Prisoner's Dilemma

13
Games Students Play: Incorporating the Prisoner's Dilemma in Teaching Business Ethics Kevin Gibson ABSTBJVCT. The so-called "Prisoner's Dilemma" is often referred to in business ethics, but probably not well understood. This article has three parts: (1) I claim that models derived from game theory are significant in the field for discussions of prudential ethics and the practical decisions managers make; (2) I discuss using them as a practical pedagogical exercise and some of the lessons generated; (3) more specula- tively, I suggest that they are useful in discussions of corporate personhood. KEY WORDS: classroom exercise, decision, experi- ential learning, game theory, negotiation, pedagogy, prisoner's dilemma, prudential ethics Introduction Game theory might be better called "strategy theory" because it deals with choosing options in the absence of sure knowledge about the actions of others. First, numeric values are assigned to possible choices. Then, a rational plan can be developed based on ranking the outcome scores. It may be applied whenever there is: (1) Kevin Gibson is an Associate Professor of Philosophy and Director of the Center for Ethics Studies at Marquette University. His research interests include business ethics, negotiation and alternative dispute resolution, with special interests in stakeholder theory, corporate personhood and moral accountability. His work has appeared in the Journal for Business Ethics, Business Ethics Quarterly, Teaching Business Ethics, Business Horizons, the Journal of Applied Ethics, Organization Science and the Harvard Negotiation Journal. a social interaction between two or more decision makers; (2) the outcome depends on the choices of all the players; and (3) every player has preferences among the possible outcomes.' Such games are extremely popular with social scien- tists and economists, among others, since they give simple ways for subjects to be placed in various sorts of interdependence and yield objective, quantitative data. In this paper, I wiU briefly outline a few forms of two-party games and show some of the dynamics they illustrate. Secondly, I will describe using a game in a classroom setting and some typical student reactions. Finally, I make the case for applying games in the classroom setting. I do this by suggesting there may be a valid form of prudential ethics (if not for humans, then for businesses), and by making the claim that, even if ethics has no prudential element, it is still important for us to recognize and appreciate the dynamics at work in game theory. Two-person games Foremost in the literature is the so-called "prisoner's dilemma" that matches personal gain against mutual benefit.^ The story goes that two prisoners are separated and given the chance to cooperate with the authorities by confessing. This sets up a two-by-two matrix, where both confess, both remain silent, or one confesses while the other keeps quiet. If both refuse to talk, then they both go free. If one confesses while the other doesn't, then the cooperator gets off lightly and the non-cooperator gets a heavy penalty. The Journal of Business Ethics 48: 53-64, 2003. © 2003 Kluwer Academic Publishers. Printed in the Netherlands.

description

Prisoner's Dilemma and Ethics

Transcript of Prisoner's Dilemma

Page 1: Prisoner's Dilemma

Games Students Play:Incorporating the Prisoner'sDilemma in TeachingBusiness Ethics Kevin Gibson

ABSTBJVCT. The so-called "Prisoner's Dilemma" isoften referred to in business ethics, but probably notwell understood. This article has three parts: (1) Iclaim that models derived from game theory aresignificant in the field for discussions of prudentialethics and the practical decisions managers make; (2)I discuss using them as a practical pedagogical exerciseand some of the lessons generated; (3) more specula-tively, I suggest that they are useful in discussions ofcorporate personhood.

KEY WORDS: classroom exercise, decision, experi-ential learning, game theory, negotiation, pedagogy,prisoner's dilemma, prudential ethics

Introduction

Game theory might be better called "strategytheory" because it deals with choosing optionsin the absence of sure knowledge about theactions of others. First, numeric values areassigned to possible choices. Then, a rational plancan be developed based on ranking the outcomescores. It may be applied whenever there is: (1)

Kevin Gibson is an Associate Professor of Philosophy andDirector of the Center for Ethics Studies at MarquetteUniversity. His research interests include businessethics, negotiation and alternative dispute resolution,with special interests in stakeholder theory, corporatepersonhood and moral accountability. His work hasappeared in the Journal for Business Ethics, BusinessEthics Quarterly, Teaching Business Ethics,Business Horizons, the Journal of Applied Ethics,Organization Science and the Harvard NegotiationJournal.

a social interaction between two or moredecision makers; (2) the outcome depends on thechoices of all the players; and (3) every player haspreferences among the possible outcomes.' Suchgames are extremely popular with social scien-tists and economists, among others, since theygive simple ways for subjects to be placed invarious sorts of interdependence and yieldobjective, quantitative data.

In this paper, I wiU briefly outline a few formsof two-party games and show some of thedynamics they illustrate. Secondly, I will describeusing a game in a classroom setting and sometypical student reactions. Finally, I make the casefor applying games in the classroom setting. I dothis by suggesting there may be a valid form ofprudential ethics (if not for humans, then forbusinesses), and by making the claim that, evenif ethics has no prudential element, it is stillimportant for us to recognize and appreciate thedynamics at work in game theory.

