The emergence of hyper-altruistic behaviour in conflictual situations
TThe emergence of hyper-altruistic behaviour in conflictualsituations
Valerio CapraroSeptember 18, 2018
Center for Mathematics and Computer Science (CWI)1098 XG, AmsterdamThe [email protected]
Abstract
Situations where people have to decide between hurting themselves or anotherperson are at the core of many individual and global conflicts. Yet little is knownabout how people behave when facing these situations in the lab. Here we reporta large experiment in which participants could either take x dollars from anotheranonymous participant or give y dollars to the same participant. Depending on thetreatments, participants could also exit the game without making any decision, butpaying a cost. Across different protocols and parameter specifications, we provideevidence of three regularities: (i) when exiting is allowed and costless, subjects tendto exit the game; (ii) females are more likely than males to exit the game, but onlywhen the cost is small; (iii) when exiting is not allowed, altruistic actions are morecommon than predicted by the dominant economic models. In particular, againstthe predictions of every dominant economic model, about one sixth of the subjectsshow hyper-altruistic tendencies, that is, they prefer giving y rather than taking x > y . In doing so, our findings shed light on human decision-making in conflictualsituations and suggest that economic models should be revised in order to take intoaccount hyper-altruistic behaviour. Part of the secret of the enormous success of human societies is our ability to cooperatewith others and help less fortunate people[1, 2, 4, 3]. Sharing food and cooperatingduring hunt have played a fundamental role in the early evolution of human societies[5]and modern variants of these attitudes play a major role still nowadays: we help friendswhen they need, we make donations to less fortunate people, we collaborate with ourpartner to build a family, we cooperate with our colleagues to finish the work faster andat higher standards. Lab experiments show that our pro-social abilities go far beyondthe five rules of cooperation[6]: people show pro-social behavior also in one-shot labexperiments with anonymous participants[7, 8, 9, 10, 11, 12, 13, 14, 15] and even inlarge groups[16]. 1 a r X i v : . [ q - b i o . P E ] D ec major consequence of our pro-social abilities is that our social network is far moreconnected than that of any other animal species. While this dense spatial structurehas numerous well known positive consequences[17, 18, 19], it also generates a painfulparadox: with all the people we are connected with, it is often difficult to make everyonehappy: sometimes the goals of two people are just not aligned; sometimes we have tochoose between hurting Person A or hurting Person B; perhaps even worse, sometimeswe have to choose between hurting ourselves or hurting someone else - and sometimes,this someone else is a close friend, or a close relative, or our romantic partner.Despite the practical importance of such conflicts, little is known about how real peoplebehave in these situations in the ideal scenario of a lab experiment with anonymoussubjects. To the best of our knowledge, only one study[20] addressed this problem,showing that most people are “hyper-altruistic”, that is, they evaluate others’ painmore than their own pain: they pay to avoid an anonymous stranger receiving an electricshock twice as much as they pay to avoid themselves receiving an electric shock.Here we go beyond real physical harm and we show that hyper-altruistic behavior canbe observed also in simple economic decisions where no real physical harm is involved.A major upside of this purely economic approach is that it provides a straight proofthat economic models are somehow incomplete, since (as it will be shown in the Resultssection) they do not predict existence of hyper-altruistic subjects.More precisely, here we report experiments on two types of conflicts, those with an exitoption and those without an exit option. The typical conflict with no exit option involvestwo people, person A and person B. Person A has to decide between two allocationsof money ( s , o ) and ( s , o ), the amount s i being for himself and the amount o i forPerson B. Person B has no active role and only gets what Person A decides to give.The two allocations of money are assumed to be in conflict, that is s > > s and o < < o . Conflicts with an exit option differ from those without an exit option inthat Person A can exit the game without making any decisions, but paying an amount e ≥
0. Thus here Person A has a third choice available, that is the allocation of money( − e, cost of the exit option to be c = ( e − s ) − ( s − e ) = 2 e − s − s ,that is, the difference between the benefit that Person A would get by exiting the gamecompared with the worst case scenario, and the loss that Person A would incur if hetakes the exit instead of maximising his income.We are interested in testing three hypotheses. First, in line with the results presented inref. 20, we expect to observe hyper-altruistic behavior to a larger extent than predictedby economic models. Second, motivated by the results reported in ref. 21, which showthat a substantial proportion of subjects prefer exiting a Dictator game rather thanplaying it, we expect to see a preference for opting out also in our conflictual situations,at least when the exit option is costless. Third, motivated by the pretty well establishedfact that females are more giving than males in the Dictator game[22, 23, 24, 25, 26,27, 28, 29, 30, 31], we suspect that there might be gender differences also in behaviourin conflictual situations. 