null null

Third party punishment increases cooperation in children
through (misaligned) expectations and conditional
cooperation
Philipp Lergetporera,b, Silvia Angerera, Daniela GlätzleRützlera and Matthias Sutterc,d
03 February 2014
a
University of Innsbruck, Department of Public Finance, A-6020 Innsbruck, Austria.
b
Ifo Institute at the University of Munich, D-81679 Munich, Germany.
c
European University Institute, Department of Economics, I-50133 Firenze, Italy.
d
University of Cologne, Department of Economics, D-50923 Köln, Germany.
The human ability to establish cooperation even in large groups of genetically unrelated
strangers depends upon the enforcement of cooperation norms. Third party punishment
is one important factor to explain high levels of cooperation among humans whereas
there seems to be no evidence that other animal species use this mechanism for
promoting cooperation. We study the effectiveness of third party punishment to increase
children’s cooperative behavior in a large scale cooperation game. Based on an
experiment with 1,120 children, aged seven to eleven years, we find that the threat of
third party punishment more than doubles cooperation rates, despite the fact that
children are rarely willing to execute costly punishment. We can show that the higher
cooperation levels with third party punishment are driven by two components. First,
cooperation is a rational (expected payoff-maximizing) response to incorrect beliefs
about the punishment behavior of third parties. Second, cooperation is a conditionally
cooperative reaction to correct beliefs that third party punishment will increase a
partner’s level of cooperation.
S1
Significance statement
Cooperation among humans depends strongly upon the willingness of others to take costly
action in order to enforce the social norm to cooperate. Such enforcement behavior is often
coined third-party punishment. Here we show that third-party punishment is already effective
as means to increase cooperation in children. Most importantly, we can identify why this is
the case. First, children expect (mistakenly) third parties to punish quite often and therefore
become more cooperative. Second, however, the presence of third parties lets children
become (rightfully) more optimistic about the cooperation levels of the interaction partner in a
simple prisoner’s dilemma game. As a reaction to more optimistic expectations, children
cooperate more themselves. The experiment has been run with about 1,100 children aged
seven to eleven years.
S2
Human cooperation rates among genetically unrelated strangers in large groups are unusually
high and exceed cooperation among all other animal species by far (1,2). Already in
childhood, humans begin to conform to cooperative social norms (3). This raises the question
how such social norms are enforced, because norm enforcement may explain why humans
cooperate more than other animals. While in repeated interactions reciprocity (4,5) may
account for the higher cooperation rates – given that, unlike chimpanzees, humans become
reciprocal already in early childhood (6) – in non-repeated (i.e., one-shot) settings a different
mechanism must be at work (7). In fact, a growing body of literature suggests that the
punishment of defectors is key to trigger and sustain cooperation in such contexts (7-11).
Punishment can take on the form of second party punishment, where those who are the
victims of defection can punish norm-violators (7,12-18), or third party punishment, where
unaffected bystanders can execute sanctions against norm-violators, even though the
bystanders are not materially affected by a norm violation (8,19-29). While both humans and
other animals use second party sanctioning to promote cooperation (30,31), it is interesting to
note that humans also engage in third party punishment in order to increase cooperation rates.
In contrast, it is unclear whether aggression against conspecifics among non-human animals
ever aims to punish non-cooperators or whether such behavior can even be called third-party
punishment at all.(*) Some recent evidence even shows that chimpanzees do not engage in
third party punishment (36). Thus, the deliberate punishment of defectors by third parties is a
likely candidate to explain higher cooperation rates in humans. In this paper, we study the
ontogeny of cooperation and third party punishment during childhood, focusing on the
questions whether third parties increase cooperation rates already in children and, if they do,
for which reasons. So far, the experimental literature has documented positive effects of third
party punishment on human cooperation exclusively for adult (typically student) populations
(26-29). Since cooperation norms become internalized much earlier, already in childhood
(3,37), we consider it as important to study how norm enforcement works in childhood. In
their recent paper on the ontogeny of cooperative behavior (unrelated to norm enforcement
through punishment), House et al. (3) conclude that future research should examine
“institutions that influence cooperative behavior and how their acquisition and application
shapes children’s behavior across development” (p. 14590). We work in this direction by
studying the effects of punishment institutions on the cooperative behavior of children.
We are particularly interested in whether children become more cooperative when a third
party (of the same age) may punish them. If so, we try to disentangle the reasons for such a
behavioral response by examining whether children become more cooperative because they
S3
are afraid of getting punished, or because they expect the partner in the cooperation game to
cooperate in the presence of a third party, in which case third party punishment works through
the channel of conditional cooperation (38-40). Studying the behavior of children, as players
in a cooperation game or as third parties, allows determining whether norm enforcement
through third party punishment works already at a young age where many norms are only
beginning to become internalized (3,41). This is of particular interest as potential third party
intervention is important among peers in school (for instance, punishment threats of peers
towards free-riders in cooperative learning environments in order to foster student’s
commitment; 42), but it is unclear to date whether it shifts expectations about others’ behavior
or whether the third party punishment itself promotes norm enforcement.
The experiment was run in the city of Meran/Italy, with more than 1,100 primary school
children, aged seven to eleven years. We chose these age cohorts because (i) important
behavioral and economically relevant traits evolve during this period of life (43), (ii) peer
interactions in primary school classes prepare children for their adult roles, teaching each
other values and attitudes such as cooperation (44,45), and (iii) middle childhood (starting
from around age 6) may be when children begin to conform to cooperative social norms (3).
A necessary prerequisite for strategic interaction experiments to provide reliable results is that
participants can understand them and, in our case, have developed the ability to take another
person’s perspective. In contrast to non-human primates like chimpanzees who “do not have a
full-blown human-like theory of mind” (46, p. 156) both conditions are entirely met in
humans in the age cohorts considered in this paper (3,46-49). The participating children
represent 86% of all primary school children in grades two to five in this city with its 38,000
inhabitants.
Following previous literature on adults (26), we let our subjects play a one-shot,
simultaneous prisoner’s dilemma game as a baseline (see Fig. 1 for an illustration of the game
and Methods as well as Supplementary Information, SI, for details). Mutual cooperation
yields the Pareto-efficient outcome of [4,4]. However, both players have a dominant strategy
to defect, leading to the inefficient Nash-equilibrium of [2,2].
554 children played this two-player game in a control treatment (CTR) without any third
party. Matching was random and anonymous and pairs were always formed from the same
age cohort. After having played the game once (but before being informed about the choice of
their partner in the PD), these children were asked to act as third parties for another set of
children. Children did not know about this additional task before completing play in the PD.
S4
A different set of 566 children was assigned to a third-party punishment treatment (TPP).
Children in TPP were randomly and anonymously paired (within their age cohort) and then
played the PD once. Each child in a pair of TPP was assigned one child from CTR as the third
party, and children in TPP were aware of this before making decisions. Of course, they were
not informed about the third party’s decision before choosing to cooperate or defect. The third
party (the child in CTR) had to decide whether to invest a token to punish the assigned child
(in TPP) in case this child would defect in the PD (†). As a consequence of punishment, the
child in TPP lost all gains from the PD-experiment if it had defected. If the child in TPP had
cooperated, or if the third party had not invested its token, then the third party kept the token
which could be exchanged into a reward (‡).