Two-person games

Foremost in the literature is the so-called"prisoner's dilemma" that matches personal gainagainst mutual benefit.^ The story goes that twoprisoners are separated and given the chance tocooperate with the authorities by confessing.This sets up a two-by-two matrix, where bothconfess, both remain silent, or one confesseswhile the other keeps quiet. If both refuse to talk,then they both go free. If one confesses while theother doesn't, then the cooperator gets off lightlyand the non-cooperator gets a heavy penalty. The

Journal of Business Ethics 48: 53-64, 2003.© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

Page 2: Prisoner's Dilemma

54 Kevin Gibson

worst mutual outcome is if both confess, andboth get sentenced. This could be representednumerically:

Confess Keep Quiet

ConfessKeep Quiet

- 1 , - 1-2 , +2

+2, -2+ 1, +1

If a rational prisoner had full information andtrusted the other party, he or she would realizethat the obvious choice with the best overalloutcome is to remain quiet. But in the absenceof such information and trust, and with therealization that the other prisoner also haschoices, the decision moves from maximizingbenefit to minimizing harm, in which case sheshould confess quickly. A shrewd, amoralprisoner might do the best for himself bypromising eternal allegiance and then defectingas quickly as possible. The desirability of payoffsfor A is first, mutual silence, but if not that thenconfessing while his partner stays silent, thenmutual confession, and the last thing he wants isto remain quiet while being implicated by hisbuddy. In game theoretic terms, confessing is thedominant strategy because he is better offchoosing it regardless of his partner's choice. Thesame is true, of course, for the partner.Paradoxically, if they are both rational, theywould both confess rather than stay quiet, whichdooms them to jail rather than walking free.

The rational and ethical become mixed whenwe use language like "cooperate" or "defect"since it assumes a relationship between theparties. There has been some discussion aboutwhether the background narrative alreadyassumes too much (such as the institution ofimprisonment and tangible benefits of confes-sion), but nevertheless, once described, thesedilemmas become, apparent in both our everydayand business lives.

For example, the New York Times (July 9th2002) reported that several airlines had increasedleisure fares by $20, only to reduce them laterin the day when it became clear that Northwestwould not do the same. Airlines cannot officiallycollude, but can certainly send signals into the

market and get a rapid response. If aU the airlineshad increased fares, then they would all reapmore profits, but if they act individually then themaximum benefit will go to whoever defectsfirst.

The same sort of dynamic occurs withcommon goods, such as international fishinggrounds. If they are being overfished, then itwould be reasonable for all the fishing fieets toagree on a system of self-restraint. In the absenceof strong regulation there is a temptation to breakthe agreement and do what is best for you alone,but that would only be successful if everyone elsecomplies with it. If aU parties make an agreementand then routinely ignore it, then everyone'slivelihood will eventually suffer.

The basic prisoner's dilemma as posed isrelatively simple. It has the pedagogical benefitof forcing players to choose from a finite set ofoptions and stimulating a discussion that dwellson the dynamics of the game without becomingbogged down in the particulars of the narrative.StiU, we should recognize the potential richnessand sophistication of these simulations.

Preference ordering

In a purely competitive zero-sum game[Defect/Defect, or D/D], a benefit for one sidewill necessarily lead to a loss for the other (thatis, +1 is always accompanied by - 1 , hence a 0result) and there will be no motive to cooperate.While these have some interest in negotiationstudies, they are not very fertile for businessethics. Similarly, in a pure coordination game, thedesired result coincides exactly [Cooperate/Cooperate, or C/C]. The prisoner's dilemma isactually a specific case of a mixed motive game,where sometimes it is rational to cooperate [C]and sometimes it is rational to defect [D]. Butnot all mixed motive games are prisoner'sdilemmas. Whenever there is a temptation in amixed motive game to defect when either theother cooperates [the "squealer" DC is better (>)than CC] or defects [DD > CD] (where mutualdefection is preferable to the "sucker" result),there will be a range of preferred outcomes,which have been given individual labels.^

Page 3: Prisoner's Dilemma

Games Students Play 55

DC>CC>DD>CDDC>DD>CC>CDDC>CC>CD>DDCC>DC>DD>CD

Prisoner's DilemmaDeadlockChickenStag Hunt

Deadlock. Deadlock occurs when the partnersdecide that they prefer mutual defection tomutual cooperation. For example:

Cooperate Defect

CooperateDefect

1, 13 ,0

0, 32 ,2

The cardinal issue is that although they wouldlike to gain, mutual defection is not a greatproblem and, in fact, gets them an outcomepreferable to cooperation. This occurs whenneither side really wants to cooperate - they areeither happy to stay where they are or to havethe other side change its stance, but aren't wiUingto budge at aU. Echoed in business and politics,this is a useful but limited lesson to dramatize.

Stag hunt. The dynamics of the stag hunt are thatthe individuals initially cooperate toward a goalthat none of them could attain alone. Along theway, though, one side sees a lame rabbit thatwould be an easy personal victory, but it wouldmean dropping out of the mutual enterprise.

Stag Rabbit

StagRabbit

3,32, 0

0 , 21, 1

Mutual cooperation will yield the greatestbenefit, and the temptation to go it alone onlyemerges if there is a feeling that the other sidewill do the same, or, if the group is sufficientlylarge, it is likely that there will be someone whowill go against the grain. In effect, it demon-strates the problem with poor teamwork, sinceeveryone is better off doing their own thing ifthere is not complete unanimity of purpose.

Chicken. The challenge in Chicken is to be asclose to peril as possible, especially in competi-tion with someone else. It is "brinkmanship"where the dare is to go as near the edge aspossible to show the opponents you are wiUingto risk all in order to beat them. The classic caseis where American teenagers in the 1950s woulddrive cars at each other on a coUision course. Theloser swerves off first, allowing the winner tokeep going straight. One way to win the dare isprobably to make a big show of throwing thesteering wheel and an empty whisky bottle outof the wfindov^ at the beginning of the run, thusdemonstrating full intent and a certain reckless-ness. Honor may be lost if one or both sides backdown from the challenge, or swerve very early.Yet the worst mutual outcome is for neither sideto swerve at all, and the zero in the payoff scalecould easily be a very large negative number.

