian (1994). Pathologies of rational choice theory: A critique of applications in political science. New Haven: Yale University Press.