2o test these hypotheses, we have conducted three studies (Studies 1, 2, and 3) toexplore human behaviour in two-person conflicts with or without an exit option andwith different parameter specifications. Details about the designs are reported in theMethod section and results are reported in the Results section. Here we anticipate thatwe have found evidence of four regularities:(i) In the conditions with a costless exit option ( c = 0), the majority of subjects exitthe game;(ii) In the conditions with a costly exit option ( c > hyper-altruistic subjects cannot be explained by any of the dominant economic models, in-cluding Levine’s model of altruism[32], Fehr & Schmidt’s and Bolton & Ockenfels’ in-equity aversion models[33, 34], Charness & Rabin’s efficiency maximisation model[35],and others[39, 40, 36, 37, 38, 41]. More precisely, since every participant was asked todescribe the reason of his or her choice, with the help of a coder we could analyse themotivation underlying each participant’s decision. We have found evidence that hyper-altruistic participants are likely to have some sort of non-consequentialist moral prefer-ences: they either think that taking money from someone else is wrong, or that givingmoney to someone else is right - independently of the economic consequences.This finding suggests that increasing the moral weight of the decision problem mayhave a positive effect on altruistic behavior. In particular, it is possible that takingmoney from an anonymous person and split it with a third party is perceived to beeven “more wrong” than just taking money from an anonymous person. Motivated bythis observation, we have conducted one more study (Study 4) to investigate whetherthere is a behavioural transition when passing from two-person conflicts to three-personconflicts. Here, in the condition with no exit option, Person A has to decide amongthree allocations of money, ( x, x, − x ) , ( x, − x, x ), and ( − x, x, x ), with x >
0, thefirst component being for himself and the other two components for Person B andperson C, who have no active role. In the condition with an exit option, Person Ahas a fourth option available, according to which he or she can exit the game at nocost, which corresponds to the allocation (0 , , Method
A total of 2.379 subjects living in the US were recruited using the online labour marketAmazon Mechanical Turk (AMT)[42, 43] and participated in one of four experimentsinvolving money.In Study 1, 601 subjects earned $0 .
30 for participation and were randomly assigned toone of six conditions. In the no-exit condition participants were asked to decide between stealing
Person B’s participation fee or donating their participation fee to Person B.Subjects in the role of Person B participated in the guess-no-exit condition and theyhad to guess Person A’s decision with a $0 .
10 reward in case they made the right guess.The free-exit and guess-free-exit conditions were similar, with the difference that therewas a third choice available to Person A, that is, exit the game without doing anything.In this case both subjects would keep their participation fee. Finally, the costly-exit and guess-costly-exit conditions differed from the free-exit conditions in that exiting thegame costed $0 .
05 to Person A. After making their decision, participants entered thedemographic questionnaire, where we asked for their gender, age, and education level,and the reason of their choice. Full instructions are reported in the SupplementaryInformation.Since AMT does not allow experimenters to manipulate participation fees, Study 1actually involves deception: participants’ choices did not have a real impact on their finalbonus. Moreover, one may contest the use of the verb “to steal”, which, having a strongmoral weight, might have driven some participants away from selfish behaviour for otherreasons than their altruism. Analysing participants’ free responses to the question “Whydid you make your choice?”, we did not find any evidence that participants were awareof the risk of deception; however, we have found evidence that the use of the verb“to steal” may have affected participants’ choices. Indeed, several participants, whendescribing their choice, declared “I am not a thief”, or similar statements.To exclude the risk that our results were driven by either of those two causes, Study 2replicates the no-exit condition of Study 1 under slightly different conditions.Specifically, in Study 2, 583 subjects kept their participation fee and were given ad-ditional $0 .
30 as a bonus. Then participants in the role of Person A were asked todecide between taking the other participant’s bonus or giving their bonus to the otherparticipant. Full instructions are reported in the Supplementary Information.4bserving altruistic behaviour in the no-exit condition of Study 1 and in Study 2 willallow us to conclude that there are some subjects who care about the payoff of theother person at least as much as their own. The purpose of Study 3 (395 subjects) is tostrengthen this conclusion showing that a substantial proportion of subjects is hyper -altruist: they care about the payoff of the other person more than their own. Thus inStudy 3, participants kept their participation fee, were given additional $0 .
10, and wererandomly assigned to either the exit-condition or the no-exit condition. In the no-exitcondition, participants in the role of Person A were asked to decide between givingtheir money to the other person or taking the money from the other person. In thelatter case, the money would be doubled and earned by themselves. The exit conditionwas very similar, a part from the fact that participant were allowed to exit the gamewithout making any decision and paying any cost. Full instructions are reported in theSupplementary Information.Finally, Study 4 (600 subjects) investigates a three-person conflict with or withoutcostless exit option. Here, participants kept their participation fee, were given additional$0 .
30, and were randomly assigned to either the exit-condition or the no-exit condition.In the no-exit condition, participants in the role of Person A were asked to decidebetween giving their money to two other people ($0 .
15 each) or taking one of thesepeople’s $0 .
