‘Does attractiveness have an influence on the cooperation rate in the prisoners dilemma, and what are the differences between gender.’

Master Thesis

014

Studentnumber: 10422811

Msc Economics: Behavioral economics

Supervisor: 1e Theo Offerman, 2e Arthur Schram

‘Does attractiveness have an influence on the

cooperation rate in the prisoners dilemma, and what

Table of content

Introduction: ... 4

State of the art: Prisoners Dilemma. ... 6

Repeated PD ... 8

One –shot game ... 9

Gender differences ... 11 Attractiveness ... 13 Research method ... 14 Procedure... 14 Design ... 16 Results ... 18 Hypothesis 2: ... 34 Conclusion ... 36 Discussion. ... 37 Acknowledgement. ... 38 Literature. ... 39 Appendix. ... 41 Instructies ... 41 3

Introduction:

Inspired by the episode of the TV show Split or steal of 14 march 2008, I started wondering how people make decisions in a prisoners dilemma. In this episode a man and a woman had the possibility to split more than $100.000. Although during the communication both stated to split, the woman choose to steal. She took all the money back home and left an angry man with nothing. His trust in her was unjustified, but what are the factors influencing this final decision? Was her attractiveness maybe decisive to him? This situation has led me to the question ‘Does attractiveness have an influence on the cooperation rate in the prisoners dilemma, and what are the differences between genders.’ Evidence shows that attractive people are judged more positively on a wide variety of dimensions (Dion, Berscheid and Walster, 1972). Daniel Hamermesh has done many studies towards the role of attractiveness in different perspectives. In one of these studies (Hamermesh, 1993) he found significant results showing that attractive people earn more money than average looking people, and average looking people in turn earn more than less attractive people. Differences can go up to fifteen percent. There is a debate ongoing that this is discriminating opposed to less attractive people. Pfann (2000) generated results which could be used as an argument in favour of this discrimination. His results show that although attractive persons earn more, their attractiveness yields extra revenue exceeding these higher earnings. This result is the strongest when people are interacting face to face with each other. Then their attractiveness raises the value of the products they sell.

Game theory is a perfect method to simulate real life situations in a controlled game where influences of specific factors could be measured. By implementing a prisoners dilemma resembling the design of ‘split or steal’ I will be able to investigate the influence of attractiveness in the decisions made by people. Many factors have already been investigated and are described in the literature regarding the prisoner dilemmas. An interesting factor is ‘gender’. In a number of studies it is found that men and women behave differently in decision making. However, attractiveness has never played a role in previous studies related to the prisoners dilemma. This will be the added value of this research. The results of this research will create new insights which could be used in the ongoing debate towards the discrimination concerning attractiveness

The first hypothesis which will be tested is ‘Men behave differently than women in the prisoners dilemma’. Investigating this hypothesis is relevant because this will lay the basis necessary to understand the influence of attractiveness which will be investigated afterwards. The results shows that the extent to which communication is allowed, before decision making, is influencing the cooperation rate significantly. When the communication is unrestricted by rules, women show a higher cooperation rate. Men seem to be more aware of the context of the game. In contradiction to the ‘split or steal’ case, the results of this study show that the ability to estimate the opponent’s behaviour is in favour of men. Where men seem to base their decision more on the obtained information during the interaction, women seem to base their decision on their inner tendency to show empathy. Since the gender specific behaviour in the prisoners dilemma is now clear, the following hypothesis is tested: ‘Attractiveness does have an influence in decision making, this influence is in advance of more attractive people’. The results of this thesis show that although the belief of participants is not influenced by attractiveness, they experience more guilt and make more positive promises when they face attractive people. Only women face the tendency to cooperate significantly more when the other participant is more attractive. And in contrast to men, only attractive women benefit compared to less attractive women because the outcome of the prisoners dilemma is in their advance.

The structure of the paper is as follows: first, the state of the art regarding the prisoners dilemma literature will be summarized. The purpose of this section is to mention important results of earlier studies. Thereafter, the research method will be described. The design, implementation and procedure used for this research will be explained. This method is used to generate the data necessary to conduct the further research. Third, the results of the data analysis will be presented. Based on the results, the hypotheses will be tested on their validity. Finally, the conclusion will include a summary of the main results. On the basis of these main results the main question and sub question posed as research questions will be answered.

State of the art: Prisoners Dilemma.

In real life, people are often engaged in negotiations with others. Cooperation is usually paying off. However, people often are inclined not to cooperate in order to maximize their own welfare at the expense of joint welfare. For example when two students have to hand in together an article before a specific deadline, it pays off to work together to meet the deadline easily. But when one student leaves it to the other to work hard to meet the deadline, he is maximizing his own welfare (he can afford to be lazy). A problem occurs when both students are not working hard and fail to hand in the article before the deadline. There is a model to research situations like this, it is the strategic game well known as the ‘Prisoners Dilemma’ (PD) (Scodel et al, 1959).

The PD can be shown in a two-player game or as a multiple-player game. For simplicity, the focus in this research will be only on the two-player game. As shown in the paper of Scodel et al. (1959) the game is played by two players (figure 1). Each player has two choices; α1 and α2 for player 1 and β1 and β2 for player two. There are four possible outcomes as shown in figure 1.

Figure 1: classic prisoner’s dilemma (Scodel et al. (1959))

In each outcome, the first value (xi) denotes the value addressed to player 1. The second value (xj) denotes the value addressed to player 2. In the classic prisoners dilemma both xi and xj are subject to the following restrictions:

As long as these restrictions hold, the game can be called a PD. Assuming that both players are rational, both players have a dominant strategy. These strategies (α2 for player one and β2 for player two) are dominating the other strategy (α1 and β1) while the dominant strategy is best for their own welfare, independent of the choice of the other player. When both players play their dominant strategy, the Nash equilibrium outcome will be α2, β2. In this case the return to each player is x4. Yet the return to each player when the outcome (α1, β1) occurs is x1. So, as is one characteristic of strategic games, the Nash equilibrium outcome is not always the best outcome for total welfare. One may be inclined to claim that while (α2, β2) is rational in some formal sense, reasonable players will prefer (α1, β1). The following example of a PD will be used to make it easier to understand.

In each couple of values in figure 3, the first value in blue denotes the value addressed to player 1 (you). The second value in orange denotes the value addressed to player 2 (opponent). As already discussed, the values are subject to restrictions. By filling in the restrictions, one can see that the restrictions hold.

