Communication, Beliefs and Cooperation
in a One-Shot Prisoner’s Dilemma
Master Thesis
Business Economics – Organization Economics Aviva Heijmans - 5876567
Octobre 31, 2014
Abstract
The role of communication on beliefs and cooperation is studied in a one-shot two-person Prisoner’s Dilemma (PD) game. Subjects were divided in four treatment groups based on different communication levels: no, silent, restricted and full communication. Before and after communication subjects were asked to predict the cooperation probability of their opponent. To some extent subjects in the restricted and full communication groups can predict the decision of their opponent in the PD game. Verbal communication significantly increases cooperation. Full communication has the strongest effect. Subjects correctly predict these communication effects. In the full communication group a promise to cooperate is very informative about actual behaviour. Subjects also perceive this as a credible signal of cooperativeness. Women are more likely to cooperate, but subjects do not believe this.
1 Introduction
There are robust findings that pre-play communication increases cooperation in social-dilemma experiments. In a meta-analysis, Sally (1995) showed that
communication is the most important factor that fosters cooperation in Prisoner’s dilemmas (PD). It is unknown which part of communication is responsible for this effect. Earlier studies did not investigate the relative importance of the underlying factors of communication and little research was done on beliefs about others’ behavior in PD situations.
In this paper, we study the role of communication on beliefs and cooperation. The positive effect of communication on cooperation could have different sources. We investigate three possible sources. Non-verbal communication could play a role in fostering cooperation. On the other hand, verbal communication might be a
prerequisite to achieve cooperation and in specific a statement of intended behaviour could be an important factor in increasing cooperation. Further we investigate the relation between subjects’ beliefs about others and differences in cooperation behavior.
Our study differs from existing studies in a few ways. Many studies recognize the positive effect of communication on cooperation. Almost no study focuses on the different features of communication that might explain this effect. Some studies look at subjects’ beliefs about others to examine if subjects can accurately predict the behavior of others. Contrary to those studies we also investigate if cooperation behavior can be explained by subjects’ beliefs.
In our experiment subjects played a one-shot two-person prisoner’s dilemma (PD) game. Before making a decision in this game, subjects were given the possibility to communicate with each other. We defined four treatment groups with different communication conditions: no (NC), silent (SC), restricted (RC) and full
communication (FC). These treatments allowed us to investigate three possible sources of the effect of communication on cooperation. With the SC treatment we investigated the effect of non-verbal communication on cooperation. The RC
treatment allowed us to examine the effect of verbal communication whereas the FC treatment also includes statements of intended behavior. The NC treatment was used as control group. Subjects were asked to state their beliefs about the cooperation
probability of their opponent. Predictions were elicited twice, before and after communication. This experimental design allowed us to investigate how different levels of communication influence cooperation and beliefs. We also estimated the accuracy of subjects’ predictions. Further, we regressed subjects’ behavior on different observable characteristics and compared the outcomes with a regression of subjects’ predictions on the same characteristics. We did this to examine if subjects pick up the right cues of cooperation.
Our most interesting findings relate to subjects’ cooperativeness and their predictions about the cooperativeness of their opponent. We found that subjects can to some extent accurately predict the decision of their opponent in the PD game, but only if they could verbally communicate with their opponent before making their prediction. Verbal communication significantly increases cooperation and full communication has by far the largest effect. We also found that women are more likely to cooperate. Subjects correctly perceive the effect of verbal communication but do not recognize that women are more cooperative. In the FC group, a promise to cooperate is very informative about the subjects’ intended behavior. In the RC group where promises are not allowed, stating that you care about the payoff of the other is a credible signal of cooperative intent. However, subjects do not believe that such a statement is credible while they do believe that a promise is trustworthy. This finding can partly explain the difference in cooperation behavior between these treatments.
The paper is organized as follows. Section 2 discusses the related literature. Section 3 describes the methodology and the hypotheses. Section 4 presents the results, and the final section concludes.
2 Related Literature
Almost all economic models assume that people are purely self-interested and only care about maximizing their own monetary payoff. There is much evidence that this assumption is violated in many situations. If all people are selfish then everybody is predicted to play the dominant strategy in a one-shot PD game, which is to defect, and communication is presumed to have no effect. However there are robust findings that communication increases cooperation. The meta-study of Sally (1995) considers
theory communication is the most important factor in fostering cooperation. Others also showed that communication has a positive influence on cooperation in PD games (Dawes et al., 1977; Bohnet and Frey, 1999; Miettinen and Suetens, 2008).
Most of the literature on the effect of communication on cooperation focuses on one form of communication or unrestricted communication compared to no
communication (e.g. Miettinen and Suetens, 2008; Ellingsen and Johannesson, 2004). Only a few papers pay attention to the different features of communication. Miller et al. (2002), for example, varied with time and symbols to communicate via computers. They found a positive correlation between these variables, labelled as the processing and language complexity, and cooperation. Bohnet and Frey (1999) investigated the effect of silent identification on solidarity in a four-person PD game. They showed that mutual identification of participants compared to anonymity, suffices to increase solidarity. Verbal communication increases solidarity even more. The role of different levels of communication on behavior in an eight-person PD game is investigated by Dawes et al. (1977). They showed that relevant communication (discussion about the dilemma) significantly increased cooperation compared to silent identification and irrelevant communication (discussion of an unrelated topic).
Our paper is most related to the study of Dawes but differs in experimental setup and variables of interest. There are two main differences between our
communication conditions and Dawes’. Firstly, we included a no communication condition to investigate the effect of silent communication on beliefs and cooperation. We assume that non-verbal communication might be sufficient to induce cooperation, which is confirmed by Bohnet and Frey (1999). Secondly, we did not explicitly ask subjects about their intended behavior since it has been shown by Belot et al. (2010) that such statements have no effect on cooperation. Most importantly, while Dawes focuses on correlations between subjects’ own behavior and their predictions about others’ behavior, we investigate if cooperation behavior can be explained by subjects’ beliefs.
Previous research showed that communication is not just cheap talk, meaning that communication does not directly affect the payoffs of the game. Ellingsen and Johannesson (2004) showed that promises in particular are very informative about intended behavior. There is also evidence that a promise to cooperate is highly correlated with cooperation in PD games (Belot et al., 2010; Reed et al., 2012). Belot
promises made in response to an explicit question regarding intended behavior. She found that voluntary promises are credible signals of cooperative intent, while elicited promises are not. This finding might explain the result of Dawes that relevant
communication with declaration of intended behavior did not yield any more cooperation than relevant communication without such a declaration.