After having made their own decisions, we asked children about their beliefs. Both in
CTR and in TPP they were asked whether they expected the partner in their pair to cooperate.
In TPP they were additionally asked whether they expected the third party to punish
defection. Correct guesses were rewarded with one token. By asking children in TPP both
about the expected punishment and the expected cooperation of their partner, we can check
whether they cooperate in order to avoid punishment or because they expect the partner to
cooperate as well.
Results
Fig. 2 shows cooperation rates across the four age cohorts for players in CTR and in TPP.
Overall, we find that 58% of players cooperate in the one-shot PD in TPP, while only 25% do
so in CTR (P = 0.000 overall and in each age group separately, χ2-tests). This means that the
presence of a third party with an opportunity to punish defectors more than doubles
cooperation rates. Looking at cooperation rates across age cohorts, we find no significant age
effects within any treatment (P = 0.339 in CTR and P = 0.552 in TPP, Cuzick’s Wilcoxontype test for trend).
Fig. 3 illustrates players’ beliefs about their partner’s likelihood of cooperation. We find
that beliefs significantly differ across treatments: 64% of subjects in TPP, but only 51% in
CTR belief that their partner will cooperate (P=0.000 across all age groups, χ2-test). This
means that subjects anticipate that the presence of third parties will have an impact on the
partner’s willingness to cooperate. In TPP, in all four age cohorts the expected likelihood of
cooperation matches the actual rate of cooperation fairly closely. Comparing the dark grey
bars across Fig. 2 and Fig. 3 does not yield significant differences in any age cohort (P = 0.16
for 7/8 years; P = 0.16 for 8/9 years; P = 0.09 for 9/10 years; P = 0.42 for 10/11 years;
S5
McNemar’s tests). However, in CTR all four age cohorts are too optimistic, because expected
cooperation rates are always significantly higher than actual cooperation rates (see the light
grey bars in Fig. 2 and Fig. 3; P < 0.01 in each cohort; McNemar’s tests). This latter result
suggests an intention to free-ride on the (expected) contributions of partners, which implies a
willingness to accept advantageous inequality (50,51). As soon as a third party is present (in
TPP), however, expectations and actual behavior with respect to cooperation get well
calibrated. This effect can explain why (potential) third party punishment increases
cooperation rates if subjects are conditional cooperators (38-40). For someone who conditions
the level of cooperation on the interaction partner’s willingness to cooperate, third party
punishment shifts the expectations upwards, and hence triggers more cooperation, even in the
absence of actual punishment. In fact, Fig. S3 in SI shows that the average expectations of
conditional cooperators (that is, subjects whose belief about the cooperative behavior of the
partner is aligned with their own decision) are significantly higher in TPP than in CTR (P =
0.000 in each cohort; χ2-tests).
Fig. 4 juxtaposes actual and expected punishment rates. It turns out that third parties use
the punishment option very rarely, overall in less than 10% of cases (§). Players in TPP
expect third parties to punish in 51% of cases on average, however. The difference is highly
significant throughout (P = 0.000 in each age cohort; χ2-tests), indicating a strong mismatch
between beliefs and actual punishment behavior. A similar mismatch, albeit of smaller size,
has been found in previous studies of third party punishment when subjects share a pie very
unevenly in a simple allocation task (26). Hence, the mismatch is not an artefact of our design
or subject pool (¶). Moreover, the fact that only a small fraction of primary school children
incurs costs in order to punish defectors is consistent with the finding that children’s behavior
is typically closer to payoff-maximization than adult behavior (53; see also the evidence of
payoff-maximizing behavior of chimpanzees; 55).
It is interesting to note that, given actual punishment behavior, players in TPP have
higher expected payoffs from defection than from cooperation in all age cohorts, meaning that
it would be a payoff-maximizing strategy to defect. Hence, if only actual punishment was
important, it should not have any effect on cooperation rates (contrary to what we see in Fig.
2). However, given expected punishment rates, cooperation yields higher expected payoffs
than defection for all cohorts, except for the oldest (where cooperation and defection have
practically the same expected payoff; see Tab. S2 in SI).
Hence, cooperation in TPP is driven by two components. First, it is a rational (expected
payoff-maximizing) response to incorrect beliefs about the punishment behavior of third
S6
parties. Second, it becomes more likely as a conditionally cooperative reaction to an increase
in the expected cooperation rate of a subject’s partner. The latter increase, in turn, is due to
the presence of third parties (||).
From a societal point of view, TPP is more efficient than CTR. Given the actual
cooperation rate of 24.6%, the expected payoff of a player is 2.49 tokens in CTR. Taking into
account the 8% chance of losing all earnings through punishment, and considering the
cooperation rate of 58% in TPP, a player in TPP earns on average 3.01 tokens. Subtracting
from this the average costs of 0.03 tokens for the third party through punishing (8% of the
42% of defectors get punished by third parties, which costs them 1 token), yields a net surplus
of 2.98 tokens, which is 20% higher than in CTR.
Discussion
Third party punishment increases cooperation already among children, aged seven to eleven
years. Across these age cohorts, we have found no significant differences in reactions to
potential punishment. Most noteworthy, third party punishment works through two channels,
one of which relies on a misalignment of actual and expected punishment behavior. Subjects
expect to get punished for defection much more often than third parties are actually willing to
incur the costs of punishment. Yet, the expectation of punishment suffices to increase
cooperation rates. This misalignment between actual and expected punishment may, in fact,
explain why field data suggest that third party sanctioning is hardly observed whereas
cooperation rates are found to be substantial at the same time (56). The second channel
through which third parties increase cooperation rates is their effect on expected cooperation
rates of other players. The presence of third parties with a punishment option is expected to
make others more cooperative, which in turn triggers own cooperation as a consequence of
conditional cooperation. In fact, while the prevalence of conditional cooperation has been
shown for adults (38,39), our study can be interpreted as showing that already children are
conditional cooperators. Moreover, our study establishes a link between conditional
cooperation and the cooperation-enhancing effect of third party punishment.
Among the avenues for future research we see three straightforward extensions of our
work. First, it would be interesting to see whether third party reward is equally efficient in
increasing cooperation as is third party punishment or whether positive and negative
incentives work differently (57). Second, it would be a worthwhile project to study even
younger children than we did in this paper. We consider it an intriguing question whether at a
very early age (potential) third party punishment would be executed on the one hand, and
S7
would be effective to increase cooperation on the other hand. Third, studying how the
presence of third party observers who cannot punish affects cooperation in children would be
interesting. As we measure the joint effect of observation and punishment on cooperation
rates, disentangling both channels contributes to the understanding whether the presence of
third parties, the possibility of punishment, or the interaction of both promotes cooperation
among humans.
Methods
We conducted our experiment in all fourteen elementary schools in Meran (South Tyrol,
Italy) in November 2012. Meran is the second largest city in the province of South Tyrol, with
about 38,000 inhabitants, of which roughly 50% are Italian speaking and 50% German
speaking. Our experiment was part of a larger research project which investigated economic
decision making of elementary school children. In Italy, elementary school comprises grades
1 to 5.