1) 2*2 > 0 + 3 > 2*1 1) 2x

> x

+ x

> 2x

2) 3 > 2

2) x

> x

3) 3 > 1

3) x

> x

4) 1 > 0

4)

x

> x

Figure 2: Example of classic prisoners dilemma

Because the restrictions hold for the values used in figure 2 one calls this a classical PD. If restriction two and four hold, the rational choice for both players is option Y. If both choose rational, the Nash equilibrium outcome is (1, 1). Although this is the rational outcome, one can see that both will prefer to end up in (2, 2) because then for both returns are higher. Repeated PD

The literature of the PD makes a clear distinction between the one-shot PD and the repeated PD. As the name already reveals, the one-shot PD is only played once while the repeated PD is played more than once. Most literature about PD has focused on the repeated PD. This is seen as a more realistic simulation of real life because people often interact more than once with each other (Bó, 2005|Andreoni and Miller, 1993| Trivers, 1971| Mckelvey and Palfrey, 1992). Just like in the one-shot PD, the dominant strategy is option Y. When both players are rational and implement backward-induction arguments (Andreoni and Miller, 1993), they will choose Y every trail leading to the Nash equilibrium outcome for all trails. In the “prospect theory” it is stated that people in real life have incomplete information about other players (Kahneman and Tversky, 1979). Kreps et al. (1982) show that if there is incomplete information about the types of players, cooperation early in the game can be consistent with rational behaviour. Suppose that both players believe that there is a small chance that their opponent may be altruistic, for instance that the opponent may get extra pleasure from mutual cooperation or may even adopt a tit-for-tat strategy. In the tit-for-tat strategy, it could be in each player’s best interest to pretend, at least for some time, to be an altruistic player in order to build a reputation for cooperation. This could continue until the game eventually unravels to mutual defection (Andreoni and Miller, 1993).

In many researches significant evidence was found that reputation building and altruism are important reasons for cooperative behaviour (McKelvey and Palfrey (1992)| Palfrey and Rosenthal (1988). In a research of Tullock (1985) the role of reputation became very evident. When people may choose with whom one will play the game, defectors have a higher chance to be rejected.

In the literature of the Repeated game, there is a clear distinction between the finite and the infinite repeated game. In a finite repeated game, the last round could be seen as a regular one-shot game in which the role of reputation does not influence the choice of strategy anymore. When it is known when the game will stop, people tend to anticipate on

this with their behaviour. In the infinite version of the repeated game, the game will continues forever. In order to simulate an infinite repeated PD, research is done in situations where the players do not know when the game stops. In the paper of Bó (2005), significant evidence is found that the higher the probability of continuation, the higher the levels of cooperation. We know now why people cooperate in repeated PD, therefore it is interesting to see whether people still cooperate in the one-shot PD.

One –shot game

A one-shot game can be used as a clean method to control a lot of factors. In the repeated game, participants are influenced by the outcomes of previous rounds. This influence is different between participants and hard to control. The one-shot PD is used as a method to control for factors like this, and therefore a good method to measure the influence of one-shot PD specific factors. In the paper of Frank et al. (1993) it is argued that there are many examples of real life situations which can be compared with the one-shot PD. For this reason it is relevant to study the behaviour of persons in a one-shot PD.

In playing the one-shot game, research has been done in both anonymous and non-anonymous games. When the game is not non-anonymous, players see each other before or during playing the PD. According to Bohnet and Frey (1999) identified interaction leads to significantly higher cooperation rates compared to anonymity, because mutual identification prior to the game allows the social norm to become relevant. Human beings do not care about the other’s welfare per se, but they react to restrictions which are relevant in a specific context. The greater the extent to which a decision is taken in a social context, the more relevant manners become. When the possibility to communicate is included, the cooperation in the PD will further increase, as argued in the paper of Dawes et al (1977). In this paper it is concluded that communication results in increased cooperation in a two-person PD setting. Communication effects can have at least three reasons. First, the opportunity to communicate allows group members to get acquainted, which may raise their concern for each other’s welfare. Second, relevant information exchanged through discussion appealing for mutual cooperation could persuade group members to cooperate. Third, group members’ statements of their own intended decisions could assure other members of their good intentions, leading to higher rates of cooperation.

Another view why cooperation increases through communication is because of the change in behaviour of conditional co-operators. These people have the desire to match the cooperation level of others. In both laboratory and field experiments, half of the participants are more willing to cooperate if others do so as well (Assem van den et al, 2012). This egalitarian motives can often explain conditional cooperative coordination. In experimental studies this conditional cooperation is investigated in settings where participants have the possibility to condition their behaviour directly to the behaviour of others. But this is not realistic in real life. In real life conditional behaviour is often based on the beliefs or expectations of the other. Then the degree of coordination depends on the predictive power or available information. Before an interaction, one will have a belief of what a (unknown) opponent will choose, playing a PD. After interaction this belief may change because of the obtained information. In a research by Dawes et al (1977) it became clear that the predictive power (belief) is more accurate in conditions with communication. According to Frank (1988), individuals who have the propensity to cooperate are endowed with an advanced emotional system causing them to cooperate more, but also enables them to signal their propensity to cooperate during interaction. These individuals experience stronger emotions, such as sympathy, compassion, guilt or shame. Research showed that cooperation occurs more often when the opponent makes an explicit though non-binding promise that he/she will cooperate (Sally, 1995). This effect of promises can be explained by the combination of a preference for conditional cooperation and reluctance to lie. People like to cooperate if others do so, and a promise is a reliable signal of others’ behaviour if they are reluctant to lie. When the possibility is there, many participants make an explicit promise or make a statement towards their intention (Assem van den et al, 2012). Gneezy (2005) reports experimental evidence indicating that people do not like to lie. His participants deceive primarily if they thereby gain a lot, or impose little loss. So, when someone makes a non-binding promise, this is a reliable indicator that this person will cooperate in the PD. In the articles of Belot, Bhaskar and van de Ven (2010) evidence is found that players do not have a higher propensity to choose cooperation when their opponent makes a promise. The other player even displays a marginally significant decrease in the likelihood of cooperation (Assem van den et al, 2012).

In the anonymous played PD, both players cannot trace the choice made by the other. Still cooperation takes place (Bó, 2005). So, even when no one will ever know which

choice is made, some people still choose to cooperate. Economists have often concluded that this human behaviour occurs because of preferences different from financial incentives (Andreoni, 1995| Fehr and Gächter, 2000| Levine, 1998, among others). The paper of Croson and Gneezy (2009) mentions a list of these preferences; ’emotions, overconfidence and framing’. A very important conclusion in this paper is that there is a clear difference in the behaviour between men and women.