Swerve Straight

SwerveStraight

2 ,23, 1

1,30 ,0

In business terms, this may be seen as theequivalent of playing "hardball" or bluffing. Ifthe posture succeeds, it may have handsomerewards. Conversely, though, if the bluff is calledor a deal falls apart, then we are left with nodeal (and potentiaUy no future deals) in a situa-tion where it would be appropriate. Adversarialpositions in coUective bargaining have often takenthis form, even to the point that businesses haveceased operations when stiU viable.''

Asymmetrical dilemmas

Again, for pedagogical simplicity it is useful tohave payoffs that are symmetrical and know^n tobe so to both parties. The payoff box nominallycaptures the utility for each side, but in realitythere will be a wide variation in both the valueof the payoff amount (a rich person could moreeasily walk away from a minimal benefit than apoorer one, for example) and factors like timesensitivity and competitiveness. There are cases

Page 4: Prisoner's Dilemma

56 Kevin Gibson

where it is instructive to see that parties may havedifFerent preference orderings.

One such case, noted by Poundstone, is theBiblical dispute over the maternity of a child.^Two women claim the same child, and KingSolomon suggests splitting the child in two. Thechoices for the women are to stand firm (defect)or give in (cooperate). The real mother wantsto stand firm on her claim, and have the imposteracknowledge her deceit. But her worst outcomeis chicken, since if neither backs down, then thechild dies. On the other hand, the imposter isplaying a deadlock game: she wants to keep thechild, but would rather the child were kiUed thanbe handed to the real mother. Because theimposter has the power that comes with indif-ference to the outcome, she can effectively winby bullying - in any case, she can stand firm withrelative impunity. When faced with this posture,the only rational thing for the real mother to dois to capitulate. In the story. King Solomonrealizes that the true mother is the one whowould give up her child to save it, and he judgesaccordingly. For our present purposes, gametheory shows that there are rational decisions tobe made, and they rely criticaUy on an assessmentof our preferences.^

The prisoner's dilemma in the classroom. Thequestion at hand is whether using game theoryhas any bearing on teaching ethics, specifically,business ethics. My experience is that it can bean important tool which students find to bechallenging and enjoyable.

I believe that there are at least three ways inwhich it can be pedagogically useful: First, byraising awareness, with the assumption that whenpeople realize the dynamics of the dilemma, theywill become more aware of how pervasive it is,and take a more rational approach when theyconfront it. Naturally, posing the issue will alsospark challenges of the is/ought variety followingHume, who famously noted that the facts aboutour behavior do not necessarily lead to author-itative prescriptions.''

Secondly, it puts the issue of whether there canbe a prudential ethics front and center. One ofthe clear lessons of the dilemma is that one cando well by doing good. People realize that acting

"ethically" will lead to long-term benefits forthemselves. Essentially, this is a lesson in benignself-interest, and thus has normative implications.Some would contend that prudence plays no partin ethics, but others will assert that it is the veryfoundation of business ethics, since businessesneed to be convinced that there will be somepayoff for their actions.

Thirdly, the dilemma is descriptive since itdemonstrates the sort of position in which peopleor businesses find themselves. It is then assumedthat they will act like rational prisoners. This isthe sort of claim made by David Gauthier (1970)about ethical action in general, or by PeterFrench (1991) and Peter Danielson (1992) aboutthe moral behavior of business. The accuracy ofthe analogy to the dilemma will also give indi-cations about how parties with confiictinginterests will react, as well as what is their best,rational strategy to maximize their individual ormutual benefit.

A typical case. Games in the classroom will sufferfrom being artificial and compressed. Thus, thereis a tendency for students to treat them as gamesrather than exercises. It is of great importance atthe outset to let participants know that theyshould take the exercise seriously, that theyshould bargain in earnest, and that they will beassessed on how well they do. They are brokeninto teams, and teams are paired off, so that theyrecognize that each individual team is dealingwith only one other person throughout, and theactions of any other participants are immaterial.The intragroup dynamics represent real decision-making bodies, such as a board, and provide adiscussion point at the end of the exercise. Ifspace allows, the two teams may be separated.Everyone is then provided with a narrative thatwill set a symmetrical payoff matrix. Words like"competition" and "opposing party" are bestavoided since they tend to impose strategies thatmay not arise spontaneously. Generally, theinstructions tell participants to focus specificallyon maximizing their own outcomes.

A typical narrative might involve, say, petrolstations operating on opposite corners of anintersection. Both are franchise operations andreset their prices each week. Customers are price

Page 5: Prisoner's Dilemma

Games Students Play 57

conscious and fickle, and the profits dependdirectly on the price the station sets. If bothstations have identical profits, then they will bothmake $1200 profit every week. If one station cutsits price and the other doesn't, then the increasedsales of the lower priced petrol will bring in$1600, and the higher priced station will makeonly $400. But if both stations cut prices, eachstation wiU only net $800.^