Before starting the project we obtained approval from the Internal Review Board of the
University of Innsbruck, the South Tyrolean State Board of Education, and from the
headmasters as well as informed consent from the parents of the involved children to run a
series of six experimental sessions in the academic years 2011/12 and 2012/13. We got
permission from 86% of parents of all elementary school children in Meran. The permissions
were either granted implicitly or explicitly. Each school district separately decided whether
the parents had to sign a consent form in order to give their child the permission to participate
(opt-in) or whether participation was implicit and the parents had to sign in order to prohibit
participation (opt-out). One out of five school districts decided to implement the explicit
participation consent form, the others implemented the implicit participation. In all cases,
parents received a letter explaining the general purpose of the two-year research project prior
to the start of the experiments. Participation in each experimental session was, of course,
voluntary for children, but all except a single child consented to participate.
The experiment on cooperation and punishment was the second experiment conducted
with the children in the second year of the study. In that year we worked with children in
grades 2 to 5 (while we had grades 1 to 5 in the first year). In total, we had 1,141 children
participating in this experiment.
Each child was fetched individually from the classroom and brought to a separate room
where an experimenter explained the experiment one-to-one to the child. In this room, there
were four to eight experimenters running the experiment with four to eight children at the
S8
same time, visually separated from one another. Treatments were randomized within each
experimenter and experimenters had to memorize the instructions of the game and explain the
game orally (in the mother-tongue of the child), with detailed visual support (see SI for
experimental instructions). The duration of the experiment was approximately 20 minutes and
it was conducted with pen and paper. Following the general procedure when conducting
experiments with children (43,58), children had to repeat the rules of the game in their own
words after the explanation by the experimenter. In case of mistakes, the experimenter
repeated the respective passages, and asked the child to repeat the rules once more. 21
children did not manage to correctly repeat the rules, in particular the consequences of each
combination of actions. Given our one-on-one explanation technique, this is a reasonable rate
(43), leaving us with 1,120 children with full understanding for the analysis (see Tab. S1). Of
course, the 21 children without correct understanding were allowed to participate in the
experiment until the end. Including their choices would not change any of our results.
A subject either participated in treatment CTR or TPP. Both experimental treatments had
two stages. This was not known to children at the beginning of the experiment. Only at the
end of the first stage (after having made all decisions and after having answered our questions
on expectations) they were informed about the second stage and its rules. Children were
informed about the outcome of the two stages of the experiment only three months later when
they received the presents (**).
In each stage, a child was anonymously matched with another child from the same age
cohort (i.e., grade) and language group, but from a different school. This was common
knowledge. The baseline game in the experiment was a one-shot prisoner’s dilemma (shown
in Fig. 1). In this game, one child and its partner were endowed with two tokens each, which
could be either kept or passed on to the other player. In the latter case, the tokens were
doubled. This game was played in the first stage of treatment CTR where no external observer
was present, hence the game was played without a third party who could punish defection.
The prisoner’s dilemma game without third party was also played in the second stage of
treatment TPP. This second stage in TPP served as a within-subjects control whether subjects
who had experienced third party punishment in stage one would change their behavior when
third party punishment was removed. The change was as expected, with cooperation rates
dropping significantly in the absence of third party punishment, and the data are shown in Fig.
S1 and Fig. S2 in SI. In the paper we do not report the second stage of treatment TPP.
In the first stage of treatment TPP the prisoner’s dilemma game was extended to a thirdparty punishment experiment: each player in the prisoner’s dilemma of this stage was paired
S9
with an exclusive third-party observer. This was a subject in stage 2 of CTR. The observer
was not affected by the play of the children in treatment TPP, but was endowed with one
token. This token could either be kept by the child in stage 2 of CTR or could be spent to
destroy the whole payoff of the paired player in TPP if this player chose defection. Since
children in the role of an observer had played the game themselves before (in stage 1 of
CTR), they were familiar with the rules of the game and could easily condition their decision
on the paired player’s choice to cooperate or defect. Of course, the observers (in CTR) were
not informed about the actual choice of the observed player (in TPP) before making their
decision on how to spend the token, meaning that we implemented a so-called strategy
method (52). The decision of the observer was only implemented in case of defection. Thus,
we did not allow for spiteful punishment (61), because we are primarily interested in the
enforcement of a cooperation norm and not whether children are willing to punish cooperative
acts. All involved participants were exactly informed about the punishment mechanism. It is
also noteworthy that the observed players knew that their partner in the prisoner’s dilemma
also faced a punishment threat in case of defection, because the partner had also one
(different) child assigned as an observer with an opportunity to punish defection. At the very
end of the session children completed a post-experimental questionnaire on demographic data
(on siblings, gender and age). Total earnings in the experiment were determined by actual
decisions and also by the stated expectations. The latter were also incentivized. Subjects
earned an extra token per correct guess.
As incentives, we used sweets (lollipops, small chocolates, candies), fruits (bananas,
apples, oranges) and other small presents (stickers, balloons, pencils, wristbands). Children
could exchange the tokens earned in the experiment into items of their choice in a so-called
“experiment-store”. The cost of each item ranged from one to three tokens.
As control variables, we measured the IQ and the extent of altruism and intertemporal
preferences of our participants one to six months before the experiment on third party
punishment. IQ was elicited with a shortened version of Raven’s test. Altruism was elicited in
a dictator game. Subjects were endowed with 6 tokens and we let them decide anonymously
how many tokens to keep for themselves (and exchange them into presents in the
“experiment-store”) or to donate to one of the province’s largest charities, “Menschen in Not:
Kinderarmut durch Kinderreichtum”, respectively “Umanità che ha bisogno: famiglia
numerosa = famiglia povera?”, an initiative to support underprivileged children in South
Tyrol. This charity is run by the well-known Caritas diocese Bolzano-Bressanone. For each
token allocated to the charity we donated 50 Euro Cents to the charity. Intertemporal
S10
preferences were measured with the use of a choice list. Each child had to make three
decisions in which to choose either 2 tokens at the end of the experiment or a larger number
of tokens with a delay of 4 weeks. The delayed payoff was either 3 tokens, 4 tokens, or 5
tokens. From these choices, we can identify very impatient children (who always choose the 2
tokens immediately) and check whether they behave differently from more patient children.
We have data for 977 children who participated in the third party punishment experiment and
in both the altruism experiment and the intertemporal choice task, and these children are the
basis for the regressions shown in Tab. S3 in SI.
S11
Acknowledgements
The authors thank Rudolf Meraner from the South Tyrolean State Board of Education
(Pädagogisches Institut für die deutsche Sprachgruppe in Südtirol), the headmasters of the
participating schools (Gabriella Kustatscher, Maria Angela Madera, Eva Dora Oberleiter,
Brigitte Öttl, Ursula Pulyer, Vally Valbonesi), the parents of the involved children for making
this study possible, and the children for participation. We received helpful comments from
two referees, Loukas Balafoutas, Nikos Nikiforakis, Karl Sigmund, the audiences at the
Maastricht Behavioral and Experimental Economics Symposium 2013 and the University of
Innsbruck. Financial support from the Government of South Tyrol and the “Aktion D.