Gender differences

A number of studies indicate that women’s social preferences are different from social preferences of men. In the paper of Balliet et al (2011) a theoretical overview is provided to further understand these gender differences. Both socio-cultural and evolutionary psychological theories are central in this understanding. These theories provide unique insights into the origins and manifestations of gender differences in social dilemmas like the PD. The socio-cultural theory is built on an historical point of view. Decades ago, societies had a clear distribution of roles. This distribution occurred because of biological differences between genders. Men are on average physically stronger and faster, while women bear the cost of pregnancy and childcare. These differences led the genders to acquire culturally different sets of skills to fulfil their roles: women developed more interpersonal relationship skills because they are assumed to play a domestic role in the society. They are perceived to be more communal in orientation, less selfish, more caring, friendly and emotionally expressive. The skills of men were more agentic, focusing on high status and power, making them more independent, assertive, ambitious and dominant. Manifestations of these agentic versus communal tendencies have been demonstrated in a large body of research ( Buchan et al, 2008). For example in task-oriented small groups, men are more focused on the task whereas women tend to be more focused on social aspects. Men are more aggressive in general, whereas women are more likely to attend to their partner in a social interaction, displaying more empathy. Finally, women, more than men, emphasize equality and harmony in relationships (Buchan et al, 2008). Cooperation in a social dilemma conveys a concern for the welfare of others which is what a communal orientation is all about, while agentic orientation implies the concern for own outcomes over the outcomes of others. According to these research results one expects from the sociological theory that women will be more cooperative in society than men (Balliet et al., 2011). In research conducted by

Frank (1988) it is concluded that cooperative individuals are endowed with an advanced emotional system experiencing stronger emotions. These individuals are more able to signal their propensity to cooperate and are more able to recognize the signals of other opponents. Because skills of women resemble the description of advanced emotional system more than skills of men, one tends to conclude that women will be more accurate in predicting the behaviour of their opponent.

Prior research on gender stereotypes suggests that women are more caring about others even beyond their own society or close relationships (Schwartz and Rubel, 2005| Eagly and Steffen, 1984| Balliet et al., 2011). From the evolutionary psychological perspective it is assumed that women and men have evolved different context-specific decision rules that enable each gender to benefit from interactions with other people in different environments. These different context-specific decision rules evolved because men had to hunt, requiring tracking and killing skills whereas women needed gathering skills related to locating and recalling food sources among an array of vegetation. Due to selection men and women successful in their skills had more chance to create offspring. This selection has ensured the evolution of different skills (Balliet et al., 2011).

Based on this theoretical overview, one may expect that both women and men respond different in social dilemma settings. The level of cooperation will be different depending on whether it is mixed-sex or same-sex setting. Maccoby (1990) suggest that women’s same-sex interactions tend to be more cooperative and pro-social, whereas men’s same-sex interactions tend to place greater emphasis on social dominance. From a socio-cultural perspective, men and women spend more time interacting with same-sex during childhood. So, one may expect that women develop a particular style, making interaction among women more cooperative than men among men. In contradiction to this perspective is the evolutionary argument that men are not less cooperative, because during the hunt (in ancestral times) they also need to work together in order to improve their chance on creating offspring. In mixed-sex interactions it is difficult to predict the level of cooperation. The socio-cultural theory expects women to cooperate more because they are more communal, but for men there is a trade-off. Men must convince women that men are dominant, competitive and powerful which suggest that men will deviate more. But on the other hand, one could expect that the chance on offspring with a woman will decrease when he deceives a woman (Balliet et al., 2011). A logical outcome in this situation could be that

cooperation goes up when men (and women) see the other participants as potential for getting offpring. Attractiveness can therefore be suggested to play a role here.

Attractiveness

Evidence shows that attractive people are judged more positively on a wide variety of dimensions (Dion, Berscheid and Walster, 1972) but little research has been done into the role of attractiveness on behaviour during the prisoner’s dilemma. Two studies (Brislin and Lewis,1968| Walster et al., 1966) found a significant correlation between attractiveness and the desire to date. One of the research questions was "How sociable or outgoing does your date seem?” suggesting that attractive people are perceived to be more sociable and therefore more cooperative. From the evolutionary perspective, the correlation seems logical because attractive people have more chance to create offspring because they do better in society. We seek out attractive others because they are more aesthetically pleasing to look at and because we have been taught that ‘what is beautiful is good’ and therefore desirable (Reis et al., 1982). From the same evolutionary perspective it seems logical that in mixed-sex combinations, persons will cooperate more with attractive persons. The cooperation rate with less attractive people will be lower because the willingness to get offspring with this person is less. From a socio-cultural perspective one would expect that the correlation between attractiveness and cooperation will be higher for men with a woman as opponent. Men are more status-assertive in their social orientation. Because a partner’s attractiveness can be a potential social asset, most men will prefer beautiful women as friends (Reis et al., 1982). Because women are less status-assertive and more communal, independently of sex, the correlation between attractiveness and cooperation in playing the PD will be less for women with a man as opponent.

Research method

Procedure

The participants for the actual research were all from the University of Amsterdam. Only Dutch speaking students could enrol via ‘www.creedexperiment.nl’, a website special for experiments. 334 Students (185 men and 149 women with an average age of 22.2) took part in the experiment. For each session a maximum of 14 people could subscribe. Only 12 of these 14 people could participate in the experiment. The participants were divided into two rooms (room A and room B). Participants were not allowed to see participants in the other room. When the number of participants was higher than 12 or not even, participants were asked to leave the experiment voluntary until the number was even and to a maximum of 12, equally divided over the two rooms. If no one left voluntary, participants were randomly picked and sent away. To give the participants random places in both rooms, they had to draw a seating card (A1-A6 for room A and B1-B6 for room B).

The experiment was executed manually and had four different conditions. Part one consisted of 6 rounds. In each round, one participant from both room A and B with a corresponding seat number had to go to a meeting room where they could see each other. Before and after these rounds participants had to fill in a belief assigning which choice they expected their opponent to make on a scale from 0 – 100. A belief of zero means that the opponent will certainly not choose X. In all conditions participants had to play the PD in which they make the choice between playing X and Y, knowing what the consequences are. The order of these steps was different in all four conditions.

1. First condition: baseline treatment ‘NC’ :

First, after reading the instructions all participants had to fill in their belief about the likelihood that the opponent will choose X or Y. Secondly, participants had to make the choice themselves between X or Y. Thereafter, corresponding with their number, one participant from each room went to the meeting room where they saw each other for just 10 seconds while they were not allowed to talk to each other. After they returned to their rooms participants had again to fill in their belief about the likelihood that the opponent would have chosen X or Y.