30 and splitting it with the third person. The exit condition was very similar,a part from the fact that participants were allowed to exit the game without makingany decision and paying any cost. Full instructions are reported in the SupplementaryInformation.After collecting the decisions, bonuses were computed and paid. These experiments havebeen conducted in July 2014, while the author was still employed by the University ofSouthampton, United Kingdom. Informed consent was obtained by all participants.These experiments were approved by the Southampton University Ethics Committeeon the Use of Human Subjects in Research and carried out in accordance with theapproved guidelines.
Results
Study 1.
We start by analysing the choices made by the participants who played in the role ofPerson A. Figure 1 reports the relevant results. In the no-exit condition, 28% of the 101subjects decided to donate their participation fee. Adding the possibility to exit thegame for free had the effect that most participants took the exit. Specifically, 70% of the100 subjects who participated in the free-exit condition decided to exit the game, whileall but three of the remaining participants acted selfishly. Three people preferred todonate their participation fee. The fact that virtually nobody acted altruistically in thefree-exit condition also shows that the results of the no-exit condition were not drivenby people who did not understand the rules of the game. The costly-exit condition gave5tatistically the same results as the no-exit condition: 30% of the participants chose toexit the game; all but four of the remaining ones acted selfishly; four people donatedtheir participation fee. In all three conditions, we found that females were more likelythan males to act altruistically, although the effect was nearly significant only in thetwo conditions with an exit option (Rank-sum, p = 0 . , p = 0 . , p = 0 . p = 0 . p = 0 . p = 0 . p = 0 . Study 2.
Study 2 is a replication of the no-exit condition of Study 1 with slightly different ex-perimental instructions. A total of 583 subjects participated in Study 2 in the role ofPerson A. The results show no significant difference with the no-exit condition in Study1: some 21% of the participants preferred giving their money away rather than takingit from the other participant. This percentage does not significantly differ from that inthe no-exit condition in Study 1 (Rank sum, p = 0 . p = 0 . Study 3.
A total of 395 subjects participate in our Study 3 in the role of Person A. Figure 2reports the relevant results. In the no-exit condition, 17% of the 198 subjects, preferredthe allocation ($0 , $0 .
20) over ($0 . , $0). In the exit-condition, 13 subjects chose to actaltruistically, despite the presence of the exit. Among the remaining 184 subjects, only28% of the subjects took the exit option. There is clearly no gender differences in eitherconditions. Observe that the cost of the exit option is $0 .
10 in Study 3, compared with c = $0 .
05 in the costly-exit condition of Study 1 and c = 0 in the free-exit condition ofStudy 1 and in the exit condition of Study 4. Thus this provides evidence that, as thecost of the exit option increases, fewer and fewer people take the exit option and genderdifferences in taking the exit option tend to disappear.6igure 1: In the no-exit condition, about 28% of subjects preferred giving $0 . to ananonymous person, rather than taking the same amount of money from that person.Error bars represent the standard error of the mean. Females tended to give more,though the difference was not statistically significant. In the costly-exit condition, about30% of subjects preferred paying $0 . to exit the game without making any decision,rather than making a decision. Females were more likely than males to exit the game( p = 0 . ). In the free-exit condition, most subjects preferred to exit the game withoutmaking any decision and without paying any cost. Females were more likely than malesto exit the game ( p = 0 . ). The p-values are only nearly significant, but this is alsodue to the small sample size. Aggregating over both exit conditions, we find p = 0 . . In the no-exit condition, about 17% of subjects preferred the allocation ($0 , $0 . over ($0 . , $0) . Error bars represent the standard error of the mean. Inthe exit condition, 13 subjects acted altruistically and are not reported in the figure.Among the remaining participants, only of them took the exit. There is clearly nogender differences in either conditions. In the three-person no-exit condition, about 28% of subjects preferred giving $0 . to two anonymous people ( $0 . each), rather than taking the same amount ofmoney from one of these people and sharing it with the third one. Error bars representthe standard error of the mean. Females tended to give more, though the difference wasnot statistically significant. In the free-exit condition, about 59% of subjects preferred toexit the game without making any decision and without paying any cost. Females weresignificantly more likely than males to exit the game ( p = 0 . ).Study 4. A total of 600 subjects participated in our Study 4, where participants were asked tomake a decision in a three-person conflict instead of a two-person conflict as in Studies1, 2, and 3. Figure 3 reports the relevant results. Perhaps contrary to the expectations,we did not find any significant difference between three-person conflicts and two-personconflicts. In the no-exit condition, 28% of the subjects opted for the altruistic action,while the remaining ones chose either of the selfish options at random. Again, wefound that females were slightly more altruist than males (33% vs 24%), though, again,the difference is not statistically significant ( p = 0 . p = 0 . Distribution of choices in the condition with an exit option. When the exitoption is costless, the majority of people take the exit. This positive effect of the exitoption vanishes as soon as participants are asked to pay to exit the game. In this case,the majority of people remain in the game and act so as to maximise their payoff. Inall conditions, a small percentage of people, ranging from 3% to 7% acted altruistically,despite the presence of the exit option.Distribution of choices in the conditions with an exit option
Figure 4 summarizes the distribution of choices in the conditions with an exit option.Subjects tend to exit the game only when the exit option is costless. Even for exitoptions with a small cost ( c = $0 .