Swarovski” at the University of Innsbruck is gratefully acknowledged.
S12
Footnotes
* Several studies report behavior which resembles third-party punishment in non-human
animals (32-35). It is likely, however, that such behavior is motivated by selfishness and
yields cooperation only as a by-product (36).
† While both altruistic and spiteful punishment, i.e. the punishment of defectors and
cooperators, is usually permitted in experiments with adults (26) we restricted our
participants’ action space to altruistic punishment, because we are primarily interested in
the enforcement of a social norm to cooperate and not whether children are willing to
punish cooperative acts.
‡ We chose this binary punishment technology in order to assure comprehension. While
adults who act as third parties in such games are usually asked to choose the number of
tokens to be invested into punishment (26,29), we kept the design as simple as possible
and thus only allowed for a binary decision.
§ One possible reason for the low incidence of punishment rates could be the use of the
strategy method to decide upon punishment. Third parties had to make decisions
conditional on the player in the PD game choosing defection. This may leave the third
party in an emotionally “cold” state, while choosing punishment after having seen the
player defect in the PD game may create an emotionally “hot” state and then trigger more
punishment. A recent survey comparing the strategy method with the latter type of directresponse method has failed to find a systematic behavioral impact of the strategy method,
however (52). In their study on third-party-punishment with adults, Fehr and Fischbacher
(26) also use the strategy method for eliciting the decisions of third-party observers and
find between 21% and 50% punishment rates (depending on the decision of the
interaction partner in the PD). Our somewhat smaller punishment rates are compatible
with the finding that children’s behavior is typically closer to payoff-maximization
(which predicts no punishment) than adult behavior (53).
¶ One might also argue that the mismatch between beliefs about the behavior of third parties
and actual punishment rates is due to the difficulty of children playing the PD game to
put themselves into the role of the third party and take her perspective. However,
psychological studies on the development of Theory of Mind (47,54) show that normally
developing children are able to differentiate the other’s view from their own one by the
age of four to six years, and then take the perspectives of other persons into account.
|| In SI we show support for this with a regression (in Tab. S3) in which the likelihood to
cooperate is the dependent variable. Expecting the partner to cooperate increases a
subject’s likelihood of cooperation by 42 percentage points (P = 0.000), and expecting
the third party to punish defection raises the likelihood of cooperation by 38 percentage
points (P = 0.000). A Wald-test shows that both factors are equally strong and not
significantly different from each other. In this regression, we also control for age, gender,
IQ and other covariates. Of the latter, only altruistic giving in an experiment on voluntary
donations to a charity turns out to be significant (and positive, as expected).
** As the total earnings of each child were dependent not only on own choices, but also on
the decision of the partner in the experiment (who was from another school), it was not
possible to calculate the final earnings of the children immediately at the end of a session.
Thus the tokens earned in the experiment were handed over at our next visit, three
months after this experiment took place. Given our delayed payment procedure and the
high discount rates among children (59,60) we checked whether impatient children
behaved differently from more patient ones, because the former might have perceived the
incentives in the experiment as less valuable than the latter. To tackle this important issue
(suggested by a referee) we compared the choices (in TPP and CTR) of very impatient
children with their more patient peers (we measured patience in an independent
S13
experiment on intertemporal preferences conducted six month before the experiment for
the present paper was conducted). Table S3 in SI shows that children who are categorized
as more impatient do not behave differently in the PD game.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Richersen PJ, Boyd R (2005) Not by Genes Alone: How Culture Transformed Human
Evolution (Univ of Chicago Press, Chicago).
Fehr E, Fischbacher U (2004) Social norms and human cooperation. Trends Cogn Sci 8:
185-190.
House B et al. (2013) Ontogeny of prosocial behavior across diverse societies. Proc Natl
Acad Sci USA 110: 14586-14591.
Hamilton WD (1964) The genetical evolution of social behavior. I, II. J Theor Biol 7: 152.
Trivers RL (1971) The evolution of reciprocal altruism. Q Rev Biol 46: 35-57.
House B, Henrich J, Sarnecka B, Silk JB (2013) The development of contingent
reciprocity in children. Evol Hum Behav 34: 86-93.
Fehr E, Gächter S (2002) Altruistic punishment in humans. Nature 415: 137-140.
Boyd R, Gintis H, Bowles S, Richerson P J (2003) The evolution of altruistic
punishment. Proc Natl Acad Sci USA 100: 3531-3535.
Hauert C, Traulsen A, Brandt H, Nowak M A, Sigmund K (2007) Via freedom to
coercion: The emergence of costly punishment. Science 316: 1905-1907.
Boyd R, Gintis H, Bowles S (2010) Coordinated punishment of defectors sustains
cooperation and can proliferate when rare. Science 328: 617-620.
Mussweiler T, Ockenfels A (2013) Similarity increases altruistic punishment in humans.
Proc Natl Acad Sci USA Early Edition: 1-6.
Fehr E, Gächter S (2000) Cooperation and punishment in public goods experiments. Am
Econ Rev 90: 980-994.
Nikiforakis N, Normann H T (2008) A comparative statics analysis of punishment in
public-good experiments. Exp Econ 11: 358-369.
Nikiforakis N (2010) Feedback, punishment and cooperation in public good experiments.
Games Econ Behav 68: 689-702.
Dreber A, Rand D G, Fudenberg D, Nowak M A (2008) Winners don’t punish. Nature
452: 348-351.
Yamagishi T (1986) The provision of a sanctioning system as a public good. J Pers Soc
Psychol 51: 110-116.
Gächter S, Herrmann B (2009) Reciprocity, culture and human cooperation: previous
insights and a new cross-cultural experiment. Phil Trans R Soc B 364: 791-806.
Gürerk Ö, Irlenbusch B, Rockenbach B (2006) The competitive advantage of sanctioning
institutions. Science 312: 108-111.
Gintis H (2000) Strong reciprocity and human sociality. J Theor Biol 206: 169-179.
Fehr E, Fischbacher U, Gächter S (2002) Strong reciprocity, human cooperation and the
enforcement of social norms. Hum Nature 13: 1-25.
Gintis H, Bowles S, Boyd R, Fehr E (2003) Explaining altruistic behavior in humans.
Evol Hum Behav 24: 153-172.
Fehr E, Fischbacher U (2003) The nature of human altruism. Nature 425: 785-791.
Fowler J H (2005) Altruistic punishment and the origin of cooperation. Proc Natl Acad
Sci USA 102: 7047-7049.
Mathew S, Boyd R (2011) Punishment sustains large-scale cooperation in prestate
warfare. Proc Natl Acad Sci USA 108: 11375-11380.
S14
25. Shinada M, Yamagishi T (2007) Punishing free riders: direct and indirect promotion of
cooperation. Evol Hum Behav 28: 330-339.
26. Fehr E, Fischbacher U (2004) Third-party punishment and social norms. Evol Hum Behav
25: 63-87.
27. Carpenter J P, Matthews P H, Ong’ong’a O (2004) Why punish? Social reciprocity and
the enforcement of prosocial norms. J Evol Econ 14: 407-429.
28. Charness G, Cobo-Reyes R, Jiménez N (2008) An investment game with third-party
intervention. J Econ Behav Organ 68: 18-28.