2. Second condition: silence treatment ‘SC’:

First, after reading the instructions all participants had to fill in their belief about the likelihood that their opponent will choose X or Y. Then, corresponding to their number participants went to the meeting room where they saw each other for 10 seconds while they were not allowed to talk with each other. Third, after they returned to their rooms, participants had to make themselves the choice between X or Y and had to fill in the belief about the likelihood that their opponent would have chosen X or Y again.

3. Third condition: restricted communication (No promises allowed ) ‘RC’:

First, after reading the instructions all participants had to fill in their belief about the likelihood that their opponent will choose X or Y. Second, corresponding to their number participants from each room went to the meeting room where they see each other for two minutes. They were allowed to talk with each other but there was one restriction. The participants were not allowed to make promises which they cannot guarantee. This means that both participants were not allowed to promise to cooperate ore deviate while they could still choose otherwise. This is clearly stated in the instructions. Finally, when they returned to their room participants had to make themselves the choice between X or Y and had to fill in the belief about the likelihood that their opponent would have chosen X or Y again.

4. Fourth condition: unrestricted communication ‘FC’

First, after reading the instructions all participants had to fill in their belief about the likelihood that the opponent will choose X or Y. Second, corresponding to the round number participants went to the meeting room where they see each other for two minutes. They were allowed to talk with each other without restrictions. Finally, when they returned to their room participants had to make themselves the choice between X or Y and had to fill in the belief about the likelihood that their opponent would have chosen X or Y again.

For every condition, there were corresponding instructions (see Appendix 1). These instructions were repeated orally to assure that participants had the same instructions. Participants having questions could raise their hands and their questions were then answered in private. Participants filled out their choices on different sheets, which they had to put in an envelope.

When part one ended, participants in every setting received instructions and corresponding questions which they had to answer on their own (appendix 2). Participants were given enough time to answer these questions. When ready, they had to put answers in an envelope.

Before part three started, one participant of each room had to go to the meeting room to see which question of part two was selected with consecutive payment based on results. This was done by a coin flip. If question two was selected, another coin flip determined which of the two options was paid. When the participants returned to their own room they had to tell the other participants which question had been selected with consecutive payment. Part three existed of a questionnaire (appendix 3) which the participants had to fill in during the time their payment was calculated. When payments were calculated, participants were sent one by one to the meeting room to collect the earned money. Participants of room A and B were not allowed to see each other during this procedure. The payment was calculated as follows: all participants received a show-up fee of 6 euro + the outcome of the PD according to the choices of the couple + one of the selected questions of part two.

Design

When reading the design, one should keep in mind that this research is part of a bigger research. While different perspectives will be researched by colleagues, some parts included in this design where not relevant for this specific research.

In each condition participants had to play a classic one-shot Prisoners Dilemma (PD) of which the assigned values are shown below. Each participant had to make a choice between A and B. The choice of A corresponds to a cooperative strategy, choice B corresponds to a defective strategy. Given that you are ‘you’, the amount of money to be earned in each cell you earn is given in blue.

Figure 3: design prisoners dilemma

The beliefs participants had to fill in, both before and after playing the PD, gives the opportunity to see whether the beliefs of people change after seeing the participant. Beliefs had to be between 0 and 100 with integer steps of 1 (0 and 100 are allowed). When participants assigned 0, it meant that the participant is certain that the other participant will not choose A. When participants assigned belief 100 this meant that these participant were certain that the other participant will choose B. If a participant assigned 50 this meant that the participant did not know what the other participant have chosen or thought that the participant were equally likely to choose A as to choose B.

In part two participants had to answer some additional questions where they could earn more money. These questions were made to measure the risk preferences of the participants. In this paper those results are not taken into account because the risk preferences are not part of this research. So no further analysis will be done on the outcomes. Part three existed of a questionnaire (appendix 3) with some additional questions of which the answers are important for the research. Like general questions about gender, age, study, study year and country of birth of each participant.

After all participants had played the one-shot prisoners dilemma data could be collected and analysis made. During all treatments, the interactions between participants in the meeting room were recorded. Before the instructions of part one were distributed all participants had the chance to leave the room if they were against recording, and they had to sign a consent form in case of no objection. To create robust data, all the interactions were coded individually by four persons. The coding scheme can be found in appendix 4. In case three out of four persons gave the same code to a participant, this code was considered valid. In the case that two codes were coded both by two persons, the professor judged the final code. The analysis of the data will be described in the next section ‘Results’.

Results

After dropping data with problems (appendix 6) a data set remained based on n 308 participants of which 163 men and 145 women. During the experiment, these participants had been divided over the four different treatments (NC, SC, RC and FC). The gender deviation is shown in table 1.

Table 1: distribution of participants in different treatments Treatment % of female Number of participants

NC 50,00% 56 SC 57,14% 84 RC 40,91% 88 FC 41,25% 80 Total 47,08% 308

Figure 4: Percentage of cooperation in four different treatments

N= 308, the blue bars represent men whereas the pink bars represent women

Figure 4 shows an upward trend of cooperation between the four different settings. In the NC treatment, total cooperation rate is 21.4% which means that 21.4% of the participants choose for the cooperative option x while the remaining participants choose option y. Despite the upward trend, the only significant increase between two subsequent treatments is from the RC treatment to the FC treatment ( Mann-Whitney u test P=0.000) where the

0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% NC SC RC FC Male Female 18

cooperation rate increases from 40.9% to 78.8%.

Contradicting the current literature, the differences between male and female cooperation rate is not significant for the first three treatments (Mann-Whitney u test: P value for NC= 0.1967, SC=0.4056 and RC=0.1513). Only the fourth treatment, where full communication is allowed, the women cooperate significantly more (10%-level, P=0.0965). Based on these results, one cannot say that women cooperate higher in all treatments. Still one can see that the cooperation rate is higher in all treatments except for SC-treatment. Question is thus whether other factors may have played a role.

A potential reason for the insignificant difference between the cooperation rate of women and men, and the insignificant increase in cooperation rate in subsequent treatments may be the binary choice system (x or y).the binary choice has large standard errors as a consequence.

Belief

During the experiment, participants had to assign a belief between 0-100 with integer steps of 1. Because of this much more detailed scoring possibility compared to the binary choice s, one might expected that standard errors are smaller. Based on data collected, analysis could be made how convinced participants were about their opponent’s behaviour. Next to that, the accuracy of their beliefs could be measured.