05 in Study 1 and c = $0 .
10 in Study 3), behaviourseem to reverse: the majority of people act selfishly. Across all conditions, we note asmall percentage of people, ranging from 3% to 7%, who acted altruistically, despitethe presence of an exit option. The nature of these people is at the moment unknown.The analysis of participants’ free responses (we asked the participants to describe theirchoice in Study 1 and Study 3, but not in Study 2 and Study 4) suggests that someof these people did not understand the rules of the decision problem. Interestingly,the remaining ones described themselves as particularly generous. However, the totalnumber of people making this choice is so small that at the moment it is impossible todraw general conclusions.
Economic models do not predict hyper-altruistic behavior hyper-altruist if he evaluates others’payoff more than his own[20]. Formally, this corresponds to saying that a person strictly prefers the allocation of money (0 , y ) over ( x, x ≥ y , where the first compo-nent is for himself and the second component for an anonymous stranger he is matchedwith. In this section we show that1. About one-sixth of our subjects acted hyper-altruistically;2. None of the dominant economic models predict existence of hyper-altruistic people.We note that the first statement is not an obvious consequence of our experimentalresults, since it might be possible that some subjects are indifferent between ( x,
0) and(0 , y ). Half of these subjects would statistically choose the allocation (0 , y ). As it willbe shown later, this behavior would be consistent with Bolton & Ockenfels’ inequityaversion model[34] and with Charness & Rabin’s efficiency maximisation model[35].However, we now show that this is not case: virtually all people who chose (0 , y ), madethis choice because they strictly preferred (0 , y ) over ( x, < y ≤ x be fixed, Person A has to decidebetween the allocation of money ( x,
0) and (0 , y ),the first component being for himselfand the second component for Person B. Person B has no active role and only gets whatPerson A decides to give.We start by analysing the predictions of Levine’s model of altruism[32]. This modelassumes that, given an allocation of money ( x , x ), Player 1 gets an utility of u ( x , x ) = x + a + λa λ x , where 0 ≤ λ ≤ − < a , a <
1. In particular, the second condition meansthat no player has a higher regard for his opponents than for himself. It is easy tosee that this property implies that Person A strictly prefers the allocation ( x,
0) over(0 , y ), independently of the parameters of the model. Indeed u ( x,
0) = x ≥ y > ( a + λa ) y/ (1 + λ ) = u (0 , y ). This prediction is rejected by the results of Studies 1,2,and 3.We now consider Fehr & Schmidt’s inequity aversion model[33]. This model assumesthat, given an allocation of money ( x , x ), Player 1’s utility is u ( x , x ) = x − α max( x − x , − β max( x − x , , where β ≤ α and 0 ≤ β <
1. In our case, we have u ( x,
0) = x − β x and u (0 , y ) = − α y . Now assume x = y , as it is in Studies 1 and 2, since β ≤ α , it follows thatevery decision maker prefers ( x,
0) over (0 , x ). This prediction is rejected by the resultsof Studies 1 and 2.Then we consider Bolton & Ockenfels’ inequity aversion model. This model assumesthat, given an allocation of money ( x , x ), with x + x >
0, Player 1’s utility is u ( x , x ) = α x − β (cid:18) x x + x − (cid:19) , where α ≥ β > u ( x,
0) = α x − β / u (0 , y ) = − β /
8, which implies u ( x, ≥ u (0 , y ). Consequently, Bolton &Ockenfels’ model predicts that every player either prefers the allocation ( x,
0) or she isindifferent between the two allocations ( x,
0) and (0 , y ); in other words, no player strictlyprefers (0 , y ) over ( x, , y ) over ( x, x , x ), Player 1’s utility is u ( x , x ) = (1 − α ) x + α ( β min( x , x ) + (1 − β )( x + x )) , where α , β ∈ [0 , u ( x,
0) = (1 − α ) x + α (1 − β ) x and u (0 , y ) = α (1 − β ) y . Since x ≥ y , one always have u ( x, ≥ u (0 , y ). Thus also Charness &Rabin model predicts that no players strictly prefer (0 , y ) over ( x, x,
0) over (0 , x ). Similarly, also the cooperative equilibriummodel[37, 38, 16] reduces to the money maximisation model in case the total welfareis constant across choices. Regret minimisation models[39, 40] instead assume thatplayers compare the payoff obtained when a certain strategy profile is played with thebest payoff they could have gotten choosing another strategy and leaving the strategiesof the other players constant. Then they try to minimise this regret . It is evident thatalso this model predicts that every player should prefer the allocation ( x,
0) over (0 , x ).Finally, the recently proposed model with translucent players[41], which is based onthe illusion of transparency[44], that is the illusion that people’s thoughts are visible toother people (who can respond punishing unfair intentions), also reduces to the moneymaximization model in the case in which the other players have no active role and sothey cannot punish.