29. Carpenter J P, Matthews P H (2012) Norm enforcement: Anger, indignation or
reciprocity? J Eur Econ Assoc 10: 555-572.
30. Hauser MD (1992) Costs of deception: Cheaters are punished in rhesus monkeys
(Macaca mulatta). Proc Natl Acad Sci USA 89: 12137-12139.
31. Clutton-Brock TH, Parker GA (2005) Punishment in animal societies. Nature 373: 209216.
32. Ratnieks FLW, Wenseleers T (2005) Evolution. Policing insect societies. Science 307:
54-56.
33. Monnin T, Ratnieks FLW, Jones GR, Beard R (2002) Pretender punishment induced by
chemincal signaling in a queenless ant. Nature 419: 61-65.
34. Ratnieks FLW, Wenseleers T (2008) Altruism in insect societies and beyond: Voluntary
or enforced? Trends Ecol Evol 23: 45-52.
35. Raihani NJ, Grutter AS, Bshary R (2010) Punishers benefit from third-party punishment
in fish. Science 327: 171.
36. Riedl K, Jensen K, Call J, Tomasello M (2012) No third-party punishment in
chimpanzees. Proc Natl Acad Sci USA 109: 14824-14829.
37. Fan C-P (2000) Theaching children cooperation – An application of experimental game
theory. J Econ Behav Organ 41: 191-209.
38. Keser C, van Winden F (2000) Conditional cooperation and voluntary contributions to
public goods. Scand J Econ 102: 23-39.
39. Fischbacher U, Gächter S, Fehr E (2001) Are people conditionally cooperative? Evidence
from a public goods experiment. Econ Lett 71: 397-404.
40. Kurzban R, Houser D (2005) Experiments investigating cooperative types in humans: A
complement to evolutionary theory and simulations. Proc Natl Acad Sci USA 102: 18031807.
41. Schmidt M F H, Tomasello M (2012) Young children enforce social norms. Current
Directions in Psychological Science 21: 232-236.
42. Joyce WB (1999) On the free-riding problem in cooperative learning. Journal of
Education for Business 74: 271-274.
43. Fehr E, Bernhard H, Rockenbach B (2008) Egalitarianism in young children. Nature 454:
1079-1084.
44. Parsons T (1959) The school class as a social system: some of its functions in American
Society. Harvard Educational Review 29: 297-318.
45. Johnson DW (1981) Student-student interaction: The neglected variable in education.
Educational Researcher 10: 5-10.
46. Tomasello M, Call J, Hare B (2003) Chimpanzees understand psychological states – the
question is which ones and to what extent. Trends in Cognitive Science 7: 153-156.
47. Astington JW (1993) The child’s discovery of the mind (Harvard University Press,
Cambridge).
48. Harbaugh W, Krause K, Berry TR (2001) GARP for kids: On the development of rational
choice behavior. American Economic Review 91: 1539-1947.
49. Call J, Tomasello M (2008) Does the chimpanzee have a theory of mind? 30 years later.
Trends in Cognitive Science 12: 187-192.
S15
50. Fehr E, Schmidt K (1999) A theory of fairness, competition, and cooperation. Q J Econ
114: 817-868.
51. Bolton G E, Ockenfels A (2000) ERC: A theory of equity, reciprocity, and competition.
Am Econ Rev 90: 166-193.
52. Brandts J, Charness G (2011) The strategy versus the direct-response method: a first
survey of experimental comparisons. Experimental Economics 14: 375-398.
53. Camerer FC (2003) Behavioral game theory: Experiments in strategic interaction
(Princeton University Press, Princeton).
54. Selman R L (1971) Taking another’s perspective: Role-taking development in early
childhood. Child Dev 42: 1721-1734.
55. Jensen K, Call J, Tomasello M (2007) Chimpanzees are rational maximizers in an
ultimatum game. Science 318: 107-109.
56. Balafoutas L, Nikiforakis N (2012) Norm enforcement in the city: A natural field
experiment. Eur Econ Rev 56: 1773-1785.
57. Gneezy, U, Meier, S, Rey-Biel, P (2011) When and why incentives (don’t) work to
modify behavior. J Econ Perspectives 25(4): 191-210.
58. Harbaugh W, Krause K (1999) Economic experiments that you can perform at home on
your children. Working Paper, University of Oregon.
https://scholarsbank.uoregon.edu/xmlui/bitstream/handle/1794/56/UO-19991_Harbaugh_Economic_Experiments%5B2%5D.pdf?sequence=1
59. Castillo, M, Ferraro, P J, Jordan, J L, Petrie, R (2011) The today and tomorrow of kids:
Time preferences and educational outcomes of children. J Public Econ 95: 1377-1385.
60. Bettinger E, Slonim R (2007) Patience among children. J Public Econ 91: 343-363.
61. Herrmann B, Thöni C, Gächter S (2008) Antisocial punishment across societies. Science
319: 1362-1367.
S16
Author Contributions P.L., S.A., D.G.-R. and M.S. contributed equally to all parts of the
research.
Additional Information
The corresponding author hereby confirms that all experiments were performed in accordance
with relevant guidelines and regulations.
Competing financial interests: The authors declare that they have no competing financial
interests.
S17
Figure 1: In the prisoner’s dilemma, players can either cooperate (C) or defect (D). While
mutual cooperation yields the socially optimal outcome of 4 tokens per player, each subject
has an incentive to defect as a dominant strategy. Defection of both players is the Nash
equilibrium of the game, yielding a payoff of 2 for each player. The first (second) number in
each cell indicates player 1’s (player 2’s) payoff.
Figure 2: Average cooperation rates by age and treatment (N=554 in CTR and N=566 in
0
relative frequency of cooperation
.2
.4
.6
.8
1
TPP). Error bars, mean ± SEM.
7/8
8/9
9/10
10/11
Age in years
CTR
TPP
S18
Figure 3: Average expectation of cooperative behavior of the partner by age and treatment
1
.8
.6
.4
.2
0
relative frequency of expecting the other player to cooperate
(N=554 in CTR and N=566 in TPP). Error bars, mean ± SEM.
7/8
8/9
9/10
10/11
Age in years
Belief (CTR)
Belief (TPP)
Figure 4: Average actual punishment rates of third parties (CTR; N=554) and expected rates
0
relative frequency of punishment
.2
.4
.6
.8
1
of punishment by players in TPP (N=566) by age. Error bars, mean ± SEM.
7/8
8/9
9/10
10/11
Age in years
Actual punishment (CTR)
Expected punishment (TPP)
S19
Supplementary Information
Third party punishment increases cooperation in children
through (misaligned) expectations and conditional
cooperation
by
Philipp Lergetporer, Silvia Angerer, Daniela Glätzle-Rützler and
Matthias Sutter
Content
Supplementary tables and figures
p. S2
Experimental instructions
p. S7
S1
Supplementary tables and figures
Table S1: Number of participants by age and gender with the number of
children not understanding the game in brackets.