N=308, beliefs before are assigned before interacting with their opponent. Beliefs after are assigned

after the interaction related to the different treatments

0 10 20 30 40 50 60 70 80 NC SC RC FC Beliefsbefore Beliefsafter

Figure 5: average beliefs on a scale 0-100 assigned by participants in different treatments.

Both beliefs before and after show an upwards trend between each treatment. When the participants assign their belief before the PD starts, they have not yet seen their opponent and are therefore not yet influenced by their opponent. For this reason, the belief given by the participants can be seen as their belief of cooperation rate in society. The increase in ‘beliefs before’ between different treatments tells that the participants belief that cooperation is increasing when the interaction possibilities increase. The difference between beliefs before and beliefs after reflects that the participants have seen each other. The beliefs after are therefore influenced by the interaction in each different treatments. 86.04% of the participants changed their beliefs between before and after the interaction.

When participants are able to “predict” what the other participant is going to choose, the beliefs after should be correlated to the behaviour of their opponents more than the beliefs before.

Figure 6 shows that the correlation of ‘beliefs after’ is indeed always higher than ‘beliefs before’, the correlations are however small for the first two treatments. The correlations become more convincing in the RC and FC treatment.

So, participants seem to expect that cooperation is increasing when interaction becomes more intensive. The influence of the interactions is supported by the change in beliefs. More interaction leads to more accuracy is indicated by the increase in correlation.

-0,2 0 0,2 0,4 0,6 NC SC RC FC before after Figure 6: Correlation between participant’s belief (before and after) and decision made by participant’s opponent for all treatments

If one looks at the difference between gender one can see that men seem to expect that the rate of cooperation will be higher when the interaction has been more intensive. This

conclusion is supported by the steeper increase of assigned beliefs before each treatments and because the correlation between beliefs before and decision made by the opponent is higher for men than for women ( 20.93 men > 12.48 women, appendix 7). The beliefs after assigned by both gender shows the same increase between treatments. Although the beliefs after assigned are higher for women, the differences are not significant (Mann-Whitney U test P=0.238). There is no significant difference in the number of men and women changing their beliefs ( Mann-Whitney U test P=0.936). But this does not yet indicate anything about the accuracy of the change in their beliefs.

To check which gender is more accurate or becomes more accurate in the beliefs they assign between treatments, the following method has been applied. If participants are able to signal the behaviour of their opponent, one could expect that the beliefs after assigned by participants are higher when the decision of their opponent will be x compared to their beliefs after when the decision of the opponent will be y. By measuring these conditional beliefs, the accuracy of men and women could be compared for all treatments. Beliefs before will not be taken into account, these beliefs are not influenced by the

opponent yet. For this reason these beliefs could not contribute to the measurement of accuracy.

N=308. The figure on the left: the ‘beliefs before’ assigned by both male and female on a scale of 0-100 for all four treatments.The figure on the right :‘beliefs after’ assigned by both male and female

on a scale of 0-100 for all treatments The linear lines added signifies the increase between

Figure 7: change in beliefs before and beliefs after between all four treatments, for both men and women.

Figure 8: Conditional beliefs after assigned by man and woman in all treatments.

N=308. The first blue bar for all treatments is the average belief after assigned by men when the opponent has chosen x, the second blue bar in every treatment is the average belief after

assigned by men when the opponent has chosen y. The same holds for women (pink bars). = significant difference between average belief after assigned to opponent who chose x and y.

Men seem to be more accurate in the NC treatment while women seem to be more accurate in the SC treatment. However, these results are not significant (Mann-Whitney u test, P >0.05). The results are significant for both gender in the RC-treatment which means that both gender are accurate in predicting the behaviour of the opponent (Mann-Whitney u test: Men: P= 0.0377 and Women: P= 0.0073). But contradicting literature, men are more accurate during the FC-treatment than women. The difference between ‘belief after’ for both x and y are no longer significant for women in the FC- treatment (Mann-Whitney u test: Men: P=0.0002 and Women: P > 0.05). The difference in accuracy between men and women during the FC-treatment is big and convincing to the disadvantage of women. So one cannot say that women are more accurate with their beliefs compared to the men. Instead of analysing the correlation between belief of a participant in the behaviour of the other participant, and the actual behaviour of that other participant. One could also analyse the correlation between participants’ belief of the behaviour of the other participant with the participants own behaviour. The correlation of 0.749 (appendix 5) means that the final decision made by participants is highly predictable by their assigned ‘belief after’. The chance that a participant will choose x increases by 1.09% if the belief after he assigns increases by one point (p=0.000). Figure 9 shows that the correlation is increasing between treatments. 0 10 20 30 40 50 60 70 80 90

Men Women Men Women Men Women Men Women

NC SC RC FC Bel iev es a ft er 22

It is interesting to see that the correlation of belief-before with own choice made is decreasing while the correlation of belief-after with own choice made is increasing. The explanation lies in the increasing interaction possibilities. The more interaction possibilities exist, the more the participant is influenced by the behaviour of the other. Participants tend to have a belief highly correlated with their own choice in the “after” scoring when interaction increases. One can see this clearly comparing the scores in the different settings. The correlation of the belief after is almost the same as the belief before during the NC-treatment. Here the participants did not have any influence yet on the decision made by the other participant. In the following treatments, the interaction influences the beliefs and choices made by the participants. It is logical that the ‘belief after’ in the other three treatments is therefore far more correlated with the decision made by participants themselves because both scores are made after seeing the other participant.

Figure 10: change in correlation (between belief after and own decision) between treatment for both gender and Total.

Figure 10 shows that the correlation between belief after and own decision is higher for men than for women in all treatments. This signifies that men base their decision more on their beliefs based on the information obtained during interaction.

0 20 40 60 80 NC SC RC FC

Men belief after Women belief after 0 0,2 0,4 0,6 0,8 NC SC RC FC before after

Figure 9: correlation between one’s belief of behaviour other participant with own decision

Promises

The only difference between treatments RC and FC is that participants are allowed to make promises towards each other in the FC treatment. For this reason, it is possible to analyse the influence of this factor on the behaviour of the participants very accurate. In the questionnaire, participants had the possibility to rate (between 1 and 4) the extent in which a promise had been made by themselves and by the other participant. In the following table, an overview of these answers will be provided:

Table 2: behaviour of participants based on promise possibility in the FC-treatment

N=80

It became clear that it is difficult for participants to estimate what kind of promise the opponent made. In many cases (43,75%), a participant filled in that the opponent made another promise than the opponent filled in about himself.