One of our main interests in this research is subjects’ predictions about their opponent. Different studies investigated subjects’ beliefs about others (Dawes et al., 1977; Frank et al., 1993; Brosig, 2002). In these studies subjects were given the possibility to communicate before playing a PD game and were asked to predict the decision of their opponents. They all maintain, although to a different extent, that subjects are able to predict their opponent’s decision with an accuracy rate above chance. A main comment on their research is the eliciting of coarse predictions. Subjects were asked to predict the exact decision of their opponents, which is ‘defect’ or ‘cooperate’. The methodology they used for examining the accuracy of the
predictions is also questionable. This is explained well by Belot et al. (2012, p. 248). In recent work Reed et al. (2012) confirmed the above findings. He found that subjects’ predictions are related to their opponent’s decision in a one-shot PD game. In contrary to the earlier mentioned studies Reed used a more precise measure for the predictions, namely a 7-point Likert-type scale. On the other hand, in a bargaining experiment of Ockenfels and Selten (2000) was found that observers were not able to predict the private cost information of subjects by detecting involuntary truth-telling signals. Like Belot, we asked subjects to state probabilistic beliefs. This is a more precise measure than the coarse measure used in previous research and it has another main advantage; it allows us to investigate how subjects form their beliefs. Belot also investigated how subjects formed their beliefs in response to observable
characteristics. In specific, she examined subjects’ predictions about the decisions of players in a PD game on a TV show and compared these findings with an analysis on how these observable characteristics influenced actual behavior. Her results show that women and players who make a voluntary promise are more likely to cooperate and that subjects correctly predict this. Age and attractiveness on the other hand have insignificant effects on cooperation and subjects correctly do not associate these characteristics with cooperativeness. We conducted a similar analysis in this study.
Some researchers addressed the possibility that beliefs about others might be a reflection of own behavior (Dawes et al., 1977; Messé and Sivacek, 1979; Belot et al., 2012). They all find some evidence that subjects’ own behavior effects their
predictions about others. Dawes compared predictions of subjects actually making the decisions in a PD game with predictions of observers. He argues that if subjects’ beliefs reflect own behavior the overall variance of subjects’ predictions should be greater than the variance of the observers’ predictions. This assumption is confirmed for the treatment group in which subjects could communicate about an irrelevant topic. The treatment group in which subjects could communicate about the dilemma yielded no difference in variance. Subjects in the study of Belot were observers predicting the choices of the players in the TV show. After making their predictions subjects were asked to make a choice in a PD game themselves. Subjects’ own decision appeared to be the best predictor of their beliefs. This questions the findings of Dawes since the predictions of the observers in their experiment might also (partly) reflect their own theoretical decision in the PD game.
There are different explanations why subjects might base their predictions of the cooperativeness of their opponent on their own behavior. One explanation for the positive correlation between predictions and own behavior is based on a theory of inequity aversion. Fehr and Schmidt (1999) found that if subjects care about fairness, they are likely to cooperate if they believe that their opponent has the intention to cooperate as well. Messé and Sivacek (1979) discriminate between two mechanisms that might account for the tendency of attributing one’s own behavior onto others; self-justification and false consensus. Subjects might feel they need to justify their own behavior (self-justification) or they might think that everybody else behaves like them (false consensus). Their results are ambiguous; they found some evidence for both explanations. One drawback of their study is the acquaintance period before playing the PD game. They assembled 4 subjects for a very short period of time without communication. Subjects were then matched to one of the others for playing the PD game, but not knowing to whom of the others. Under these conditions it seems very plausible that subjects use their own decision as a reference for their predictions, since they have no or little information about their opponents. Our main interest is not to investigate the causality between beliefs and behavior. However, we considered above findings when making any assumptions on this matter.
3 Methodology
The experiment was conducted in May 2013 at the University of Amsterdam with a total of 334 students from different fields of study. The experiment took place in three different rooms; room A, room B, and a meeting room. Room A and B were used to separate subjects into two even groups. The meeting room was used for a
communication phase during the experiment. In total, 38 sessions were run with a minimum of 4 and a maximum of 12 subjects per session. Subjects were randomly assigned to different sessions as well as to room A or room B.
At the beginning of each session subjects were equally divided over room A and room B. In each room, subjects drew a card with an identification number face down. During the experiment subjects were matched according to their number; number 1 from room A was matched with number 1 from room B, number 2 from room A was matched with number 2 from room B, etc.
The experiment consisted of three parts. In the first part of the experiment subjects were divided in pairs and matched as described. Each pair of subjects played a one-shot prisoner’s dilemma (PD) game. The payoff structure in Euro is depicted in Figure 1. Before making their decision in the PD game, each pair of subjects was sent to the meeting room where they got the opportunity to communicate with each other. After this communication phase, subjects returned back to their room and made a decision to cooperate (option X) or to defect (option Y). We elicited subjects’ beliefs about others by asking them to indicate the cooperation probability of their opponent and to indicate this with a number between 0 and 100 with increments of 1. This was done twice, once before communication and making the decision in the PD game, and once after.
Part two of the experiment consisted of two different questions, regarding social value orientation and risk attitude. The experiment ended with a short
questionnaire regarding subject specific characteristics. Subjects’ earnings were based upon a fixed show up fee, the outcome of the PD game, and the outcome of one of the two questions of part two. The written instructions given to the subjects can be found in the Appendix.
Figure 1 – Payoff table Other X Y X You Y
In this study we suggest three possible sources of the positive effect of communication on cooperation. First, subjects might be able to non-verbally signal their cooperative intentions to others, leading to more cooperation. Second,
information exchanged with verbal communication could persuade subjects to cooperate. Third, a promise to cooperate could give subjects more trust in their opponent’s cooperativeness. Based on these three sources we defined four treatment groups in our experiment: no communication (NC), silent communication (SC), restricted communication (RC), and full communication (FC). The level of
communication during the communication phase depended on the treatment. In the NC group subjects had the opportunity to see each other for 10 seconds but were not allowed to speak. Except for the other treatment groups, in the NC group this
communication phase was after making the decision in the PD game. The SC group was identical to the NC group except that the communication phase took place before making the decision in the PD game. In the RC group subjects had the opportunity to talk to each other for 2 minutes but were not allowed to make any promises. Subjects in this treatment were instructed not to make a statement that could become a lie if afterwards they would make a decision that was not in accordance with that statement. The emphasis here is on could. So even if subjects planned to act in accordance with their statement, they were not allowed to make such a statement. In the FC group there were no restrictions to the communication and subjects could talk to each other for 2 minutes during the communication phase. These treatments
allowed us to investigate the role of non-verbal communication (SC treatment), verbal 8 , 8 0 , 12
cooperation. In all treatment groups, subjects were videotaped during the communication phase.
This experimental setup allowed us to investigate the role of communication on beliefs and cooperation in a one-shot prisoner’s dilemma game. Our main variables of interest are subjects’ decisions in the PD game and their predictions of the
cooperation probability of their opponent before and after communication. Our first hypothesis is based on the large amount of literature showing the positive effect of communication on cooperation.
Hypothesis 1: More communication leads to more cooperation
We assume that if subjects have more possibilities to communicate to each other they will have more information available on which they can base their prediction. This assumption forms our second hypothesis.
Hypothesis 2: More communication leads to a more accurate prediction of the opponent’s cooperativeness
Finally, we examined which observable features influence the decision to cooperate or not and how these features affect subjects’ beliefs.
4 Results
4.1 Descriptive statistics
In total, 334 observations are gathered of which 21 observations are dropped because a problem occurred (e.g. subjects did not adhere to the rules of the experiment). From the NC, SC, RC and FC treatment groups respectively 56, 80, 97 and 80 observations were used for statistical analysis.