Gender
Age (in years)
Female
Male
7/8 years
126 (5)
139 (9)
8/9 years
121 (1)
164 (2)
9/10 years
131 (2)
153 (1)
10/11 years
133 (0)
153 (1)
ALL (N=1,141)
511 (8)
609 (13)
In total, 21 children (10 in CTR and 11 in TPP) were excluded from the analysis because they
did not understand the experiment.
Table S2: A comparison of expected payoffs in TPP, given beliefs about expected behavior and
given actual behavior of the opponents.
Age (in
Observer’s
Partner’s probability
years)
probability to punish to cooperate (C)
Expected payoff
Expected payoff
(given behavior) $
(given beliefs) $
(P)
Behavior Beliefs
Behavior
Beliefs
Coop.
Defect
Coop.
Defect
7/8 yrs.
7.5%
53.8%
60.6%
67.4%
2.42
4.09
2.70
2.17
8/9 yrs.
9.1%
53.5%
57.0%
64.1%
2.28
3.89
2.56
2.12
9/10 yrs.
7.3%
52.4%
57.8%
65.3%
2.31
4.00
2.61
2.20
10/11 yrs.
7.8%
44.8%
56.6%
60.7%
2.26
3.93
2.43
2.44
$
Calculation of expected payoff
… for cooperation: C*4
… for defection: (1-P)*C*6 + (1-P)*(1-C)*2
S2
Table S3: Clustered probit regression on cooperation
Age in years
Female (=1)
German school (=1)
Belief partner (1=cooperation)#
Relative IQ$
Number of siblings
Altruism&
Impatient (=1)+
Cooperation
Cooperation
in CTR
in TPP
0.011
0.001
(0.015)
(0.020)
-0.036
0.077
(0.037)
(0.052)
0.035
0.055
(0.035)
(0.050)
0.320***
0.416***
(0.035)
(0.065)
-0.061
0.015
(0.103)
(0.127)
0.001
0.007
(0.020)
(0.023)
0.083***
0.065***
(0.015)
(0.020)
-0.051
-0.027
(0.043)
(0.058)
Belief observer (1=punishment)
0.379***
(0.076)
Belief partner*Belief observer
0.114
(0.106)
# Observations
489
488
Notes. ***, **, * denote significance at the 1%, 5%, 10% level, robust standard errors in parentheses.
Clustered on class level.
$
The IQ was measured with Raven’s Coloured Progressive Matrices. Consistent with the
mean values of the norming sample of the Raven’s Coloured Progressive Matrices, the share
of correctly solved matrices in our subject pool increases with age. Therefore, we measured
the IQ relative to the respective age group in order to avoid confounding age- and IQ-effects.
&
Number of tokens donated to a charity (0-6) in an independent experiment. We include this
variable as a proxy for social intelligence which is needed in social interactions (Kaukiainen
S3
et al. (1) find that social intelligence is significantly related to empathy).
+
Patience was measured in an independent experiment with 3 binary choice problems. If a
child decided in all three choice problems not to wait for the higher payoff, the child was
classified as being impatient.
Reference:
62. Kaukiainen A et al. (1999) The relationships between social intelligence, empathy, and
three types of aggression. Aggressive Behavior 25: 81-89.
S4
Figure S1: Average cooperation rates by age and stage in TPP. Stage 1 with third party
punishment (reported in the paper); Stage 2 without third party punishment (not reported in
0
relative frequency of cooperation
.2
.4
.6
.8
1
the paper, N=566). Error bars, mean ± SEM.
7/8
8/9
9/10
10/11
Age in years
TPP Stage 1
TPP Stage 2
Overall, 20% cooperate in stage 2 of TPP without punishment, compared to 58% with punishment in stage 1
(P=0.000 overall and in each cohort; McNemar exact test). The cooperation rate in stage 2 is not significantly
different from the rate in CTR (compare Fig. 2; P>0.05 overall and in each cohort, χ2-tests).
Figure S2: Average expectation of cooperative behavior of the partner by age and stage in
TPP. Stage 1 with third party punishment (reported in the paper); Stage 2 without third party
1
.8
.6
.4
.2
0
relative frequency of expecting the other player to cooperate
punishment (not reported in paper, N=566). Error bars, mean ± SEM.
7/8
8/9
9/10
10/11
Age in years
Belief (TPP Stage 1)
Belief (TPP Stage 2)
In the second stage of TPP (without punishment) children have significantly and much lower expectations about
their partner’s likelihood to cooperate (P < 0.01 in all age cohorts; Mc-Nemar test).
S5
Figure S3: Average expectations of conditional cooperators towards the cooperative behavior
of their partners by age in CTR (N=369) and TPP (N=418). Subjects are classified as
conditional cooperators if their beliefs about the partner’s choice are aligned with their own
relative frequency of expecting the other player to cooperate
0
.2
.4
.6
.8
1
decision.
7/8
8/9
9/10
10/11
Age in years
Belief (CTR)
Belief (TPP)
S6
Experimental instructions (translated from German/Italian)
Note: Italic font is used for the instructions to the experimenter.
Assign child to treatment 1 or treatment 2 according to the randomization-list.
Today I prepared a game for you. In this game you can earn tokens. You can buy presents in
our shop with these tokens. Bigger presents cost more than smaller ones. The presents are
different from last time.
Treatment CTR
In this game you can earn these white tokens (show tokens). Could you please repeat the rules
for the tokens? (Child must repeat that it can select presents with the tokens and that bigger
presents are more costly than smaller presents).
There are two meeples: a yellow and a white meeple (place meeples on table; the yellow
meeple must be placed directly in front of the child). You are the yellow meeple (point at
yellow meeple). The white meeple is a child which we will select randomly. This child attends
the x. grade of a German (Italian) school here in Meran just like you, but it attends a different
school (adapt explanation to grade and school-language; place white meeple on schoolcard). This child can be a boy or a girl, but you don’t know whom exactly you are playing
with. This is a secret. Your partner doesn’t know either who you are. Could you please repeat
this part? (Child must repeat all the information of this paragraph. If it misses some parts, ask
explicitly)
Stage 1: Prisoner’s Dilemma without punishment
The game works like this:
(place decision-sheet “COOP” in front of the child). At the beginning you and your partner
receive 2 tokens each (place tokens in front of yellow and white meeple). Both of you must
decide whether to send ZERO or BOTH tokens to the partner. It is important to note that I
have tokens as well (show tokens). If you send your tokens to your partner, I will add 2 more
tokens for your partner (show physically: shove the subject’s tokens towards its partner and
double the tokens upon arrival; restore original distribution after illustration). If your partner
decides to send you his/her tokens, I will double them as well (show physically). You must
decide between sending ZERO and sending BOTH tokens on this sheet (point at the
respective box on the decision sheet). Your partner has the same options as you: he/she can
either send you ZERO or BOTH tokens. Let’s go through some examples now (The child
must handle tokens in the examples).