The participant will make the decision based on what he beliefs the opponent has promised and based on what his own promise is. For this reason, the behaviour of participants will be measured according to this formed belief.

Table 3: Cooperation rate of a participant dependent on promises made by both him and opponent.

N= 79. For example, 41 participants filled in that a promise to cooperate had been made both by the opponent as by the participant himself. * in the questionnaire the participant filled in that he made and other promise than his opponent. For example: he made a promise to deviate while he believes

that his opponent made a promise to cooperate.

As can be seen above, the cooperation rate is highest when a participant is convinced that both he and his opponent promised to cooperate. 97,65 % of these participants cooperated.

Promise My behaviour other participants behaviour

Promise to cooperate 47 45

Conditional promise (if you,.. then) 21 20

Promise to defect 5 11

Did not make a promise 7 4

Participants beliefs of made promises: Number Cooperation rate participant %

Both promised cooperation 41 97,56%

Both promised deviation 4 75%

Both promised conditional promise 14 64,29%

Both promised nothing 4 75%

Both promised something different* 16 56,52%

It is hard to judge the other results because the number of observations is low. But one can see that regardless of what promises are made, the cooperation rate in the FC-treatment is higher compared to the total cooperation rate in the RC-treatment (figure 4). So one might conclude that the possibility to talk unrestrictedly raises the cooperation rate.

If we compare promises made and actual behaviour between both genders, the interesting case worth mentioning is where men and women make a conditional promise. In that case women show significantly more cooperation then men, 87.5% compared to only 38.46% for men (Mann-Whitney u test, P=0.0314). The only problem here is that it is unknown whether the participants made a conditional promise to cooperate of to deviate. For the other three possible promises made, there is no significant difference in the final choice made by men and women (Mann-Whitney u test, P>0.05). So, one cannot say that one of the gender adheres more to the promises made. There are no significant differences in the number of promises made between both genders (Mann-Whitney u test, P >0.05).

Table 4: difference between gender in promises made and actual cooperation rate.

Guilt

In the questionnaire, participants had to react to the statement ‘it feels bad when you are not meeting the expectations of the other participant ‘on a 1-7 scale in which 1 is considered as not true and 7 is considered as true. The results show a significant upwards trend in guilt between the four different treatments. The correlation between actual decision and guilt is 0.466 (appendix 5), and on average an increase in guilt by 1 on the scale increases cooperation with 11.6%. For this reason guilt can be seen as a good indicator for the choice participants will make.

Promise made self: Total cooperation rate Men Cooperation rate Women Cooperation

cooperate 47 93,60% 26 92,31% 21 95,24% conditional promise 21 57,14% 13 38,46% 8 87,50% promise to deviate 5 60% 3 66,67% 2 50,00% made no promise 7 57,14% 5 60,00% 2 50% total 80 47 72,34% 33 87,88% 25

Figure 11: ‘Guilt on a scale 1-7 measured for male, female in all four treatments.

Women on average assigned significantly more ‘Guilt’ than men (Mann-Whitney u test P=0.000). The average difference in assigned ‘Guilt’ over the treatments is more than one point. If this had been the only indicator to create a difference between men and women, the cooperation rate of women would have been on average 11.6% more in all treatments

Payment

Figure 12 the difference in payments participants earned on average in different treatments.

The total payment between treatments increased 76.5 cents on average. The only significant increase in total payment between two subsequent treatments is between the RC treatment and FC treatment. The increasing trend in payment between treatments is as expected because in figure 1 we already showed an increase in cooperation rate in-between treatments. Because the more cooperative outcome in the PD is best for total welfare, more cooperation means more payment.

Accept for the tiny difference in payment for the SC-treatment, men earned more than women in all treatments. It is however worthwhile mentioning that there is a clear upward trend in payment for women while this is not the case for men. However, the differences in payment between gender are insignificant for all treatments (Mann-Whitney u test, P>0.05). 0 1 2 3 4 5 6 NC SC RC FC Male Female 26 0 2 4 6 8 NC SC RC FC Pa ym en t Male Female total

Hypothesis 1: Men behave differently than women in the prisoners dilemma.

Cooperation is significantly increasing between treatments which support the influence of the social context. Both women and men cooperate more when interaction is increasing. The only significant (10% - level) difference in the cooperation rate exists during the FC-treatment, where women cooperate more. The question is however whether women have good reasons for this high cooperation rate. The results of the ‘beliefs before’ show that men are more aware of the context of the game. They foresee that the cooperation is increasing when the interaction increases. This awareness is also supported by the accuracy of the beliefs assigned by men in comparison with the beliefs assigned by women. Although both women and men are significantly accurate during the RC-treatment, the FC-treatment shows that women are not accurate anymore when the participants have the possibility to interact unrestricted. It seems that they lose the ability to estimate the opponent’s behaviour and just cooperate. That women have the tendency to cooperate more could be supported by the fact that their feelings of guilt (empathy, caring) are significantly higher. The higher correlation between ‘belief after’ and own decision made by men tells that they are able to base their decision more on the obtained information. Although the difference in payment is not significant, one can see that men in the FC-treatment receive higher payments than women.

These results lead to affirmation of the first hypothesis. According to my results, men seem to behave different than women in the prisoners dilemma. Although women increase the cooperation rate in the prisoner’s dilemma, women are less aware of the context of the game, loose accuracy in estimating opponent’s behaviour when the interaction possibilities are high and therefore earn less than men.

Attractiveness

All participants were asked to judge the attractiveness of their opponent on a scale of 1-7 with the remark that a 7 is very attractive and a 4 could be seen as the average attractiveness of the Dutch population. Twelve participants did not judge the attractiveness of their opponent, for this reason these participants were excluded.

Table 5: judgement of attractiveness

N=296

The women were significantly judged more attractive than men, both by men and women (Man Whitney u test P=0.000). Another notable fact is that women judged their opponents as more attractive than men do. Table 5 shows the distribution of attractiveness for both gender and for the total group.

Table 6 judgement of both genders on a scale 0-7

The number of participants judged with 1 or 7 for attractiveness is very low. Because this could result in less significance of outcomes in the following research, judgements were clustered in three groups:

average attractiveness of Judged by women judged by men judged in total

Women 4,535 4,377 4,457

Men 3,96 3,81 3,878

Attractiveness on scale 1-7 Men Women Total

1 7 1 8 2 17 7 24 3 21 13 34 4 67 48 115 5 32 50 82 6 11 19 30 7 1 2 3 Total: 156 140 296 28

Table 7: the distribution of participants over three new attractiveness groups

This part of my research focuses on the influence of attractiveness on decision making: does attractiveness influence decision making? The first step is to investigate whether attractiveness of the opponent influences the beliefs of a participant. Only the scores in the ‘belief after’ situation are researched since this is the only belief possibly influenced by the attractiveness of the opponent.