The dilemma is usuaUy given in iterated form,so that there are repeat encounters with the sameside with known payoffs. This leads to partici-pants deciding not only on an initial strategy, butalso on how they wiU react to moves by the otherside. They have to decide on an opening posture:If both sides cooperate, then they have done weU,but need to sustain their relationship against anever-present temptation to defect for a "quickbuck." If one or both sides fail to cooperate, thenthere is a real problem of how they can move tocooperate. Presumably, if we were, as Rawls(1971) suggests, rational and risk averse, thenonce we realize the dynamics involved we wouldidentify and adopt an appropriate strategy oflong-term cooperation given the payoff matrixabove. In my experience, this rarely happens.Sometimes, although rarely, participants fail tonotice that cooperation over repeated plays is thebest approach. But even when they determine anoptimal strategy the parties routinely defect oneach other. When allowed to negotiate theyroutinely lie about future intentions and indi-vidual short-term gains are taken as the gaugeof success, even if the instructions explicitlydemand long-term maximization. After anumber of rounds, the class is reassembled andtotals tabulated for each pair of teams. Given thenature of a prisoner's dilemma, even cooperativegroups have been tempted to defect. Hence it israre for groups to have maximized their poten-tial returns, and often they have migrated tomutual defection. This is highlighted when thereare a finite number of interactions. Why does thishappen?

One reason could be akrasia — weakness of thewill. We may intellectuaUy realize what is best,but our emotions or greed take over. While thiscould account for some of the results, it isperhaps too simplistic an explanation.

Another reason could be a version of the"surprise test" paradox. There, students are toldthat there will be a surprise test by Friday duringthe next week. They reason that it wouldn't bea surprise if it came on Friday itself, so it wiUhave to be between Monday and Thursday. Butthen it can't be held on Thursday, since thatwouldn't be a surprise now either, and so on,until it is impossible to have a surprise test at all.In an iterated game played by students, theyrealize there wiU be a finite number of rounds.In the last round it is irrational to cooperate,since they can achieve maximum individual gainswithout reaping the penalty in further play. Butif both sides think alike, then they believe thatneither will cooperate in the final round, andtherefore the appropriate place to begin defec-tions is in the penultimate round, and so on, untilit makes sense to defect as quickly as possible.

A further possibility arises from what might becalled the " 'scape whipping" problem. (Hamlet:"Use every man after his desert, and who should'scape whipping?") Players reason that they arefallible human beings who, although trusting,may not always be trusted. Extrapolating fromtheir own temptations, they project these ontothe other party, which leads them to take theleast vulnerable path.'

There is often a "signal/noise" issue, too, sinceplayers have little substantive information aboutthe other's actions except the result with whichthey are presented. Players will interpret theother's actions in various speculative ways andmay weU excuse their own actions by identifyingsuperficial signals ("he was smiling at me whenwe talked") as substantive and pernicious.

Non-cooperation is also the stance that makesplayers least vulnerable, and they often report thatthey are willing to forsake gains because theydidn't want their opponent to take advantage ofthem. This is reminiscent of a psychological testwhich tells participants that they can have $10between them if they can agree on a divisionwithin a rainute or two. If someone is offeredsomething other than a 5/5 split - say, 9/1, thenthey will often refuse to agree, althoughrationaUy they should take something rather thannothing.'"

Some participants report that once they were

Page 6: Prisoner's Dilemma

58 Kevin Gibson

ahead, they needed to stay ahead, and in so doinggave up greater gains. This effect has beendescribed as the "Red Queen Paradox."" InAlice's Adventures Through the Looking Glass, theRed Queen grabs Alice and runs with herconstantly yelling "Faster! Faster!" It takes all therunning you can do to keep in the same place.Because everyone else is running, the Queenfears that if she stops, then she will fall behind.However, if everyone stopped running, then shewould, at worst, remain where she was. Utilityis wasted in the competition; it has no benefitother than providing an institutional contextwhere everyone is afraid that they will be hurtif they stop. Another common result is thatparticipants ignore the injunction to concentrateon their own profits, and instead becomeobsessed with merely doing better than others.'^

It is useful to play out the game, if only tosee the dynamics that players get into and howthese reflect their "real life" experience. Giventhe setup of a prisoner's dilemma, cooperationcan be a very rewarding experience. However,players who start noncooperatively (and it is aninteresting discussion point to explore why theydo so) and those who move to noncooperationwill find it difficult to change their posture.Typical debrief questions might include:

• How the starting price affected the finaloutcome.

• How emotions of both parties contributedto the process.

• How information is critical, and how it iscollected and assessed.

• How easy it is to get involved in an"auction dynamic".

• Whether you can trust someone whodoesn't have your interests at heart.

• How rational the group was in making aninitial bid and in its overall strategy.

Some lessons from the classroom. One way tounderstand why the dilemma is actually adilemma rather than a rational exercise is torecognize that one of the key variables in deter-mining one party's actions with regard to theother is the degree of information and trust. Ifboth sides had instant access to the other's dis-

cussions, or knew that failure to cooperate wouldresult in prolonged torture, then choice is effec-tively foreclosed. On the other hand, the com-munication of information and bestowal of truston others is not a purely rational action, andindeed is probably best considered value-laden.This opens the door to non-rational factors suchas emotion and intuition.