S7
I)
What happens if you tick this box here (point at box with ZERO tokens) and your
partner sends you ZERO tokens? (Answer: no tokens will be sent and the
experimenter does not add any tokens) And how many tokens will you earn in this
case? (Answer: 2) And how many tokens will your partner earn? (Answer: 2)
II)
What happens if you tick this box here (point at box with BOTH tokens) and your
partner sends you BOTH of his tokens? (Answer: the child and its partner send
both tokens to one another and the experimenter doubles each transaction) And
how many tokens will you earn in this case? (Answer: 4) And how many tokens
will your partner earn? (Answer: 4)
III)
What happens if you tick this box here (point at box with ZERO tokens) and your
partner sends you BOTH of his tokens? (Answer: the child does not send tokens to
partner, but the partner sends his/her tokens to the child. The experimenter
doubles the transaction from the partner to the child) And how many tokens will
you earn in this case? (Answer: 6) And how many tokens will your partner earn?
(Answer: 0)
IV)
What happens if you tick this box here (point at box with BOTH tokens) and your
partner sends you ZERO of his tokens? (Answer: the child sends both tokens to
partner, but the partner sends no tokens to the child. The experimenter doubles the
transaction from the child to the partner) And how many tokens will you earn in
this case? (Answer: 0) And how many tokens will your partner earn? (Answer: 6)
Do you already know how many tokens your partner sends you? (Answer: No.) Exactly.
Likewise, your partner does not know how many tokens you sent him when he decides on
how many tokens to send you. Could you please repeat how the game works? (Child must
exhaustively repeat the rules of the game. If it misses some parts, ask explicitly)
We don’t know yet the exact number of tokens you will earn in this game. You receive the
tokens which you keep and those which your partner sends you. Since we don’t know yet how
many tokens your partner will send you, you will receive the tokens from this part not today
but only when we visit you next time. It is very important that your decision in this game is
secret: the other children will never know how many tokens you sent.
Please take your decision now. Take as much time as you need. In the meantime I will turn
around so that I don’t disturb you. Just call me when you are done. (Hand over the pen to the
child so that it can decide. Turn around and wait until the child signals that it has finished.)
S8
Thank you for your decision. You ticked this box (point at the respective box). Could you
please explain what that means (Child must explain what the [possible] consequences of its
decision are)?
I still have another question for you. What do you think decides your partner? If your guess is
correct (meaning that your partner really does what you think), you will receive one extra
token. If your guess is not correct, you don’t earn an additional token. What do you think, will
your partner send you ZERO or BOTH of his/her tokens (Register answer of the child in
computer)? Thank you for your decision. This game is over now. (Put away the material of
this stage)
Stage 2: Punishment Decision
(Place two white meeples in front of the child and put orange tokens on the table) Here I
prepared a second game for you. In this game you can observe the play of two other children
in a game which works just like the one we played before. These children attend the x. grade
of a German (Italian) school here in Meran just like you, but they attend a different school
(adapt explanation to grade and school-language; place white meeples on school-cards). We
selected these children randomly and both are new children, i.e. none of them is your partner
from before. You are the observer of one of these children (point to the left meeple from the
subject’s point of view). You do not know yet how this child decides in this game, but there
are two possibilities: This sheet (place explanation sheet 1 in front of the child) shows the first
option the child has. Can you tell me how many tokens this child would send to its partner
(Answer: 0). On this sheet (place explanation sheet 2 in front of the child and to the right of
explanation sheet 1) you see the second option the child has. Can you tell me how many
tokens this child would send to its partner (Answer: 2). As an observer you have the option to
deduct tokens from the child if it does not send the tokens to its partner (point at explanation
sheet 1). Deducting tokens from the child works like this (place orange-token-card (left) and
deduction card (right) in front of the child): You can either pick the card with the orange
token (point at orange-token-card) or the deduction-card (point at the deduction card). If you
pick the card with the orange token (point at orange-token-card), this token is yours and with
this token you can select a present which costs 1 token when we visit you next time. In this
case, the child you observe can keep all its tokens. If you pick the deduction-card instead
(point at deduction-card), you don’t receive the orange token but we will deduct all the tokens
from the child which you observe. In this case, these tokens are lost and nobody gets them.
This means that the child receives zero tokens in the game. If the child chooses to send the
tokens to its partner (point at explanation sheet 2) you will receive the orange token and we
S9
will not deduct the tokens from the child no matter which card you picked. This means that
you can only deduct points from a child who does not send its tokens.
It is important that the child knows at the time of deciding that you might deduct its tokens if
it does not send them. The partner of the child you observe (point at right meeple) has an
observer on its own and this observer can deduct the tokens of the partner if he/she does not
send the tokens. In the game you played before nobody had the possibility to deduct your
tokens. Could you please repeat how the game works? (Child must exhaustively repeat the
rules of the game. If it misses some parts, ask explicitly). It is very important that your
decision in this game is secret: the other kids will never know which card you picked. The
other children really exist and you can really deduct the tokens of a child which sends nothing
by picking the deduction-card.
Now you may choose between the two cards. Please leave the card which you want to pick on
the table as it is so that I can see its picture and turn the other card upside down. If you pick
the card with the orange token, just flip the deduction card (demonstrate). In this case you will
receive the token in any case and we don’t deduct the tokens from the child in any case. In
contrast, if you pick the deduction card, flip the card with the orange token (demonstrate). In
that case you will not receive the orange token and we will deduct the child’s tokens if it
doesn’t send it. If it sends its tokens, you will receive the orange token and we don’t deduct
any token from that child. Please take your decision now. Take as much time as you need. In
the meantime I will turn around so that I don’t disturb you. Just call me when you are done.
(Hand over the pen to the kid so that it can decide. Turn around and wait until the child
signals that it has finished. Then register answer of the child in the computer.)
Thank you for your decision. You picked this card (point at the respective card). Could you
please explain what that means (Child must explain what the [possible] consequences of its
decision are).
Treatment TPP
In this game you can earn these white tokens (show tokens). Could you please repeat the rules
for the tokens? (Child must repeat that it can select presents with the tokens and that bigger
presents are more costly than smaller presents).
There are two meeples: a yellow and a white meeple (place meeples on table; the yellow
meeple must be placed directly in front of the child). You are the yellow meeple (point at
yellow meeple). The white meeple is a child which we will select randomly. This child attends
the x. grade of a German (Italian) school here in Meran just like you, but it attends a different
S10
school (adapt explanation to grade and school-language; place white meeple on schoolcard). This child can be a boy or a girl, but you don’t know whom exactly you are playing
with. This is a secret. Your partner doesn’t know either who you are. Could you please repeat
this part? (Child must repeat all the information of this paragraph. If it misses some parts, ask
explicitly)
Stage 1: Prisoner’s Dilemma with punishment
The game works like this:
(place decision-sheet “COOP_PUN” in front of the child). At the beginning you and your
partner receive 2 tokens each (place tokens in front of yellow and white meeple). Both of you
must decide whether to send ZERO or BOTH tokens to the partner. It is important to note that
I have tokens as well (show tokens). If you send your tokens to your partner, I will add 2 more
tokens for your partner (show physically: shove the subject’s tokens towards its partner and
double the tokens upon arrival; restore original distribution after illustration). If your partner
decides to send you his/her tokens, I will double them as well (show physically). You must
decide between sending ZERO and sending BOTH tokens on this sheet (point at the
respective box on the decision sheet). Your partner has the same options as you: he/she can
either send you ZERO or BOTH tokens. Let’s go through some examples now (The child
must handle tokens in the examples).