Discussion
Here we have provided evidence of the following three regularities: (i) a substantialproportion (about one sixth) of people is hyper-altruist, that is, they prefer givinga certain amount of money to an anonymous stranger, rather than taking the sameamount of money from the same person; (ii) the majority of people prefer to avoid thisconflictual decision and exit the game, but only when the exit-option is costless; (iii)females are more likely than males to exit the game, even when it is costly, but thisgender difference tend to vanish when the cost of the exit option increases.Existence of hyper-altruism is certainly our major result, since it is not predicted byany of the dominant economic models, including Levine’s model of altruism[32], Fehr13 Schimdt’s and Bolton & Ockenfels’ inequity aversion models[33, 34], Charness & Ra-bin’s efficiency maximisation model[35], and others[39, 40, 36, 37, 38, 41]. We are notaware of any model predicting existence of hyper-altruistic people. As a consequence,it is important to understand what psychological and economic motivations led a sub-stantial percentage of people away from the theoretical predictions. Our results providea starting point in that they suggest that hyper-altruistic behaviour is driven by threedifferent (though probably connected) forces: desire to do the right thing; desire not todo the wrong thing; desire to be generous. Further research is necessary to understandhow these forces interact and how they can be incorporated into a model of humanbehaviour.A recent paper[20] makes a point similar to our point (i). There, Crockett et al. showthat most people evaluate others’ pain more than their own pain: they pay to avoidan anonymous stranger receiving an electric shock twice as much as they pay to avoidthemselves receiving an electric shock. Though similar, our results are different in theway that they point out that there is no need of real physical harm to observe hyper-altruistic behaviour. In our experiment, a substantial proportion of people value others’monetary outcome more than their own, without any real physical harm involved.Another paper[21] makes a point similar to our point (ii), that is that most peopleprefer to exit the game, rather than making a decision that would harm either of theparties. There the authors show that about 28% of subjects prefer to exit a dictatorgame with $9, rather than playing it in the role of the dictator with an endowment of$10. More precisely, participants in ref. 21 played a two-stage game: Stage 1 was astandard Dictator game where participants in the role of the dictator had to decide howto allocate $10 between them and an anonymous recipient, knowing that the recipientwould not have any active role. After making the decision, but before telling it to therecipient and before telling to the recipient that they were playing a Dictator game inthe role of the recipient, the dictators played Stage 2, in which they were asked whetherthey wanted to stick with their decision or leave the game with $9. In this latter case, therecipient would not be informed of the fact that they were supposed to be the recipientin a Dictator game. The authors found that 11 subjects (corresponding to 28% of thetotal) preferred to exit the game. Our results extend this finding to conflictual situationsand they also make a little step forward: in ref. 21, only two of the 11 subjects whodecided to exit the game had decided to keep the whole endowment for themselves inthe first stage of the game. Thus it is possible that the fact that the strategy space hasbeen changed from Stage 1 to Stage 2, and the fact that the recipient does not havecomplete knowledge of the decision problem, have changed some people’s preferences,which in Stage 2 act just as money-maximising. Contrariwise, our experiment is a one-stage experiment where both parties have complete knowledge of the decision problemand so it shows that a substantial proportion of people truly have preferences for optingout.Our point (iii) is reminiscent of the pretty well established result that females are moregiving than males in the Dictator game[22, 23, 24, 25, 26, 27, 28, 29, 30, 31]. However,14t goes beyond it, suggesting that females are not only more sharing than males, butthey also have a stronger tendency to exit from a conflictual situation even at a personalcost. While this result is intriguing, we recommend caution on its interpretation. Study3 suggests that when the cost of the exit option increases, gender differences in takingthe exit option tend to disappear. Further research may help understand how robust isthe result that females are more likely than males to exit from a conflict and how far itcan be generalised.We also believe that further research should be devoted to see whether there are be-havioral differences between two-person conflicts and N-person conflicts, with
N > unmitigated communion , that is the extreme focus on others without the balance of afocus on self[46]. Since unmitigated communion is known to cause anxiety, depressivesymptoms, lower self-esteem, and poorer physical health[47, 48, 49], it would be impor-tant to understand the extent to which it can be captured by simple economic gamessuch as the ones we have introduced.