As can be seen in table 8, there is no clear trend in the influence of attractiveness on beliefs. This is the case regardless of when the opponent is a woman or a man. None of the results shown in table 8 are significant.

Table 8: “belief after” in relation to attractiveness of opponent

N=296

The beliefs which participants assign after interaction are with 51.8% strongly correlated to the beliefs assigned before the interaction (appendix 5). It could therefore be interesting to see whether the change in beliefs before and beliefs after is influenced by attractiveness. Table 9 shows that there are no significant results found in the change of beliefs influenced by attractiveness. For this reason it is logically that the correlations between beliefs and attractiveness shown in appendix 5 are low.

New groups attractiveness: Men Women Total 1: less attractive (originally 1-3) 45 21 66 2: average attractive (originally 4) 67 48 115 3: very attractive (originally 5-7) 44 71 115

Total 156 140 296

Belief when the opponent is: Attractiveness of opponent

1 2 3

Men 49,727 49,776 50,968

Women 39,04 55.121 52.846

Total 46,16 52,85 51,817

Table 9: Average change in belief between before and after (influenced by seeing opponent), categorized by attractiveness

For example, the 8.14 in table 9 means that the belief of men changed (belief after – belief before) on average with 8.14 after seeing less attractive women. N=296.

To analyse whether the beliefs after of participants are more accurate when their opponent is attractive compared to a less attractive opponent, the method to measure accuracy with the use of conditional beliefs has been used. The results are shown in table 10 (appendix 5).

Table 10: accuracy of beliefs when the opponent is a woman/man based on attractiveness categories

The table above shows that the accuracy of beliefs for both X and Y is not influenced by attractivenss in case the opponent is a woman. None of the three groups shows significantly more accuracy compared to the other groups (Mann-Whitney u test, P>0.05). When we look at the different “attractiveness”groups of men, we see that attractiveness group 2 is most accurate for both X (highest average belief) and Y(lowest average belief). However even

change in belief of: Attractiveness of opponent

1 2 3

Men 8,14 5,69 -0,1269

Women 3,636 11,1 6,44

Total 6,636 8,8 2,84

Accuracy of beliefs when the opponent is a Woman

Attractiveness opponent: 1 2 3

choice

opponent: X 57,67 72,1 61,04

Y 26,15 40,97 44

Accuracy of beliefs when the opponent is a Man

Attractiveness opponent: 1 2 3

choice

opponent X 65,65 73,2 60,407

Y 39,7 32,18 43,89

these results are not significant (Mann Withney u test, P>0.05). So based on this research I can not conclude that attractiveness of a opponent does have a significant positive influence on the accuracy of beliefs of the participants. When the accuracy of beliefs was researched between the four different treatmetns, I had to conclude that the number of observations was too small to draw firm conclusions. The potential influence of the indicator attractiveness on beliefs after interaction between participants is now fully elaborated.

The next step has been to see whether decisions made by the participants are influenced by the attractiveness of their opponent. The NC-treatment will be excluded because in that treatment participants made a choice before the opponent was seen, and they could therefore not been influenced by the attractiveness of the opponent.

In the three treatments where attractiveness could play a role in decisions made, significant results were found regarding the influence of the attractiveness of a female opponent on the decision made by participants (table 11). Participants interacting with a very attractive female opponent were on average 12.35% more likely to choose x compared to average attractive female opponents, whereas participants interacting with average attractive female opponents choose x 17.76% more likely compared to less attractive female opponents. The correlation between decisions made and attractiveness of female opponent is therefore logically high (with 19.97 as could be seen in appendix 5). As could be seen in table 11, this upwards trend is significant for women playing with more attractive women while this is not significant for men playing with more attractive women.

Table 11: Shows how cooperative both men and women are playing against women in different attractiveness groups

N= 112. There is a significant increase in cooperation for all opponents playing with more attractive women. In contrast to men, only female opponents show an increase in cooperation when they are playing against more attractive participants.

In table 12 is shown that there is no significant influence of the attractiveness of men on the decisions made by the opponents (men, women and total). Still, one could see that the cooperation rate of women playing with less attractive men is much higher than women playing with average and more attractive men.

Table 12: Shows how cooperative both men and women are playing against men in different attractiveness groups

N= 128. There is no significant increase or decrease in cooperation for all opponents playing with more attractive men (women, men and total).

As expected, in this study the influence of attractive female opponent is proved to be higher than the influence of attractive male opponents. Since it is already elaborated that beliefs could not be used as an explanation for this behaviour, the factors ‘guilt’ and ‘promises’ will now be elaborated to see whether they can explain the results of higher cooperation when interacting with more attractive women.

Guilt

Figure 13: influence of opponent’s (men, women and total) attractiveness on participants ‘guilt disappointment on a scale 0-7

The results of figure 13 are clear positive trend. Participants feel significantly more guilt when their opponent is more attractive (P=0.059). The increase of guilt between attractiveness groups for participants with a woman as opponent is higher than for

0 1 2 3 4 5 6 1 2 3 G ui ltd is ap po in tme nt Attractiveness component Women Men 32

participants with men as opponent. Because feelings of guilt are strongly correlated with the final choice participants make, this factor could partly explain the increase of cooperation when the opponent is more attractive.

Promises

Table 13: overview of promises made dependent on the attractiveness of opponent

We assume that a promise to cooperate can be seen as a positive promise. This assumption is made because the outcome of playing the PD is always higher when your opponent is choosing for the cooperative choice x. It is important to note that the number of promise to cooperate to an attractive opponent is much higher than the number of such promises made to the average attractive and less attractive participants. (3 = 75.76%, 2= 40% & 1= 53,3%). When comparing the number of positive promises made by both men and women playing with the different attractiveness groups, one could see that both men and women make the most positive promises when the opponent is more attractive (table 14). While the number of observations is rather small, the results are not significant.

Table 14: overview of promises made by women and men dependent on the attractiveness of opponent

Promises made by

women attractiveness of the opponent: Promises made by men attractiveness of the opponent:

0 1 2 0 1 2

promise to cooperate 4 4 12 promise to cooperate 5 8 13

conditional promise 2 4 2 conditional promise 3 7 3

promise to deviate 0 1 1 promise to deviate 0 3 0

no promise made 0 0 2 no promise made 2 3 0

Total 6 9 17 Total 10 21 16

Promise made by participant: attractiveness of the opponent:

1 2 3

promise to cooperate 8 12 25

Conditional promise (if you...