To investigate differences in cooperation behavior between the treatment groups we compare the percentages of subjects that cooperated. The cooperation rates
more cooperation in the RC group (p = 0.012; t-test, one-tailed) and in the FC group (p = 0.000; t-test, one-tailed). The same holds compared to the SC group and there is no difference in cooperation between the NC and SC groups. Also compared to the RC group there is significant more cooperation in the FC group (p = 0.000; t-test, one-tailed). These findings confirm Hypothesis 1.
Beliefs are important to investigate if they are related to subjects’ decisions. Therefore we compare the beliefs of subjects who cooperated with the beliefs of subjects who defected. Figure 3 shows that subjects hold different beliefs about the cooperation probability of their opponent given their own decision in the PD game. Subjects who cooperate show a higher mean prediction after communication than subjects who defect. This difference is significant in all treatment groups.
To investigate how different levels of communication influence beliefs, we look at the beliefs before and after communication. Mean predictions of the
cooperation probability are different before and after the communication phase (Figure 4). Compared to the mean prediction before communication, the mean prediction after communication is lower in the NC and SC groups and higher in the RC and FC groups. It seems that subjects change their beliefs downwards in the NC and SC groups while upwards in the RC and FC groups. But the difference in prediction, before and after communication, is only significant in the FC group (p = 0.000; t-test, two-tailed). The mean prediction before communication in the FC
21% 28%
39%
79%
NC SC RC FC
Treatment
group is significantly different from the other treatment groups (p < 0.05; t-test, two-tailed). The same holds for the RC group compared to the NC group but only at the 10% significance level. This indicates that ‘knowing’ the possibility of verbal communication influences subjects’ beliefs on forehand. When looking at the mean predictions after communication, the communication in the RC and FC groups positively influences beliefs compared to the other treatment groups (p < 0.05; t-test, two-tailed). 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% NC SC RC FC Treatment
Figure 3 - Mean predictions after
communication
Cooperate Defect Subject decision: 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% NC SC RC FC TreatmentFigure 4 - Mean predictions
Before
communication After
If subjects are rational in their beliefs, overall predictions of the cooperation probability before communication should equal overall predictions after
communication. Table 1 shows the mean predictions before and after communication per treatment group. In the NC, SC and RC groups, subjects are rational in their beliefs (p > 0.10; t-test, two-tailed). But in the FC group predictions are biased in a systematically way (p = 0.000; t-test, two-tailed).
Table 1 - Cooperation probabilities
NC SC RC FC
Prediction before communication 0.38 0.41 0.45 0.53 (0.23) (0.23) (0.26) (0.23) Prediction after communication 0.34 0.39 0.49 0.72 (0.27) (0.3) (0.35) (0.29)
Actual cooperation 0.21 0.28 0.39 0.79
(0.41) (0.45) (0.49) (0.41)
Standard deviations are in parentheses.
If we compare the actual cooperation rate with the predicted cooperation after communication (Table 1), then subjects tend to overestimate the cooperation probability in the NC, SC, and RC groups (p < 0.01; t-test, two-tailed), while underestimating the cooperation probability in the FC group (p = 0.026; t-test, two-tailed).
4.2 Adjustment of beliefs
More information is available when subjects have more communication possibilities. We expect that subjects who have more information adjust their beliefs more. To examine this we look at the mean absolute adjustment per treatment. Table 2 shows that subjects adjust their beliefs more when they have more communication
possibilities, although the adjustment is somewhat higher in the RC group compared to the FC group. The differences in adjustment across treatment groups are significant (p < 0.01; t-test, two-tailed), except from the difference between the NC and SC groups and the difference between the RC and FC groups.
If subjects can to some extent predict their opponent’s behavior we expect that the adjustment of subjects’ beliefs is related to their opponent’s decision. In specific, we expect a positive adjustment of beliefs when the opponent cooperated and a negative adjustment of beliefs when the opponent defected. In the second and third row of Table 2 the mean adjustments of subjects’ beliefs given their opponent’s decision are shown. On average subjects seem to update their belief in the right direction; upward when their opponent cooperates and downward when their opponent defects.
Table 2 - Adjustment of beliefs
Opponent decision NC SC RC FC
Mean absolute
adjustment 15.46 18.33 27.64 26.99
(14.72) (17.73) (19.79) (22.3)
Mean adjustment Cooperate 0.45 3.32 11.57 24.63
(23.39) (32.03) (31.31) (27.63) Mean adjustment Defect -3.02 -3.84 -0.78 -4.29
(21) (22.57) (34.8) (27.01)
The adjustment of beliefs is calculated by the prediction after communication minus the prediction before communication. Standard deviations are in parentheses.
It is difficult to interpret the magnitude of the mean adjustment. For example, if your prediction before communication is 99%, your prediction after communication is 100% and your opponent cooperates. In this case, your adjustment of the prediction is in the right direction (positively), your prediction after communication is totally accurate, but the adjustment of the prediction is only 1. Because we cannot easily interpret the magnitude of the adjustment, we will look at the percentage of subjects that correctly versus incorrectly update their belief given their opponent’s decision. Were correctly means that subjects update their prediction positively when their opponent cooperates and negatively when their opponent defects. Figure 5 shows that in the treatment groups with more communication possibilities a higher percentage of subjects correctly adjust their prediction.
The adjustment of subjects, whose prediction is updated from 100% before communication to 100% after communication when their opponent cooperates, is classified as correct. The same holds for the adjustment of subjects who updated their prediction from 0% before communication to 0% after communication when their opponent defects.
4.3 Accuracy of beliefs
We now investigate if subjects can predict whether their opponent is going to cooperate or not. As a measure of the accuracy of the predictions we use the following formula:
𝜆𝜆 = Ρ�𝑋𝑋��𝑋𝑋� − Ρ�𝑋𝑋��𝑌𝑌� (1)
where Ρ�𝑋𝑋��𝑋𝑋� is the mean prediction of the cooperation probability of the opponent given that the opponent cooperates and Ρ�𝑋𝑋��𝑌𝑌� is the mean prediction of the cooperation probability of the opponent given that the opponent defects. We use subjects’ prediction after communication to estimate the accuracy of the prediction, because we assume that subjects base their decision on their prediction after
communication rather than their prior belief.
Shortly we explain the implications of equation (1). If subjects are totally accurate, they should predict a 100% cooperation probability when their opponent cooperates and a 0% cooperation probability otherwise. In this case, Ρ�𝑋𝑋��𝑋𝑋� = 100, Ρ�𝑋𝑋��𝑌𝑌� = 0 and 𝜆𝜆 = 100 accordingly. When subjects are totally inaccurate 𝜆𝜆 = -100. So 𝜆𝜆 has a range from -100 till 100 and the higher 𝜆𝜆, the more accurate the overall predictions. 0% 20% 40% 60% 80% 100% NC SC RC FC Treatment
Figure 5 - Adjustment of beliefs
Correctly Incorrectly
If subjects can accurately predict their opponent’s behavior, then subjects’ belief after communication should be related to their opponent’s decision.