Acknowledgments
This paper was presented at the Human Cooperation Lab Meeting at the Department ofPsychology at Yale University. We thank all participants, in particular Antonio AlonsoAr´echar, Jillian Jordan, Gordon Kraft-Todd, and David Rand, for numerous usefulcomments. We thank Giorgia Cococcioni for assistance with coding participant free-responses. The author was partly funded by the Dutch Research Organization (NWO)grant 612.001.352. 15 eferences [1] Trivers, R. The evolution of reciprocal altruism.
Q Rev Biol , 35-57 (1971).[2] Axelrod, R. & Hamilton, W.D. The evolution of cooperation. Science , 1390-1396 (1981).[3] Fehr, E. & Fischbacher, U. The nature of human altruism.
Nature , 785-791(2003).[4] Rand, D.G. & Nowak, M.A. Human cooperation.
Trends Cogn Sci , 413-425(2013).[5] Kaplan, H. & Gurven, M. The natural history of human food sharing and cooper-ation: a review and a new multi-individual approach to the negotiation of norms.In: Moral sentiments and material interests: The foundations of cooperation in eco-nomic life , [Gintis, H., Bowles, S., Boyd, R. & Fehr, E. (eds.)] [75-113] (MIT Press,Cambridge 2005).[6] Nowak, M.A. Five rules for the evolution of cooperation.
Science , 1560-1563(2006).[7] Kahneman, D., Knetsch, J. L. & Thaler, R. T. Fairness And The Assumptions OfEconomics.
J Bus , S285 (1986).[8] Forsythe, R., Horowitz, J. L., Savin, N. E. & Sefton, M. Fairness in Simple Bargain-ing Experiments. Games Econ Behav , 347-368 (1994).[9] Andreoni, J. & Miller, J. Giving According to GARP: An Experimental Test of theConsistency of Preferences for Altruism. Econometrica , 737-753 (2002).[10] Wong, R.Y. & Hong, Y.Y. Dynamic Influences of Culture on Cooperation in thePrisoner’s Dilemma. Psychol Sci , 429-434 (2005).[11] Artinger, F., Fleischhut, N., Levanti, V. & Stevens, J.R. Cooperation in riskyenvironments: decisions from experience in a stochastic social dilemma. Proceedingsof the 34th Conference of the Cognitive Science Society , 84-89 (2012).[12] Engel, C. & Zhurakhovska, L. When is the Risk of Cooperation Worth Tak-ing? The Prisoners Dilemma as a Game of Multiple Motives.
Available athttp://ssrn.com/abstract=2132501 , 2012 (Date of access: 17/12/2014).[13] Khadjavi, M. & Lange, A. Prisoners and their dilemma.
J Econ Behav Organ ,163-175 (2013).[14] Capraro, V., Jordan, J. J. & Rand, D. G. Heuristics guide the implementationof social preferences in one-shot Prisoner’s Dilemma experiments. Sci Rep , 6790(2014). 1615] Capraro, V., Smyth, C., Mylona ,K. & Niblo, G. A. Benevolent characteristicspromote cooperative behaviour among humans. PLoS ONE , e102881 (2014).[16] Barcelo, H. & Capraro, V. Group size effect on cooperation in social dilem-mas. Available at SSRN: http://ssrn.com/abstract=2425030 , 2014 (Date of access:17/12/2014).[17] Santos, F. C. & Pacheco, J. M. Scale-Free Networks Provide a Unifying Frameworkfor the Emergence of Cooperation.
Phys Rev Lett , 098104 (2005).[18] Perc, M. & Szolnoki, A. Coevolutionary games - A mini review. Biosystems ,109-125 (2010)[19] Rand, D. G., Nowak, M. A., Fowler, J. H. & Christakis, N. A. Static networkstructure can stabilize human cooperation. Proc Natl Acad Sci USA , 17093-17098 (2014).[20] Crockett, M. J., Kurth-Nelson, Z., Siegel, J. Z., Dayan, P. & Dolan, R. J. Harm toothers outweighs harm to self in moral decision making.
Proc Natl Acad Sci USA .DOI: 10.1073/pnas.1408988111 (2014).[21] Dana, J., Cain, D.M. & Dawes, R.M. What you don’t know won’t hurt me: Costly(but quiet) exit in dictator games.
Organ Behav Hum Dec Process , 193-201(2006).[22] Eckel, C.C. & Grossmann, P. Are Women Less Selfish than Men? Evidence fromDictator Experiments.
Econ J , 726-735 (1997).[23] Andreoni, J. & Vesterlund, L. Which is the fair sex? Gender differences in altruism.
Q J Econ , 293-312 (2001).[24] Dufwenberg, M. & Muren, A. Gender Composition in Teams.