Than...) 5 11 5

promise to deviate 0 4 1

no promise made 2 3 2

Total 15 30 33

Payment

The potential influence of attractiveness on all important indicators has now elaborated. The last step left to investigate is whether attractiveness has influence on the final payment which the participants earned during the PD. The NC-treatment will not be taken into account.

The results in figure 14 show that participants on average earn most when they are more attractive than average. There is however not an even upward trend because the average attractive participants earn less than the less attractive participants. The differences between the groups of attractiveness are therefore insignificant (P >0.05). However, the differences are significant within the women groups of attractiveness (P = value 0.0032). One can conclude that the results of this research show that the effect of attractiveness of women on their final payment is convincing. This is not the case for men, there is not a clear trend visible and in contrast to women, less attractive men earn the most in playing the PD.

Hypothesis 2: ‘Attractiveness does have an influence in decision making, this influence is in advance of more attractive people’

The results show that the attractiveness of participants does not have significant influence on the beliefs formed by the opponents. However, attractiveness does have an influence on the factors guilt and promises made. Opponents experience more guilt and make more positive promises when the other participant is more attractive. The influences are however not the same for men as for women. The cooperation rate is significantly higher when interacting with more attractive woman, the consequence is that their payment is significantly higher as well. These two results do not hold for men. The cooperation rate is

Figure 14: Payment earned by participants based on attractiveness

34 0 2 4 6 8 Pa ym en t l ev el o f at tr ac tiv e p ar tic ip an ts

Attractiveness level op participant

Women Men

not significantly higher when interacting with more attractive men, and payment for men does not show an upward trend related to the level of attractiveness.

So, although there seems to be no reason to cooperate more with more attractive women compared to less attractive women, results show that in general the cooperation rate does increase. And since cooperation by the opponent is always to the advance of a participant, one could say that more attractive women perform better than less attractive women. To evaluate the hypothesis, attractiveness does have an influence on the factors ‘guilt’ and ‘promise’. The difference in gender is that women are more influenced by the attractiveness of their opponent. Attractive women do have an advantage compared to less attractive women because their opponents show a higher cooperation rate on average. This does not hold for men.

Conclusion

The results found during the analysis of the data have brought some new insights concerning decision making in the prisoners dilemma, and the influence of important factors. As already found in previous studies, this research also confirms that the cooperation rate is increasing when there are more interaction possibilities. Results show that in the most interactive treatment (the FC-treatment) women cooperate significantly more. Statements made in earlier research that women cooperate more because they are more caring, show more empathy and are focusing more on social aspects could explain this higher cooperation rate as well as the fact that women experience higher levels of guilt and therefore tend to cooperate. Besides this gender differences, the results found the following differences between men and women. First, this research shows that men are more aware of the context and the influences of the interaction. Second, they are able to estimate the intentions of the opponent more accurate in the treatment which simulates the real life situation most, the FC setting. Third, they base their final decision more on their beliefs formed by the obtained information during the interaction. Finally, the payment earned based on decisions made is in favour of men. So based on these results, gender differences are proved to be significant.

This research shows that of the level of attractiveness does have a significant influence on the cooperation rate in the prisoners dilemma. However it seems to be gender specific. Results from this study shows that the attractiveness of men is not convincingly influencing decisions contrary to the attractiveness level of women. Factors like guilt and promise are positively influenced favouring on average more attractive people of both gender. However, the most convincing result that the cooperation rate differs around 30% in favour of the average more attractive participants only holds for women. The consequence for these on average more attractive women is that their payment are around 40% higher than on average less attractive women. The attractiveness does not have an influence on the assigned beliefs by opponents. For this reason, this study seems to support the fact that that people just like to be nice to more attractive women, even while they know that they do not behave differently from average less attractive women.

Discussion.

In the research of Brosig (2002) the accuracy of participant’s beliefs are measured with an overall accuracy rate. The participants had to fill in whether their opponent would choose x or y in a PD. Because in this research participants filled in their belief on a scale 0-100, the measurement method of conditional beliefs is used. Although the method is different, both research methods showed that participants are able to predict the behaviour of their opponents. For further research I would recommend to make a deliberated choice between these methods, best could be to use both.

Although the data set for this research contains 334 participants, data collected were insufficient to measure in some cases. One main reason was that participants had to be divided over four different treatments making every single group rather small. Data collected about the influence of factors such as attractiveness or “promises made” in one single setting could thus offend up with less than ten observations. This was especially the case when participants had also to be divided based on gender.

As discussed earlier, the PD experiment is done for different purposes, some elements like risk valuation, were included but had no additional value for my research. The answers on the questionnaire at the end of the experiment could maybe be different influenced by the additional elements. For example, due to the unnecessary extra length of the experiment, the attractiveness one felt by its opponent could be weakened or forgotten. If this research is reproduced, one should consider leaving these elements out.

Researches of Boone et al (1991a|1991b) indicated that factors like age and culture have significant influence on the cooperation rate of participants. In this research, students participating were mainly between 18 and 24 (75%) years old. In this research participants represented 22 different nationalities, but most (300) were Dutch. The potential influence of these two factors is not taken into account in this research. So one has to be aware when interpreting the outcomes this research especially when conclusions are drawn that differences exist between women and men.

This research applies the one-shot game prisoner’s dilemma. For follow up study I suggest in particular to research the influence of attractiveness on the behaviour of people using the repeated PD methodology. One may expect that attractiveness plays a more

important role when persons interact more frequently with each other and thus may influence decision making to a larger extent.

As indicated in the literature study, the behaviour of participants will be different when they are interacting with an opponent of the same sex compared to an opponent of the other sex. The number of observations was too small to study whether attractiveness has a different influence for same sex groups compared to mixed sex groups. For follow up study, I suggest that research will be done in which only the FC-treatment is used. When more observations are obtained, one could conduct a more specific research about the influence of attractiveness in both groups.

Acknowledgement.

I would like to thank all of the students, family and friends who supported me in conducting my thesis. I am especially thankful to my two professors, Theo Offerman and Jeroen van de Ven, who supported and advised me during the whole process. Tanks to Arthur Schram for reading my theses as the second supervisor. I have also benefited from the useful

collaboration with PhD student Semin and Master students Jacco van Mourik, Aviva Heijmans and Harrie Beek. I wish them good luck with the finalization of their projects.