Accordingly, subjects’ prediction of the cooperation probability of their opponent should be higher for opponents that cooperate compared to the prediction for opponents that defect. Figure 6 shows the mean predictions after communication given the opponent’s decision and Table 3 shows the outcome of equation (1) per treatment group.
Table 3 - Accuracy of predictions
NC SC RC FC
λ 5,9 3,8 21,2*** 37,5***
*** corresponds to 1% significance level.
Figure 6 shows that subjects have a higher mean prediction of the cooperation probability after communication when their opponent cooperates compared to when their opponent defects, as a result λ is positive in all treatment groups, which is also shown in Table 3. Except from the SC group, the more communication possibilities there are, the higher is λ and so the more accurate the predictions are. In the RC and FC groups λ is significant which means that only in these treatment groups subjects can to some extent predict the decision of their opponent. These estimates confirm Hypothesis 2. 0% 20% 40% 60% 80% 100% NC SC RC FC Treatment
Figure 6 - Mean predictions after
communication
Cooperate Defect Opponent decision:
4.4 Causality beliefs and decision
The causality between subjects’ beliefs and their decisions is not straightforward. It might be convenient to assume that subjects base their decision on their predictions about the behavior of others. However, some researchers found that beliefs are partly influenced by one’s own behavior.
In this paragraph we investigate if subjects’ beliefs are a reflection of their own decision rather than a reflection of the decision of their opponent. We do this by looking at the adjustment of subjects’ beliefs given their opponent’s decision and their own decision (Table 4). If subjects’ beliefs reflect their opponent’s behavior, we expect a positive adjustment of beliefs when their opponent cooperates and a negative adjustment otherwise. On the contrary, if subjects’ beliefs reflect their own decision, we expect a positive adjustment of beliefs when they cooperate and a negative adjustment otherwise.
Table 4 - Mean adjustment of beliefs
Opponent decision Subject decision NC SC RC FC
Cooperate Cooperate 5 24.63 19.8 27.45 (21.21) (29.87) (27.14) (26.7) Cooperate Defect -0.56 -8.86 1.88 2.14 (24.93) (27.17) (33.86) (26.28) Defect Cooperate 5 8.07 20 17.14 (31.12) (18.58) (23.01) (23.07) Defect Defect -5.15 -7.64 -9.69 -19.3 (17.45) (22.59) (35.38) (18.3)
The adjustment of beliefs is calculated by the prediction after communication minus the prediction before communication. Standard deviations are in parentheses.
It seems that the adjustment of subjects’ beliefs are more linked to subjects’ own decision than to the decision of their opponent. Subjects’ prediction should be adjusted positively when their opponent cooperates and negatively when their opponent defects, assuming that subjects can predict their opponent’s decision. However, Table 4 shows that subjects’ adjustments of beliefs move in these directions given their own decision; subjects with a positive adjustment of beliefs cooperate and subjects with a negative adjustment defect. Although in the RC and FC groups
subjects update their belief positively when their opponent cooperates, irrespective their own decision. However, the mean adjustment is much lower when subjects’ own decision is to defect, suggesting that even here subjects’ beliefs are related to their own decision.
Our finding that the adjustment of beliefs seems to be linked to subjects’ own decision rather than their opponent’s decision does not directly imply the causality of this link. We do not know whether subjects adjust their belief based on their own decision or that subjects base their decision on an incorrect adjustment of their belief. However, given the consistency of the direction of the adjustment and subjects’ own decision, in particular in the NC and SC groups in which subjects cannot predict their opponent’s decision, we argue that subjects partly base their belief on their own decision rather than systematically adjust their belief incorrectly.
Our experimental set up does not allow us to fully investigate the causality between beliefs and decisions and to what extent beliefs influence decisions or decisions influence beliefs, but we found some evidence for mutual causality. In the RC and FC groups subjects are to some extent able to predict their opponent’s
decision, which indicates that subjects’ predictions are not entirely based on their own decision. On the other hand, Table 4 shows some evidence that subjects’ beliefs are influenced by their own decision.
4.5 Regression analysis
We now investigate which factors influence the decision to cooperate and which factors subjects perceive as cooperative traits. We consider the decision to cooperate as a function of observable characteristics. The same we do for subjects’ prediction after communication. We use a probit model to predict the decision to cooperate, where the dependent variable is denoted as 𝑌𝑌 = 1 if a subject cooperates and 𝑌𝑌 = 0 otherwise. We estimate the following equation:
𝑃𝑃𝑃𝑃(𝑌𝑌 = 1|𝑋𝑋) = Φ(X’β) (2)
after communication of subject i regarding the cooperation probability of opponent j, 𝑝𝑝𝑖𝑖𝑖𝑖, is modelled as follows:
𝑝𝑝𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛽𝛽’𝑋𝑋𝑖𝑖+ 𝜀𝜀 (3)
where 𝛼𝛼 is a constant, 𝑋𝑋j is a vector of observable characteristics of opponent
j, and 𝜀𝜀 is an error term. 𝛽𝛽can be interpreted as the average adjustment of the prediction in response to a signal 𝑋𝑋.
Since we are interested in factors that might influence the decision to cooperate as well as subjects’ beliefs about their opponent, we only include
observable characteristics of the subjects. There might be other factors that influence subjects’ decision to cooperate like risk attitude and social value orientation, but since these characteristics are not observable they are not likely to influence subjects’ beliefs about their opponent. We suggest that subjects might discriminate on the basis of gender, race and/or attractiveness. Whereas women are found to be more likely to cooperate (Belot et al., 2012), we expect that race and attractiveness might affect subjects’ beliefs about their opponent rather than affecting their cooperativeness. Therefore we control for the following observable characteristics in both equations: gender, attractiveness and skin colour. The variable attractiveness is measured by asking subjects how attractive they think their opponent is on a scale from 1 (very unattractive) to 7 (very attractive). Skin colour is judged by the researchers by watching the videotapes. The variable used is ethnicminority which is coded 1 if a subject is non-white and 0 otherwise. Since we are particularly interested how communication affects cooperation and beliefs, we also included the type of communication. We did not include age in our model, since there are almost no differences in age among the subjects in our sample.
The results are presented in Table 5. Column 1 reports which variables influence the decision to cooperate. The coefficients in this column are the marginal effects of the probit estimates. Verbal communication significantly enhances
cooperation. In the RC and FC groups, subjects are respectively 22 and almost 60 percentage points more likely to cooperate compared to subjects in the NC group. Women are 9 percentage points more likely to cooperate than men. Silent
cooperation behavior. Column 2 shows how the same variables influence subjects’ prediction after communication. Subjects correctly perceive the positive relationship between verbal communication and cooperation. In the RC group subjects have a significantly higher prediction after communication compared to subjects in the NC group. In the FC group this prediction is even higher. Subjects do not believe that women are more cooperative, although this correlation is only significant at the 10% significance level in the actual data. But they correctly assume that silent
communication, attractiveness and skin colour have insignificant effects on cooperation.