J Econ Behav Organ , 50-54 (2006).[25] Houser, D. & Schunk, D. Fairness, competition and gender: Evidence from Germanschoolchildren. J Econ Psychol , 634-641 (2009).[26] Dreber, A., Ellingsen, T., Johannesson, M. & Rand, D. G. Do people care aboutsocial context? Framing effects in dictator games. Exp Econ , 349-371 (2013).[27] Dreber, A., von Essen, E. & Ranehill, E. Gender and competition in adolescence:task matters. Exp Econ , 154-172 (2014).[28] Kettner, S. E. & Ceccato, S. Framing Matters in Gender-Paired Dictator Games. Available at http://archiv.ub.uni-heidelberg.de/volltextserver/16594/ , 2014 (Date ofaccess: 17/12/2014).[29] Boschini, A., Dreber, A., von Essen, E., Muren, A. & Ranehill, E. Gender andeconomic preferences in a large random and representative sample.
Working Pa-per Available at SSRN: http://ssrn.com/abstract=2443315 , 2014 (Date of access:17/12/2014). 1730] Bolton, G.E. & Katok, E. An Experimental Test for Gender Differences in Benefi-cent Behaviour.
Econ Lett , 287-292 (1995).[31] Capraro, V. & Marcelletti, A. Do good actions inspire good actions in others? SciRep , 7470 (2014).[32] Levine, D. K. Modeling altruism and spitefulness in Experiments. Rev Econ Dynam , 593-622 (1998).[33] Fehr, E. & Schmidt, K. A theory of fairness, competition, and cooperation. Q JEcon , 817-868 (1999).[34] Bolton, G. E. & Ockenfels, A. ERC: A Theory of Equity, Reciprocity, and Compe-tition.
Am Econ Rev , 166-193 (2000).[35] Charness, G. & Rabin, M. Understanding social preferences with simple tests. Q JEcon , 817-869 (2002).[36] Halpern, J.Y. & Rong, N. Cooperative equilibrium.
In Proceedings of the 9th In-ternational Conference on AAMAS , 1465-1466 (2010).[37] Capraro, V. A Model of Human Cooperation in Social Dilemmas.
PLoS ONE ,e72427 (2013).[38] Capraro, V., Venanzi, M., Polukarov, M. & Jennings, N. R. Cooperative equilibriain iterated social dilemmas. In Proceedings of the 6th International Symposium onAlgorithmic Game Theory , 146-158 (2013).[39] Halpern, J.Y. & Pass, R. Iterated Regret Minimization: a new solution concept.
Games Econ Behav , 184-207 (2012).[40] Renou, L. & Schlag, K.H. Minimax regret and strategic uncertainty. J Econ Theory , 264-286 (2010).[41] Capraro, V. & Halpern, J. Y. Translucent players: Explaining cooperative behaviorin social dilemmas.
Available at SSRN: http://ssrn.com/abstract=2509678 , 2014(Date of access: 17/12/2014).[42] Paolacci, G., Chandler, J. & Ipeirotis, P. G. Running Experiments on AmazonMechanical Turk.
Judgm Decis Mak , 411-419 (2010).[43] Horton, J.J., Rand, D.G. & Zeckhauser, R.J. The online laboratory: conductingexperiments in a real labor market. Exp Econ , 399-425 (2011).[44] Gilovich, T. Savitsky, K., & Medvec, V. H. The illusion of transparency: biasedassessments of others’ ability to read one’s emotional states. J Pers Soc Psychol ,332-346 (1998).[45] Van Lange, P. A. M., Rusbult, C. E., Drigotas, S. M., Arriaga, X. M., Witcher, B.S., Cox, C. L. Willingness to sacrifice in close relationships. J Pers Soc Psychol ,1373-1395 (1997). 1846] Helgeson, V. S. Relation of agency and communion to well-being: Evidence andpotential explanations. Psychol Bull , 412-428 (1994).[47] Drigotas, S.M., Rusbult, C.E. & Verette, J. Level of commitment, mutuality ofcommitment, and couple well-being.
Pers Relationships , 389-409 (1999).[48] Fritz, H. L. & Helgeson, V. S. Distinctions of unmitigated communion from commu-nion: Self-neglect and overinvolvement with others. J Pers Soc Psychol , 121-140(1998).[49] Helgeson, V. S. Gender-related traits and health. In Social psychological foundationsof health and illness , [Suls, J. & Wallston, K.A. (eds.)] [367-394] (Malden MA,Blackwell, 2003). 19 upplementary Information
This Supplementary Information contains two sections. In the first section we provide allthe details about the analysis of the free responses of the subjects who participated in ourStudy 3. In the second section we report the instructions used in our experiments.