A standard criticism of using prisoner'sdilemmas in business ethics is that it is an exercisein costs and benefits, whereas ethics is theprovince of the normative, that is, the "is/ought"assertion that factual claims ("stealing is wrong")do not necessarily lead to normative ones ("youshould not steal"). Some, like Robert Solomon(1999), dismiss game theory as missing the pointbecause it looks at individual benefits rather thanfocusing on the development of personal andcorporate virtues. I believe this claim is over-determined: Even if one does not believe inprudential ethics, it seems there are significantlessons to be learned from game practice — thatis, what typically happens when groups playiterated dilemmas or other games. Some partic-ipants wiU take an openly "moral" stance and saythat they would not go against their word innegotiations because no reward would justify alie. Still, many do compromise their everydayethics and act contrary to what they usuallyespouse. Some of the psychological forces atwork present interesting discussion points, sincethey promote behavior at odds with whatparticipants know they should do. (Perhaps an"ought/is" problem?). Heuristics and biases maylead individuals to buy products they don't want,and temper their conscious awareness of what iscorrect. In a prisoner's dilemma they often comeinto play, and form the basis for discussions ofthe interplay between the rational, the norma-tive, and infiuenced behavior. Notably, the gameis best played in groups where the participantsget to discuss their strategy along the way, andthe intragroup dynamics are worth analyzing toexamine compliance and excuses within a teamsetting.

Discussion points. A debrief of the game will notonly involve the internal dynamics of how playersselected and developed their strategies, but also

Page 7: Prisoner's Dilemma

Games Students Play 59

some background assumptions and conceptualissues.

Evidently games have numeric values as theirstock in trade. This will bring up the questionof w^hether such valuations can be made, and ifthey have any validity. So, for example, studentsmay question if it makes any sense to assign valueto liberty or various aspects of welfare. A standardresponse from the business school would be thatnot only are they valid, but we do them all thetime. We make trade-offs about convenience andmoney (fiying or taking a train, perhaps) and wehave a system of monetary compensation foraccidents and injuries. If students suggest that,say, we cannot assign value to songbirds orsunsets, then they have to explain a sense of valueat all, especially when others might say that thereis, in fact, a figure that people would agree topay to forgo having pollution deadly to wildlifeor dense smog obscuring the view.''' In short,students will face the countervailing claims ofinstrumental and intrinsic values. As an associ-ated point, they will have to take a position onwhether, for example, loyalty and trust can betaken as commodities, that is, marketable with anexchange value.

Certainly since Plato, and emphatically withHobbes, there has been the question of whetherone should do good out of self-interest. Some,like Tom Beauchamp (1982) have strongly arguedagainst prudential ethics, saying that, howeverlong-term or distant, any action motivated outof self-interest will not qualify as moral. I willnot argue these claims at any length, except tosay that I think prudence does play a role inbusiness ethics, and suggest that there may stiUbe significant lessons from game theory even fordeontological or virtue-based approaches tobusiness ethics.

Economists often equate rationality withprudence, and games give students the oppor-tunity to test that assumption. For example,Kantian ethics represents rationality par excellence,since to become a moral agent one acts accordingto maxims that can be consistently generalized,and individual welfare maximization is clearly notpart of the picture. The Kantian faced with theinitial prisoner's dilemma is confronted with thechoice between (a) both prisoners confessing and

(b) both keeping quiet, since the maxim has tobe generalizable to the other prisoner. It isapparent to students that choice (b), althoughrational, may not be prudent. Indeed, if one hasan agreement with the other prisoner to keepquiet and trusts her to do so, it is prudent (at leastin a single play-game) to betray her.

An assumption embedded in the game is theneo-classical economist's view that behavior ismotivated by individual self-interest. At the endof the game, though, "winners" who have actedout of rampant self-interest are rarely congratu-lated by their peers and often are in the positionof defending their actions, which initiaUy appearsparadoxical. One explanation that is worthfloating with students is whether the economistis correct, and whether there are other individualor cultural assumptions, such as the nature ofrationality or intuition which may be embeddedin the way the game is played out.

The critical element for a Kantian in the gameis consistency in terms of the reasons for action.That is, either (a) or (b) would be ethical if theplayer was sure of what the others' choice wouldbe in the same circumstances. Choice (b) appearsthe better policy for both prisoners, but it cannotbe derived from Kantian deontology. The notionof "better" here is a function of rule-utilitari-anism thought of as a criterion for optimal policy.Students typically conclude that this kind ofrule-utilitarianism may be prudent for individ-uals in the long run. This may turn out to be thecase, but teachers should emphasize that thisresult is an empirical fact rather than a norma-tive prescription. EssentiaUy part of the tensionthat students feel at the end of a game and a keylesson is that whether they are drawn to optimaloutcomes in a utilitarian sense or rational policyfrom a Kantian view, neither wiU be necessarilyequivalent to individual prudence.

The game demonstrates that prudence doeshave a motivating role in ethics for many peopleand, I suggest, the majority of businesses.'•* Thepayoffs may be more or less immediate.Interestingly, though, they may also be hypo-thetical: We can think of a person who leads anexemplary life, but is motivated by the thoughtthat there is some celestial leger where everythought, word and deed is recorded. From our

Page 8: Prisoner's Dilemma

60 Kevin Gibson

earthly perspective, the person acts like a saint,and is indistinguishable from a deontologicalaltruist. Now, it is plausible that our saint ismistaken, and it turns out there is either nohereafter or no leger. But it does not matterwhether, in fact, the eternal rewards exist. Whatdoes matter is that the individuals involved havefaith that there is a reward, and act accordingly.The question is whether we should say theperson was unethical because they were moti-vated by the (unproven) payoff.

Using the same thinking, the claim that busi-nesses doing good wiU result in them doing weUis unscientific. If they do good actions if, andonly if, there is a sure reward, then perhaps wecould discount their actions as not fully ethical.However, even in that case, we are faced withthe sum of good in the world being increasedwhich is preferable to the status quo. However,I believe that most cases are more like my hopefulsaint: the payoff is uncertain for businesses. StiU,what matters is that they have a belief in thepayoff and act accordingly. Thus, I contend thatalthough our judgments about the morality ofbusiness may be tempered by perceived prudence,it is not a mortal blow to them being ethical ataU.