Table 5 - Actual and perceived cues of cooperation
Dependent variable: Actual realization Beliefs
Cooperation decision (1) (2) Communication: SC .07 (.08) .06 (.05) RC .22 (.08)*** .17 (.05)*** FC .59 (.08)*** .38 (.05)*** Female .09 (.05)* .02 (.04) Attractiveness .03 (.02) .02 (.02) Ethnic minority -.04 (.08) .04 (.05) Number of observations 267 268 R2 0.19
Standard errors are in parentheses. * and *** correspond to 10% and 1% significance levels, respectively. Column 1: probit estimates, marginal effects. Column 2: OLS estimates. Standard errors are robust and clustered by subject pair.
We now turn to the RC and FC groups in specific, by looking at the role of communication in these treatment groups. The results show that the type of
communication in the RC and FC groups positively influences cooperation and beliefs. We investigate the content of the communication to see which part of communication drives these results. Equation (2) and (3) are used to predict
cooperation behavior and beliefs in the RC and FC groups. For both treatment groups we use the same control variables as before: gender, attractiveness and skin colour. Previous research showed that promises are positively correlated with cooperation behavior. Since subjects were only aloud to make promises in the FC group, we add the variable promise for the FC group, which is 1 if a subject makes a promise to
content of the communication in the RC group, we found some subjects saying that they care about the payoff of the other. We suggest that making such a statement might be a good alternative for saying that you are willing to cooperate. Therefore, in the RC group we add the variable caring, which is 1 if a subjects says that he/she cares about the payoff of the other in some kind of way and 0 otherwise.
Table 6 - Actual and perceived cues of cooperation
RC treatment FC treatment Dependent variable: Actual realization Beliefs Actual realization Beliefs
Cooperation decision (1) (2) (3) (4) Female .2 (.1)** -.03 (.09) .08 (.09) -.01 (.06) Attractiveness .05 (.06) .04 (.04) -.04 (.04) -.007 (.04) Ethnic minority .2 (.16) .15 (.07)** .1 (.11) -.04 (0.1) Caring .43 (.23)* .22 (.16) Promise .39 (.06)*** .4 (.12)*** Number of observations 73 74 62 62 R2 0.06 0.28
Standard errors are in parentheses. *, ** and *** correspond to 10%, 5% and 1% significance levels, respectively. Column 1 and 3: probit estimates, marginal effects. Column 2 and 4: OLS estimates. Standard errors are robust and clustered by subject pair.
Table 6 shows the actual cues (Column 1 and 3) and perceived cues (Column 2 and 4) of cooperation in the RC and FC groups. Women are more likely to cooperate in the RC group but not in the FC group. Subjects do not perceive women to be more cooperative in these treatment groups. The coefficients for female are even negative but not significant. In both treatment groups, attractiveness and skin colour have no significant effect on cooperation. Although in the RC group, subjects believe non-white opponents to be more cooperative than non-white opponents. Our main variable of interest in the RC group, caring, positively influences cooperation. Subjects who say that they care about the other’s payoff are 43 percentage points more likely to
cooperate. But subjects do not perceive such a statement to be correlated with cooperativeness. As we expected, a promise to cooperate in the FC group has a positive effect on cooperation. Subjects who make such a promise are 39 percentage points more likely to cooperate. They also correctly believe that a promise to
5 Conclusion
Our results support the existing literature on the positive effect of communication on cooperation. Verbal communication significantly increases cooperation compared to no and silent communication. Unrestricted communication, allowing subjects to make promises, has by far the strongest effect on cooperation rates. Our data do not confirm the findings of Bohnet and Frey (1999) that silent communication is sufficient to increase cooperation. This discrepancy might be explained by a difference in group size; Bohnet and Frey investigated a four-person PD game compared to our two-person PD game. With silent identification subjects might experience more social pressure or sanctions in a group of four compared to a group of two, leading to more cooperation.
The eliciting of beliefs allowed us to investigate subjects’ predictions of the cooperation probability of their opponent. Subjects’ predictions are related to their decision in the PD game. Subjects who cooperate have a higher mean prediction after communication than subjects who defect. We found this in all treatment groups. A priori, the knowledge of having the opportunity to verbally communicate influences subjects’ beliefs about the cooperativeness of their opponent positively. Verbal communication itself compared to no and silent communication increases trust in the cooperativeness of others even more.
We asked subjects to state their prediction before and after communication to examine the change in their beliefs. Subjects adjusted their prediction more when they had more communication possibilities. The adjustment of beliefs is highest in the RC and FC groups compared to the other treatment groups. In all groups, subjects on average update their belief in the right direction; they change their prediction
positively when their opponent cooperates and negatively otherwise. In the treatment groups with more communication possibilities a higher percentage of the subjects correctly changed their beliefs given the decision of their opponent.
We investigated if subjects could accurately predict the decision of their opponent. Subjects are to some extent able to predict the behavior of their opponent, but only when they have the opportunity to verbally communicate with each other. In
beliefs instead of their actual beliefs. This would bias our estimates to chance level. In this case the accuracy of the predictions measured is a lower bound of the actual accuracy that can be achieved.
A regression analysis allowed us to examine actual cues of cooperation and subjects’ perception of these cues. We found that verbal communication has the strongest effect on cooperation. Like Belot et al. (2012) we also found that women are more cooperative. Subjects correctly perceive the positive effect of verbal
communication, but they do not believe that women are more cooperative. Silent communication, attractiveness and skin colour have insignificant effects and subjects correctly do not associate these factors with cooperativeness. In an attempt to explain the effect of verbal communication on cooperation we investigated the content of the information. Promises to cooperate are very informative about subjects’ intended behavior and they are also perceived as a credible signal of cooperativeness. These findings confirm the results of Belot et al. (2010, 2012). In the RC group, stating that you care about the payoff of the other appeared to be a good alternative for making a promise to cooperate. However, subjects do not recognize that such a statement is related to cooperative behavior. This finding can partly explain the higher cooperation rate in the FC group compared to the RC group.
We found some evidence that behavior is a consequence of beliefs about others, given the accuracy of the predictions. However we also found some evidence of the reverse relation; beliefs about others are a reflection of own behavior. We did not investigate to what extent beliefs influence behavior and behavior influences beliefs. In specific, it will be beneficial to investigate to what extent the causal relation is mutual under different conditions. Based on our results we carefully hypothesize that subjects with less information about their opponents, i.e. less
communication possibilities, might base their assumptions about others more on their own behavior than subjects with more information about their opponents, i.e. more communication possibilities. We suggest that in the situation with minimal
information beliefs about others are mainly influenced by own behavior. In the situation with maximal information own behavior is primarily influenced by beliefs about others. Findings from Dawes et al. (1977) strengthen this assumption. Further studies are needed to investigate this hypothesis.
References
Belot, M., V. Bhaskar, J. van de Ven, 2010. Promises and cooperation: Evidence from a TV game show. Journal of Economic Behavior and Organization, 73 (3), 396-405.