I)
What happens if you tick this box here (point at box with ZERO tokens) and your
partner sends you ZERO tokens? (Answer: no tokens will be sent and the
experimenter does not add any tokens) And how many tokens will you earn in this
case? (Answer: 2) And how many tokens will your partner earn? (Answer: 2)
II)
What happens if you tick this box here (point at box with BOTH tokens) and your
partner sends you BOTH of his tokens? (Answer: the child and its partner send
both tokens to one another and the experimenter doubles each transaction) And
how many tokens will you earn in this case? (Answer: 4) And how many tokens
will your partner earn? (Answer: 4)
III)
What happens if you tick this box here (point at box with ZERO tokens) and your
partner sends you BOTH of his tokens? (Answer: the child does not send tokens to
partner, but the partner sends his/her tokens to the child. The experimenter
doubles the transaction from the partner to the child) And how many tokens will
you earn in this case? (Answer: 6) And how many tokens will your partner earn?
(Answer: 0)
S11
IV)
What happens if you tick this box here (point at box with BOTH tokens) and your
partner sends you ZERO of his tokens? (Answer: the child sends both tokens to
partner, but the partner sends no tokens to the child. The experimenter doubles the
transaction from the child to the partner) And how many tokens will you earn in
this case? (Answer: 0) And how many tokens will your partner earn? (Answer: 6)
Do you already know how many tokens your partner sends you? (Answer: No.) Exactly.
Likewise, your partner does not know how many tokens you sent him when he decides on
how many tokens to send you. Could you please repeat how the game works? (Child must
exhaustively repeat the rules of the game. If it misses some parts, ask explicitly)
I still have to explain you something very important: There is yet another meeple (place black
meeple next to the yellow meeple) and this meeple is your observer in the game. This child
attends the x. grade of a German (Italian) school here in Meran just like you, but it attends a
different school (adapt explanation to grade and school-language; place white meeples on
school-cards). We selected this child randomly and this is a new child, i.e. it is not your
partner but another child. Your observer can see what you do in this game and he/she can
deduct your tokens. This works as follows:
If you tick this box here (point at box with ZERO tokens) your observer can deduct all your
tokens. The deduction works as follows: Your observer must choose between the card with
the orange token (show orange-token card) and the deduction card (show deduction-card). If
he/she goes for the orange token (shove orange-token-card in front of the black meeple) then
he/she can use this token to buy a present in our shop for him/herself and you keep all your
tokens. If however, he/she picks the deduction-card (shove deduction-card in front of the
black meeple), he/she does not receive the orange token and we will deduct ALL your tokens.
In this case, your tokens will be lost and nobody gets them. That is, you won’t earn any tokens
in this game. However, your observer can only deduct your tokens if you don’t send any to
your partner. If you tick this box here (which means that you send both tokens to your
partner), your observer can’t deduct tokens from you and he/she receives the orange token for
him/herself. In this game your partner (point at the white meeple) has a separate observer
which chooses between the orange-token- and the deduction-card. Could you please repeat
how the game works? (Child must exhaustively repeat the rules of the game. If it misses some
parts, ask explicitly). What happens to the deducted tokens? (Answer: they are lost and
nobody gets them).
We don’t know yet the exact number of tokens you will earn in this game. You receive the
tokens which you keep and those which your partner sends you. If you don’t send your
S12
tokens, your observer might deduct all your tokens. Since we don’t know yet how the other
children decide you will receive the tokens from this part not today but only when we visit
you next time. It is very important that your decision in this game is secret: the other children
will never know how many tokens you sent.
Please take your decision now. Take as much time as you need. In the meantime I will turn
around so that I don’t disturb you. Just call me when you are done. (Hand over the pen to the
child so that it can decide. Turn around and wait until the child signals that it has finished.)
Thank you for your decision. You ticked this box (point at the respective box). Could you
please explain what that means (Child must explain what the [possible] consequences of its
decision are)?
I still have two more questions for you. What do you think will your partner do? If your guess
is correct (meaning that the other child really does what you think), you will receive one extra
token. If your guess is not correct, you don’t earn an additional token. What do you think your
partner will do? Remember that your partner has an observer who can deduct his/her tokens if
he/she doesn’t send you the tokens. Do you think that your partner will send you ZERO or
BOTH of his/her tokens? (Register answer of the child in computer). Thank you for your
decision.
Case 1: Subject did not send its tokens: And what do you think your observer will do? If your
guess is correct (meaning that your observer really does what you think), you will get one
extra token. If your guess is not correct, you don’t earn an additional token. What do you
think your observer will do? (Register answer of the kid in computer)
Case 2: Subject sent its tokens: And what do you think your observer would do if you didn’t
send your tokens? If your guess is correct (meaning that your observer really does what you
think), you will get one extra token. If your guess is not correct, you don’t earn an additional
token. What do you think your observer would do? (Register answer of the kid in computer)
Thank you for your decision. This game is over now. (Put away the decision sheet, the black
meeple, the cards of the observer and the white tokens)
Stage 1: Prisoner’s Dilemma without punishment
Here I prepared a second game for you. You can earn these yellow tokens in this game (show
tokens). It is important that you can’t sum tokens of different colors. Apart from that, the same
rules for the tokens apply.
There are two meeples: a yellow and a white meeple. You are the yellow meeple (point at
yellow meeple). The white meeple is again a child which we select randomly. This child
attends the x. grade of a German (Italian) school here in Meran just like you, but it attends a
S13
different school (adapt explanation to grade and school-language; place white meeple on
school-card). This child can be a boy or a girl, but you don’t know whom exactly you are
playing with. This is a secret. Your partner doesn’t know either who you are. It is important
that this partner is a new child and not your partner or observer from the game we played
before. Could you please repeat this part? (Child must repeat all the information of this
paragraph. If it misses some parts, ask explicitly). This game works exactly the same as the
one we played before. The only difference is that you don’t have an observer now and thus,
nobody can deduct your points if you decide not to send them. Your partner (point at the
white meeple) has no observer either. Could you please repeat how the game works? (Child
must exhaustively repeat the rules of the game and mention the differences to the first stage. If
it misses some parts, ask explicitly)
We don’t know yet the exact number of tokens you will earn in this game. You receive the
tokens which you keep and those which your partner sends you. Since we don’t know yet how
many tokens your partner sends you, you will receive the tokens from this part not today but
only when we visit you next time. It is very important that your decision in this game is
secret: the other children will never know how many tokens you sent.
Please take your decision now. Take as much time as you need. In the meantime I will turn
around so that I don’t disturb you. Just call me when you are done. (Hand over the pen to the
child so that it can decide. Turn around and wait until the child signals that it has finished.)
Thank you for your decision. You ticked this box (point at the respective box). Could you
please explain what that means (Child must explain what the [possible] consequences of its
decision are)?
I still have another question for you. What do you think will your partner do? If your guess is
correct (meaning that your partner really does what you think), you will receive one extra
token. If your guess is not correct, you don’t earn an additional token. What do you think, will
your partner send you ZERO or BOTH of his/her tokens (Register answer of the child in
computer)? Thank you for your decision. This game is over now. (Put away the material of
this stage)
S14
Sample Decision Sheet
1

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project