If one grants a prudential element in business,then there are obvious applications to businessethics, since there is a rational strategy based onthe utilities and iteration of the game.

Tit-for-Tat. In iterated games, (that is, whenrepeated with a known payoff matrix) variousstrategies emerge, and these have been testedagainst each other with human subjects and incomputer tournaments. When players have thepossibility of rewarding each other throughcooperation and hurting each other throughdefection, the best strategy has turned out to bethe so-caUed Tit-for-Tat. Robert Axelrod (1984)called for programs and then let them run inprolonged tournaments (over 120,000 moves).The simplest program, Tit-for-Tat, was also themost successful in terms of survival. It begins bycooperating and then reciprocates the move ofthe other party. In effect, it rewarded coopera-tion, but did not punish defection with anythingother than a penalty of parity. Axelrod summa-

rizes its success as being due to it being nice(never being the first to defect), retaliatory,forgiving, and clear. It is interesting to note thatthe simple program, uninfluenced by perceivednon-verbal messages, desire for vengeance, or theneed to out-do the other party at whatever thecost, routinely out-performs its human competi-tors.

One significant lesson of Tit-for-Tat is that inan iterated setting, cooperation is the preferablestrategy and defection wiU be punished. If busi-nesses want to survive and grow in a competi-tive environment, cutting ethical corners to getinstant gain will lead to long term problems andsub-optimal payoffs. While that is often held tobe intuitively true, there is a body of empiricalresearch which confirms it. Currently, there area number of writers in business ethics whocontend that the admittedly artificial frameworksof game theory in general and the prisoner'sdilemma in particular point out importantfindings in normative ethics. As ManuelVelasquez (1996) says

The research on prisoners' dilemmas and socialdilemmas shows that ethical behavior is moreprofitable and more rational than unethicalbehavior in terms of both the negative sanctionson unethical behavior when interactions withstakeholders are iterated, and the positive rewardsof habitually ethical behavior when stakeholderscan identify those who are predisposed to beethical."^

Corporate personhood. It is, of course, possible toplay a prisoner's dilemma with a computer, muchlike a simple chess program. Moreover, if youcannot see the other party, there is nothing todistinguish a computer from a human player. Inthis section I wiU suggest, foUowing French andDanielson, that businesses may act as indepen-dent moral agents, and can react to the dilemmain a rational way.

Individuals are the basic units of moral agency,and an institution is composed of many individ-uals. Certainly the individuals may be workingtogether (that is, coUectively) and they may makedecisions which are moral or immoral. French'sclaim is that individuals may not be the only

Page 9: Prisoner's Dilemma

Games Students Play 61

moral actors, though. A corporation may have apurpose, culture and intentionality that is distinctfrom the people that make it up. Thus a business,especially if it is large and established, will havean almost organic identity which wiU survive thecoming and going of any particular individual.Thus it makes sense to say there is "an IBM wayof doing things" or that there is a "FedEx spirit."There are shared understandings, and individualsbecome acculturated or leave. This is not to saythat individuals can abdicate personal responsi-bility, even acting in a professional or corporaterole. French believes that individuals can beculpable if there is a proper causal chain.However, there may be cases w^here no particularindividual can be isolated, and then we have toadmit that the responsibility evaporates, or thatthere is another moral entity involved. So if thereis systematic racism displayed by Denny's, arestaurant chain in the U.S.A., it may be thatthere is no written policy or dictate. Instead,v^hat we find is a culture of shared understand-ings which no one person has authored, buteveryone in the company knows and lives by. Irithat case French would suggest the corporationis acting just like a moral agent. While not com-pletely endorsing French's project on corporatepersonhood, I do want to accept as plausible hisnotion that corporations, and not just people, canbe moral actors. Claims of the sort "Exxon wasmorally responsible for polluting the environ-ment", or "Ford was to blame in the design ofthe Pinto." are perfectly meaningful if seen in thisway.

French recounts a conversation he had with arepresentative of Mazda's leasing division whocalled him about an alleged late payment. Hecomes to realize that he is not only talking toan individual, but she is doing nothing butreading from a flow-chart script provided by thecompany: he might weU have been talking to amenu driven voice synthesizer. Hence the deci-sions she (apparently) makes are the product ofcorporate policies and procedures that are trulyorganizational; since individuals have very littlepersonal discretion. So, for the sake of argument,we have entities, notably corporations, which canhave a tremendous impact on our lives, andwhich may be thought of as making decisions of

a moral kind. Admittedly, of course, the policiesare generated by humans, and there is a level ofdiscretionary judgment at a high executive level.Nevertheless, on this view corporations areproperly viewed as having a culture, dispositionor intentions; and these can have moral content.For our purposes, corporate decision-makingtends to be rational. Peter Danielson has taken astrong view along these lines. He says:

I find that . . . firms are suitable rational moralagents but humans perhaps are not . . . They havemorally crucial capacities that individual peoplemay lack. For example, firms . . . may be consti-tuted by public decision procedures, some of whichcommit them to various courses of action contraryto their interests, yet which are open to interest-based change."