Belot, M., V. Bhaskar and J. van de Ven, 2012. Can observers predict
trustworthiness? Review of Economics and Statistics, 94 (1), 246–259. Bohnet, I. and B. S. Frey, 1999. The sound of silence in prisoner’s dilemma and
dictator games. Journal of Economic Behavior and Organization, 38 (1), 43-57.
Brosig, J., 2002. Identifying cooperative behavior: Some experimental results in a prisoner’s dilemma game. Journal of Economic Behavior and Organization, 47 (3), 275–290.
Dawes, R. M., J. McTavish and H. Shaklee, 1977. Behavior, communication, and assumptions about other people’s behavior in a commons dilemma situation. Journal of Personality and Social Psychology, 35 (1), 1-11.
Ellingsen, T. and M. Johannesson, 2004. Promises, threats and fairness. The Economic Journal, 114 (495), 397–420.
Fehr, E. and K. M. Schmidt, 1999. A theory of fairness, competition, and cooperation. Quarterly Journal of Economics, 114 (3), 817– 868.
Frank, R., T. Gilovich and D. Regan, 1993. The evolution of one-shot cooperation: An experiment. Ethology and Sociobiology, 14 (4), 247–256.
Messé, L. A. and J. M. Sivacek, 1979. Predictions of others’ responses in a mixed-motive game: self-justification or false consensus? Journal of Personality and Social Psychology, 37 (4), 602–607.
Miettinen, T. and S. Suetens, 2008. Communication and guilt in a prisoner’s dilemma. Journal of Conflict Resolution, 52 (6), 945-65.
Miller, J. H., C. T. Butts and D. Rode, 2002. Communication and cooperation. Journal of Economic Behavior and Organization, 47 (2), 179-195.
Ockenfels, A. and R. Selten, 2000. An experiment on the hypothesis of involuntary truth-signalling in bargaining. Games and Economic Behavior, 33 (1), 90–116. Reed, L. I., K. N. Zeglen and K. L. Schmidt, 2012. Facial expressions as honest
signals of cooperative intent in a one-shot anonymous prisoner’s dilemma game. Evolution and Human Behavior, 33 (3), 200-209.
Appendix: Instructions
(Instructions in English are available upon request)
Instructies
Hartelijk dank voor je deelname aan het experiment van vandaag. We vragen je om de volgende instructies zorgvuldig te lezen.
Aan het begin van het experiment ontvang je €6 voor je deelname. Gedurende het experiment heb je de mogelijkheid om meer geld te verdienen, afhankelijk van je eigen keuzes en de keuzes van andere deelnemers. Aan het einde van het experiment zal je privé uitbetaald worden.
Je mag niet met andere deelnemers communiceren gedurende het experiment, tenzij we daar expliciet toestemming voor geven. We vragen je ook om je mobiele telefoon niet te gebruiken gedurende het experiment.
Vergeet niet je nummer op elk antwoordvel te schrijven.
Het experiment bestaat uit drie delen. Hier volgen de instructies voor het eerste deel. De instructies voor de andere delen worden later uitgedeeld.
Deel 1
NC, SC, RC, and FC treatments:
Er is een even aantal deelnemers in dit experiment. We hebben de deelnemers over twee kamers verdeeld. Elke deelnemer in de ene kamer wordt op willekeurige wijze gekoppeld aan een deelnemer in de andere kamer.
In deel 1 is het aantal ronden gelijk aan het aantal paren dat deelneemt. In elke ronde neemt een koppel van deelnemers hun beslissingen. Het nummer op je kaart toont de ronde waarin je je beslissing zal nemen.
Elke persoon maakt een keuze tussen X en Y. Als jij en de andere persoon beide X kiezen, ontvangt elk van jullie 8 euro. Als jij Y kiest en de andere persoon X kiest, dan ontvang jij 12 euro en de andere persoon niets. Als jij X kiest en de andere persoon Y kiest, dan ontvang jij niets en de andere persoon 12 euro. Als jij en de andere persoon beide Y kiezen, ontvangt elk van jullie 4 euro.
De mogelijke beslissingen en verdiensten worden ook in de volgende tabel getoond. In elke cel van de tabel toont het eerste getal het bedrag in euro dat jij verdient (in blauw) en het tweede getal het bedrag in euro dat de ander verdient (in oranje).
De andere deelnemer
X Y
Jij X 8, 8 0, 12
Y 12, 0 4, 4
Jij en de andere deelnemer nemen jullie beslissingen gelijktijdig, zonder dat je weet welke beslissing de ander neemt.
Voordat je je beslissing neemt, vragen we je de keuze van de ander te voorspellen. We vragen je hoe waarschijnlijk jij het vindt dat de andere persoon X kiest op een schaal van 0 tot 100. Als je 0 aangeeft, betekent dit dat je er zeker van bent dat de ander X niet kiest. Als je 100 aangeeft, betekent dit dat je er zeker van bent dat de ander X wel kiest. Als je 50 aangeeft, betekent dit dat je het niet weet en dat je het even waarschijnlijk acht dat de ander X kiest als dat de ander Y kiest. Je kan elk getal tussen 0 en 100 aangeven in stapjes van 1 (0 en 100 zijn ook toegestaan).
NC treatment:
Nadat je je voorspelling hebt gegeven vragen we je een keuze tussen X en Y te maken.
In de ronde waarin het jouw beurt is een beslissing te nemen, zullen we jou en de deelnemer van de andere kamer vragen naar de ontmoetingskamer te gaan. Dit gebeurt nadat je je voorspelling hebt gegeven en nadat je je beslissing tussen X en Y hebt genomen.
Als je in de ontmoetingskamer bent, hebben jij en de andere deelnemer de
mogelijkheid om elkaar te zien, maar je mag niet met elkaar spreken. Dit deel zal met video opgenomen worden.
Als 10 seconden verstreken zijn, zul je naar deze kamer teruggaan. Hieronder volgt een overzicht van de stappen.
SC treatment :
In de ronde waarin het jouw beurt is een beslissing te nemen, zullen we jou en de deelnemer van de andere kamer vragen naar de ontmoetingskamer te gaan. Dit gebeurt nadat je je voorspelling hebt gegeven maar voordat je je beslissing neemt tussen X en Y.
Als je in de ontmoetingskamer bent, hebben jij en de andere deelnemer de
mogelijkheid om elkaar te zien, maar je mag op geen enkele manier met elkaar communiceren. Dit deel zal met video opgenomen worden.
Als 10 seconden verstreken zijn, zul je naar deze kamer teruggaan en een keuze tussen X en Y maken.
Hieronder volgt een overzicht van de stappen.
1
•Je doet een voorspelling over de keuze van de ander
2
•Je kiest gelijktijdig met de ander tussen opties X en Y
3
•Je gaat naar de ontmoetingsruimte en ziet de ander gedurende 10 seconden
1
•Je doet een voorspelling over de keuze van de ander
2
•Je gaat naar de ontmoetingsruimte en ziet de ander gedurende 10 seconden
3
•Je komt terug in deze kamer en kiest gelijktijdig met de ander tussen opties X en Y
RC treatment:
In de ronde waarin het jouw beurt is een beslissing te nemen, zullen we jou en de gekoppelde deelnemer van de andere kamer vragen naar de ontmoetingskamer te gaan. Dit gebeurt nadat je je voorspelling hebt gegeven maar voordat je je beslissing neemt tussen X en Y.