Analysis of free responses
In order to support our conclusion that altruistic behavior in Study 3 was driven byhyper-altruistic subjects, that is subjects who evaluate other’s payoff strictly more thantheir own, rather than by indifferent subjects, who evaluate the other’s payoff the sameas their own, we asked a research assistant to code each response from the altruisticparticipants in Study 3. The coder was not informed about the purpose of the studyand the hypothesis and predictions being tested. For each statement, she was askedwhich of the following five categories best described it:The participant explicitly said that they took the action because that was theright thing to do (Rightness).The participant explicitly said that they took the action because the other actionwas wrong (Wrongness).The participant explicitly said that they took the action because they are generous(Generosity).The participant explicitly said that they took an action at random, because theywere indifferent between the two actions (Indifference).The participant said something that is not classifiable in any of the previouscategories (Not classifiable).Below we report all 30 responses. Next to each response, in parenthesis, we report thecategory to which the response was assigned.1. I’d feel bad if I took it (wrongness).2. it’s the holiday season..it is about giving, so I gave. I hope they appreciate it(generosity).3. i would not feel right taking money from a person (wrongness).4. I’m a giver, not a taker (generosity).5. Did not want to be greedy. If there was a 3rd option of keeping my 10c and theother keeping his I would have selected that (indifference).6. I like giving instead of receiving (generosity).20. Rather have someone else gain a bonus (generosity).8. Maybe the other person needs the money more than me and I won’t take it fromsomeone else’s hand (rightness).9. Just felt generous (generosity)10. i just felt like being nice for once (rightness)11. I wasn’t going to take the other person’s bonus (wrongness).12. The other person most likely needs the money more than I do (rightness).13. Although I considered taking the money, I decided that since it was such a smallamount and that I would feel guilty for taking the money, I decided to give up mymoney (wrongness).14. It’s wrong to take something away from someone else for what amounts to aninsignificant gain for myself (wrongness).15. I would not feel good about myself by taking everything while another person hadnothing (wrongness).16. I thought it would make someone happy and they might need it more than I do(generosity).17. I want to be good (not classifiable)18. Better to give than recieve (rightness).19. I am generous (generosity).20. I didn’t feel right taking all of the money (wrongness).21. I’d rather give something than take something (generosity)22. I would much rather give than take. It just doesn’t feel right to take somethingaway (wrongness)23. I feel better not taking away the others money. The benefit is less than the costof being mean (rightness).24. I didn’t think it was fair to take the other person’s money (wrongness).25. I felt that I wanted to give the amount of the bonus, because I felt it unfair totake all and assume that the other participant would gain more. I generally liketo be fair, and hope that both participants receive an equal or substantial amountfor the work done. I don’t like to be overtly greedy (rightness).26. Because I didn’t want to take from the other participant, and since the only otheroption was to give my ten cents to him/her, I decided that worked better for myconscience (rightness).27. It was the easiest choice and no conflict (not classifiable).218. It was the most kind thing (rightness).29. I had to make a choice (indifference).30. I would have felt guilty leaving someone else with nothing but by giving up my10 cents I feel as if I’ve done something small but good (wrongness).
Experimental instructions
The first screen, where the subjects were asked for their TurkID, was the same for all fourstudies. After this screen, participants entered the real game, where, in a single screen,we presented the problem and asked to make a decision. After making their decision,subjects entered a standard demographic questionnaire where we asked for their gender,age, education level, and reason for their choice (only in Study 1 and Study 3). Belowwe report the instructions used in the decision screen for each of the four studies. Foreach study, we report only the instructions used in the no-exit condition. Those for theother conditions were very similar, a part from the obvious changes.
Study 1
You have been paired with another anonymous participant. You both own $0 .
30 forparticipating in this HIT. Please choose one of the following alternatives: • Donate your $0 .
30 to the other participant. In this case you end the game withnothing and the other participant ends the game with $0 . • Steal the participation fee from the other participant. In this case you end thegame with $0 .
60 and the other participant with nothing.
Study 2
You have been paired with another anonymous participant. You are both given addi-tional $0 .
30 as a bonus. Please choose one of the following alternatives: • Give your $0 .
30 to the other participant. In this case you end the game withnothing and the other participant ends the game with $0 . • Take the other participant’s bonus. In this case you end the game with $0 .
60 andthe other participant with nothing.
Study 3
You have been paired with another anonymous participant. You are both given addi-tional $0 .
10 as a bonus. You can either give your $0 .
10 to the other participant or takehis or her $0 .
10. In this latter case, the money will be doubled and earned by you.What is your choice? 22
Give your $0 .
10 to the other participant. In this case you end the game withnothing and the other participant with $0 . • Take $0 .
10 from the other participant. In this case you end the game with $0 . Study 4
You have been grouped together with other two participants, Person A and PersonB. You are all given additional $0 .
30 as a bonus. Please choose one of the followingalternatives: • Take the $0 .
30 from Person A and share them with person B. In this case, PersonA will finish this task with $0 and you and Person B will finish this task with$0 . • Take the $0 .
30 from Person B and share them with person A. In this case, PersonB will finish this task with $0 and you and Person A will finish this task with$0 . • Give your $0 .
30 to Person A and Person B. In this case, you will finish this taskwith $0 and Person A and Person B will finish this task with $0 ..