That is, working from enlightened self-interest, corporations have realized that thebusiness environment is effectively an iteratedprisoner's dilemma, and hence not only curbtheir behavior, but do so in a transparent andpredictable way. French summarizes his versionof the argument by saying:

It is empirically demonstrable that most corpora-tions do adopt constrained strategies in the verymarketplace that for about three centuries we havebeen told is governed only by unconstrainedmaximization. They do so — some might say theyhave to do so — because the constrained approachessimply fair better. Constrained strategies are morelikely to produce market success than uncon-strained ones. That is what the empirical tourna-ment evidence in game theory indicates . . . I amcertainly not saying that corporations are auto-matically ethical, or that by being involved in themarketplace they cannot avoid becoming ethical.That would be grossly out of line with the facts.Instead, I am saying that corporations, by their veryconstitutive structures, evidence the functionalcapacities necessary to be ethical. The evidenceregarding humans is far more ambiguous.'*

If he is correct, then this is important news forteaching business ethics. The airlines notedearlier are acting like players in a prisoner'sdilemma game, but importantly they can all beseen as acting like programmed computer playersor players with artificial inteUigence. There will

Page 10: Prisoner's Dilemma

62 Kevin Gibson

be verifiable strategies for business in acting andreacting to situations in the marketplace. French,in dealing with Mazda, should recognize thatthey will respond in certain ways to his initialmoves as if they were repeatedly playing roundsin a game; and the appropriate response shouldprobably be strategic rather than intuitive. As henotes, this is probably more predictable thandealing with humans, and even offers the hopethat an organization could be ethical even if itscomponent human parts are not.

I believe that his example of the woman-as-voice-for-the-corporation is telling, and true.Even though we could grant executive over-ride,many of our (human) interactions with a businessare with people who are agents of the business.Such agents are not free to make their own(personal and individual) decisions, but rathermust respond according to the corporate script.Again, let us admit that individuals in role stillare autonomous agents who are ultimatelyresponsible for their own actions, and at somepoint they may choose to quit or blow thewhistle if actions seem egregious. However, mostworkers necessarily have to pick and choose theirbattles, and not everything rises to the level of aquittable issue. Thus, I should expect that mostpeople working for organizations will behaveaccording to a set of policies which may havemoral consequences. A university teacher may berequired to report behaviors such as plagiarismor harassment that do not exactly comport withhis or her own beliefs, for example. The gov-erning "morality" of the institution may be anorganizational set of policies, or even theunwritten and unarticulated culture of howthings are done at work. These understandingsmay be independent of any particular person andyet guide how individuals work in a corporatesetting. It is arguable, naturally, that individualsought to take moral responsibility for theiractions in an institutional setting. StiU, if manypeople do, in fact, abdicate their personal choicesin the workplace, then the corporate personhoodpoint remains: We are confronted with powerfulentities (corporations and institutions) whichhave the ability to help or harm us, and theytypically act in predictable self-interested wayswhich can be modeled by games.

What I want to take from French is the pos-sibility that there may be moral actors apart fromparticular people, and if so it is incumbent onus to realize how a corporate entity is likely tobehave. It seems that they may well be rational,and act in risk averse ways when faced withdecisions. Although I may personally be a virtuetheorist who scorns prudential ethics, I am likelyto encounter both individuals who thinkdifferently, and, significantly, a whole class ofinstitutional entities who are probably governedby the rational dynamics of game theory.

Pedagogical implications. Game theory is an artifice- a simple analogy for representing reality, andit has its fiaws. Yet despite its shortcomings, itoffers teachers of business ethics valuable service.Going through a prisoner's dilemma (or similarmatrix based game) has several potential benefitsfor the teacher. Variation of pedagogical approachhas value in altering the classroom dynamic fromone of lecturer and lectured-to. It wiU appeal toa wide range of audiences on a variety of levels.It also has sufficient content and material reflec-tion to challenge different types of learners.

For example. Honey and Mumford (2000)suggest that there are likely to be several kindsof student learners. "Activist" students learn bestwhen they have fresh tasks to confront and con-stantly seek new experiences. In contrast, the"Reflector" needs time to evaluate material."Theorists" want to connect what they learn toa wider context and a background of theory. Incontrast, a "Pragmatist" is keen to put theoryinto practice. If a classroom has a mix, then itwould be appropriate for the teacher to use sometraditional and some innovative techniques.These games offer a chance for different studentsto bring their strengths to the fore: Some willbe more attracted to the task involved, whereasothers will want to understand the dilemmas interms of the wider business world. It is visceral,in the sense that students get to have an imme-diate and sometimes passionate experience ofliving through a dilemma that mirrors thedynamics of everyday business life. The problemsfound in the prisoner's dilemma and its variantsare realistic, even if the setting is contrived, andtherefore students can extrapolate from their

Page 11: Prisoner's Dilemma

Games Students Play 63

findings. Depending on the teacher's objectivesin the particular course, there are opportunitiesfor both task completion, observation and reflec-tion on a conceptual level.

One of the significant challenges to teachersin business ethics, in my experience, has beento make a bridge between theory and (whatstudents perceive as) the "real world." Theprisoner's dilemma allows them to feel thetension between the intellectual norms theyaspire to and the temptation to put personalwelfare first. There is a strong folklore, at least,about the dominance of profit and paramountself-interest. Although these games are stylized,they do offer an alternative view involvingcooperation and mutual benefit, a view whichemerges from the student's own actions ratherthan being imposed or prescribed. The evidencefrom computer tournaments and internationaltreaties (such as limits on fishing) gives someempirical backdrop to the dynamics involvedwhich is attractive to many students. Students arenot being asked to "reinvent the wheel" sincethere is a substantial amount of quantitative andqualitative research on which they can draw.