Als je in de ontmoetingskamer bent, hebben jij en de andere deelnemer de mogelijkheid om twee minuten met elkaar te praten.
De communicatie is niet aan regels gebonden. De enige restrictie is dat je geen uitspraak mag doen die een leugen zou kunnen worden als je na afloop een keuze zou maken die niet in overeenstemming met die uitspraak is. Bijvoorbeeld, je mag niet zeggen “ik beloof X te kiezen”, omdat dat een leugen zou worden als je in plaats daarvan Y kiest. Merk op dat het criterium is dat de uitspraak een leugen zou kunnen worden, en zelfs als je van plan bent om je aan je uitspraak te houden, mag je zo’n uitspraak niet doen.
De communicatie tussen jou en de andere deelnemer zal op video worden opgenomen.
Na deze twee minuten zul je naar deze kamer teruggaan en een keuze tussen X en Y maken.
Hieronder volgt een overzicht van de stappen.
FC treatment:
In de ronde waarin het jouw beurt is een beslissing te nemen, zullen we jou en de deelnemer van de andere kamer vragen naar de ontmoetingskamer te gaan. Dit gebeurt nadat je je voorspelling hebt gegeven maar voordat je je beslissing neemt
1
•Je doet een voorspelling over de keuze van de ander
2
•Je gaat naar de ontmoetingsruimte en praat 2 minuten met de ander
3
•Je komt terug in deze kamer en kiest gelijktijdig met de ander tussen opties X en Y
Als je in de ontmoetingskamer bent, hebben jij en de andere deelnemer de mogelijkheid om twee minuten met elkaar te praten.
De communicatie is niet aan regels gebonden.
De communicatie tussen jou en de andere deelnemer zal op video worden opgenomen.
Na deze twee minuten zul je naar deze kamer teruggaan en een keuze tussen X en Y maken.
Hieronder volgt een overzicht van de stappen.
NC, SC, RC, and FC treatments:
Je zal je keuze tussen X en Y slechts één keer maken. Als alle deelnemers hun beslissing hebben genomen en alle deelnemers terug zijn gekomen van de
ontmoetingskamer, zal het tweede deel van het experiment van start gaan waarin we je enkele additionele vragen stellen. Je verdiensten in het tweede deel zullen van geen enkele beslissing in het eerste deel afhangen en je zult niet meer aan dezelfde
deelnemer worden gekoppeld.
Aan het einde van het experiment, zullen we de beslissing van de andere deelnemer aan je onthullen en je privé betalen. Deelnemers zullen één voor één gevraagd worden het gebouw te verlaten.
Steek alsjeblieft je hand op als je een vraag hebt.
1
•Je doet een voorspelling over de keuze van de ander
2
•Je gaat naar de ontmoetingsruimte en praat 2 minuten met de ander
3
•Je komt terug in deze kamer en kiest gelijktijdig met de ander tussen opties X en Y
Antwoordvel 1
Je nummer
____________
Geef op een schaal van 0 tot 100 aan hoe waarschijnlijk je het vindt dat de andere persoon optie X kiest.
Als je 0 aangeeft, betekent dit dat je er zeker van bent dat de ander X niet kiest. Als je 100 aangeeft, betekent dit dat je er zeker van bent dat de ander X wel kiest. Als je 50 aangeeft, betekent dit dat je het niet weet en dat je het even waarschijnlijk acht dat de ander X kiest als dat de ander Y kiest.
Je kan elk getal tussen 0 en 100 aangeven in stapjes van 1 (0 en 100 zijn ook toegestaan).
Voorspelling dat de andere persoon X kiest:
Antwoordvel 2
Je nummer
____________
Maak alsjeblieft je keuze tussen X en Y:
Doe dit formulier alsjeblieft in de envelop als je klaar bent.
Als het jouw beurt is om naar de ontmoetingskamer te gaan, zullen we je roepen.
Antwoordvel 3
Je nummer
____________
Jij en de andere deelnemer hebben jullie keuzes gemaakt.
Geef weer op een schaal van 0 tot 100 aan hoe waarschijnlijk je het vindt dat de andere persoon X kiest.
Als je 0 aangeeft, betekent dit dat je er zeker van bent dat de ander X niet heeft gekozen. Als je 100 aangeeft, betekent dit dat je er zeker van bent dat de ander X wel heeft gekozen. Als je 50 aangeeft, betekent dit dat je het niet weet en dat je het even waarschijnlijk acht dat de ander X heeft gekozen als dat de ander Y heeft gekozen. Je kan elk getal tussen 0 en 100 aangeven in stapjes van 1 (0 en 100 zijn ook toegestaan).
Voorspelling dat de andere persoon X heeft gekozen:
Doe dit formulier alsjeblieft in de envelop als je klaar bent. Blijf alsjeblieft zitten en blijf stil.
Deel 2: Instructies
Je nummer
____________
In dit deel ga je enkele keuzes maken voor de volgende vragen. Aan het einde van het experiment zal op willekeurige wijze een van deze vragen geselecteerd
worden en je verdiensten hangen af van je keuze(s) voor die vraag. Je verdiensten in deel 2 zullen opgeteld worden bij je verdiensten van deel 1 en het bedrag dat je voor deelname krijgt. Na dit deel is er een afsluitend deel waarin je gevraagd wordt een korte vragenlijst in te vullen.
De uitbetalingen in dit deel zijn in punten, en elk punt is €0.10 waard.
Deel 2: Vraag 1
We tonen je 24 beslissingssituaties op de volgende pagina. Voor elke situatie zijn er twee opties (A en B) die geldbedragen voor jou en een andere persoon weergeven. We vragen je aan te geven welke optie je voorkeur heeft. Geef alsjeblieft voor elk van de 24 situaties aan welke optie je verkiest. Er is geen goed of fout antwoord, het hangt af van je persoonlijke preferenties.
Na afloop zullen we je willekeurig koppelen aan een persoon in de andere kamer. Dit zal een andere persoon zijn dan de persoon die je in de ontmoetingskamer hebt ontmoet in deel 1. Deze persoon maakt keuzes voor dezelfde situaties als jij doet. Als deze vraag geselecteerd wordt voor uitbetaling, zullen we de bedragen optellen die corresponderen met de 24 keuzes die jij gemaakt hebt, en we zullen hetzelfde doen voor de persoon aan wie je gekoppeld bent. Je uitbetaling hangt af van zowel je eigen keuzes als de keuzes van de andere persoon.