Such games by themselves are unlikely toproduce much change in the students. Howeverwhen they are used in conjunction with othermaterial they can be a powerful tool to promotethe integration of an ethical dimension in thedecisions of future managers.'*

Notes

' See Coleman (1982), p. 113.^ The story of the prisoners is attributable to AlbertTucker who used it to explain a game described byMerrill Flood to a group of psychologists. Theoriginal game is from Merrill Flood "SomeExperimental Games" (1952).•* This discussion draws on Poundstone (1992).'' An example could be the Thatcher/Scargill con-frontation over the future of mining in England in themiddle 1980s.^ 1 Kings 3: 16-28.* See Jon Elster on Adaptive Preference Formationwhich suggests that our preferences can be altered byour perception of how possible they appear to beingrealized. "Sour Grapes — Utilitarianism and the

Genesis of Wants", in Utilitarianism and Beyond, ed.B. Williams and A. Sen, p. 219.' David Hume, Enquiry Concerning the Principles ofMorals, Section 1.' This is a sketch of a prisoner's dilemma presentedin Murnighan Bargaining Games (1992), p. 70.' This observation has empirical backing: Kelley andStahelski (1970) found that cooperative people believethere are cooperative, neutral and competitive peoplein the world, whereas people with a more neutraldisposition felt there were neutral and competitivesorts, and competitive individuals felt that everyoneelse was exclusively competitive too.'° These division problems are known as "ultimatumgames" and are covered comprehensively inMurnighan (1991)." See Khalil (1997).

'̂ Martin Shubik (1971) published a paper whichdescribes a "dollar auction" which is instructive inportraying how people get wrapped up in the psy-chological urge to win rather than rationally lookingat their substantive gains. A sum of money (adjustedto be non-trivial) is auctioned off, with the usual rulesexcept the second-highest bidder also has to pay, butgets nothing for it. As long as there are two or morebids over half the sum, the auctioneer wins (e.g., 50cents and 55 cents). The critical learning momentcomes when the bids near the sum (98 and 99 cents),where the second bidder realizes that he or she isforced to lose 98 cents or bid a dollar or more, atwhich point the bids routinely go over the sum atauction.

" Daniel Kahneman suggests proxy values can befound out relatively easily by asking simple questionsof the order "If you broke your leg and won thelottery in the same week, how much would you haveto win to think it wasn't such a bad week after all?"Personal conversation with the author, 1997.'•* See, for example, the work of the psychologistLawrence Kohlberg (1981), who claimed that mostpeople were at either a "preconventional" or "con-ventional" ethical state.'= Velasqez (1996) p. 202."̂ Danielson (1992), p. 198." French ( 1991), p. 86.

'* I am grateful to the editor and two anonymousreviewers whose comments have improved the article.

References

Axelrod, R.: 1984, The Evolution of Cooperation (BasicBooks, New York, NY).

Page 12: Prisoner's Dilemma

64 Kevin Gibson

Beauchamp, T.: 1982, Philosophical Ethics {McGmw-HiU, New York, NY).

Coleman, A.: 1982, Cooperation and Competition inHumans and Animals (Van Nostrand Reinhold,Wokingham, U.K.).

Danielson, P.: 1992, Artificial Morality: Virtuous Robotsfor Virtual Games (Routledge, London).

Flood, M.: 1952, 'Some Experimental Games',Research Memorandum RM-789, RANDCorporation, Washington, DC.

French, P.: 1991, Corporate Ethics (Harcourt, NY).Gauthier, D. (ed.): 1970, Morality and Rational Self-

interest (Prentice-Hall, Englewood ClifFs, NJ).Honey, P. and A. Mumford: 2000, The Learning Styles

Helper's Guide (Peter Honey Publications,Maidenhead).

Hume, D.: 1999, Enquiry Concerning the Principles ofMorals, edited by T. Beauchamp (OxfordUniversity Press, Oxford).

Kelley, H. and A. Stahelski: 1970, 'Social InteractionBasis of Cooperator's and Competitor's Beliefsabout Others', Journal of Personality and SocialPsychology 6, 66-91.

Khalil, E.: 1997, 'The Red Queen Paradox: A ProperName for a Popular GAme', Journal of Internationaland Theoretical Economics 153, 411—415.

Kohlberg, R.: 1981, The Philosophy of MoralDevelopment (Harper & Row, San Francisco).

Murnighan, K.: 1991, The Dynamics of BargainingGames (Prentice-Hall, Saddle Brook, NJ).

Murnighan, K.: 1992, Bargaining Games (Morrow,New York, NY).

Poundstone, W: 1992, Prisoner's Dilemma (BantamDoubleday, New York, NY).

Rawls, J.: 1971, A Theory of Justice (HarvardUniversity Press, Cambridge, MA),

Sen, A. and B. Williams.: 2002, Utilitarianism andBeyond (Cambridge University Press, Cambridge).

Shubik, M.: 1971, 'The Dollar Auction Game: AParadox in Non-cooperative Behavior andEscalation', Journal of Conflict Resolution 15,545-547.

Solomon, R.: 1999, 'Game Theory as a Model forBusiness and Business Ethics', Business EthicsQuarterly 9(1), 11-29.

Velasquez, M.: 1996, 'Why Ethics Matters: A Defenseof Ethics in Business Organizations', Business EthicsQuarterly 6(2), 201-222.

Marquette University,Department of Philosophy,

P.O. Box 1881,Milwaukee WI 33201-1881,

U.S.A.E-mail: Kevin. [email protected]

Page 13: Prisoner's Dilemma