Optie A Optie B Jouw keuze Situatie Bedrag Zelf Bedrag
Ander
Bedrag Zelf Bedrag Ander 1 +15,00 +0,00 +14,50 -3,90 2 +14,50 -3,90 +13,00 -7,50 3 +13,00 -7,50 +10,60 -10,60 4 +10,60 -10,60 +7,50 -13,00 5 +7,50 -13,00 +3,90 -14,50 6 +3,90 -14,50 0,00 -15,00 7 0,00 -15,00 -3,90 -14,50 8 -3,90 -14,50 -7,50 -13,00 9 -7,50 -13,00 -10,60 -10,60 10 -10,60 -10,60 -13,00 -7,50 11 -13,00 -7,50 -14,50 -3,90 12 -14,50 -3,90 -15,00 0,00 13 -15,00 0,00 -14,50 +3,90 14 -14,50 +3,90 -13,00 +7,50 15 -13,00 +7,50 -10,60 +10,60 16 -10,60 +10,60 -7,50 +13,00 17 -7,50 +13,00 -3,90 +14,50 18 -3,90 +14,50 0,00 +15,00 19 0,00 +15,00 +3,90 +14,50 20 +3,90 +14,50 +7,50 +13,00 21 +7,50 +13,00 +10,60 +10,60 22 +10,60 +10,60 +13,00 +7,50 23 +13,00 +7,50 +14,50 +3,90 24 +14,50 +3,90 +15,00 0,00
Deel 2: Vraag 2
In de tabel hieronder presenteren we zes verschillende opties. Je wordt gevraagd een van die opties te kiezen.
Je verdiensten zullen afhangen van de uitkomst van een worp met een zuivere munt. Elke optie toont het bedrag in punten dat je verdient met de uitkomst Kop en het bedrag in punten dat je verdient met de uitkomst Munt. Het bedrag dat je kan verdienen is altijd hoger met Kop, behalve bij optie 6 waarin je altijd 28 punten verdient.
Als deze vraag geselecteerd wordt voor uitbetaling, zullen we een munt opgooien en je betalen aan de hand van de uitkomst van de worp en de keuze die je gemaakt hebt.
Optie Bij uitkomst Je verdiensten in punten
1 Kop 66 Munt 0 2 Kop 60 Munt 12 3 Kop 52 Munt 16 4 Kop 44 Munt 20 5 Kop 36 Munt 24 6 Kop 28 Munt 28
Deel 3: vragenlijst
Je nummer
____________
Tenslotte, neem alsjeblieft je tijd om de volgende vragen te beantwoorden. Je antwoorden zullen vertrouwelijk blijven.
Leeftijd:
Geslacht (Man/Vrouw): Aantal broers en zussen:
Geboorte volgorde (1 = jij bent de oudste, 2 = jij bent de tweede, etc.): Geboorteland:
Geboorteland(en) van grootouders:
Kende je de persoon al aan wie je gekoppeld was in Deel 1 van het experiment (de persoon die je ontmoet hebt in de ontmoetingsruimte)? Geef hieronder aan in hoeverre je de andere persoon al kende.
1: Ik kende de andere persoon niet
2: Ik heb de andere persoon eerder gezien, maar ik heb nooit met hem of haar gesproken
3: Ik heb de andere persoon eerder gezien en gesproken, maar ik beschouw hem of haar niet als een vriend
4: Ik beschouw de andere persoon als een vriend
Als je vindt dat bovenstaande mogelijkheden je relatie met de andere persoon niet goed beschrijven, beschrijf je relatie alsjeblieft hieronder in je eigen woorden:
Hoe zou je de aantrekkelijkheid van de andere persoon aan wie je gekoppeld was in Deel 1 beoordelen op een 1-7 schaal (1 is erg onaantrekkelijk; 4 is gemiddeld ten opzichte van de Nederlandse bevolking; 7 is erg aantrekkelijk)? (Als je het je niet herinnert of verkiest geen antwoord te geven, laat deze vraag dan open).
1 2 3 4 5 6 7
Erg onaantrekkelijk
Erg aantrekkelijk
Het voelt slecht als je niet aan de verwachtingen van de andere deelnemer voldoet.
1 2 3 4 5 6 7
Heb je gehoord over het “Prisoner’s Dilemma” of het “Sociale Dilemma” voordat je aan dit experiment deelnam?
Had je al iets gehoord over het experiment van vandaag? Indien ja, wat had je gehoord?
Leg alsjeblieft hieronder uit hoe je in Deel 1 je beslissing tussen optie X and Y maakte.
Heb je nog andere opmerkingen over het experiment?
RC treatment:
Jullie hadden de instructie gekregen om tijdens het communiceren geen uitspraak te doen die een leugen zou worden als je na afloop je niet zou houden aan die uitspraak (bijvoorbeeld, door het maken van een belofte).
Geef hieronder aan in hoeverre je het eens bent met elk van de twee volgende uitspraken (als je 3 of 4 aankruist, leg dan alsjeblieft uit waarom).
De andere deelnemer hield zich aan de instructie om niet zulke uitspraken te doen.
Je oordeel:
1: helemaal mee eens 2: mee eens
3: mee oneens
4: helemaal mee oneens
Als je 3 of 4 hebt aangekruist, leg dan alsjeblieft uit waarom je het hiermee oneens bent:
jij hield je aan de instructie om niet zulke uitspraken te doen.
Je oordeel:
1: helemaal mee eens 2: mee eens
3: mee oneens
4: helemaal mee oneens
Als je 3 of 4 hebt aangekruist, leg dan alsjeblieft uit waarom je het hiermee oneens bent:
FC treatment:
Deed de persoon met wie je communiceerde een uitspraak die een leugen zou worden als hij of zij zich na afloop niet zou houden aan die uitspraak (bijvoorbeeld, door het maken van een belofte)?
Geef hieronder aan in hoeverre je het eens bent met elk van de twee volgende uitspraken.
De andere deelnemer deed zulke uitspraken.
Je oordeel:
1: helemaal mee eens 2: mee eens
3: mee oneens
4: helemaal mee oneens
Jij deed zulke uitspraken.
Je oordeel:
1: helemaal mee eens 2: mee eens
3: mee oneens
4: helemaal mee oneens
Geef hieronder op een 1-7 schaal aan in hoeverre je het eens bent met de uitspraak.
Het voelt slecht als je een belofte maakt tijdens het communiceren en daarna die belofte niet nakomt.
1 2 3 4 5 6 7
Verklaring van instemming
Mogelijk zijn we geïnteresseerd in het gebruik van de video opnames voor presentatie doeleinden en vervolgexperimenten. Hiervoor willen we je toestemming vragen. Merk op dat je achtergrondinformatie altijd vertrouwelijk blijft en nooit wordt gedeeld met deelnemers aan andere experimenten ongeacht hoe je vragen 1-4 beantwoordt.
1) Ik stem in met het gebruik van de video opnames tijdens presentaties voor een wetenschappelijk publiek.
Ja Nee
2) Ik stem in met het gebruik van de video opnames in experimenten die buiten Amsterdam plaatsvinden.
Ja Nee
3) Ik stem in met het gebruik van de video opnames in experimenten die in Amsterdam plaatsvinden.
Ja Nee
4) Ik stem in met het plaatsen van de video opnames op websites die door onderzoekers in Amsterdam worden beheerd voor gebruik in een online experiment.
Ja Nee
Nummer:
Handtekening:
Dank voor je deelname! * * *
