Spontaneous cooperation for prosocials, but not for proselfs: Social value orientation moderates spontaneous cooperation behavior

Cooperation is essential for the success of societies and there is an ongoing debate whether individuals have therefore developed a general spontaneous tendency to cooperate or not. Findings that cooperative behavior is related to shorter decision times provide support for the spontaneous cooperation effect, although contrary results have also been reported. We show that cooperative behavior is better described as person × situation interaction, in that there is a spontaneous cooperation effect for prosocial but not for proself persons. In three studies, one involving population representative samples from the US and Germany, we found that cooperation in a public good game is dependent on an interaction between individuals’ social value orientation and decision time. Increasing deliberation about the dilemma situation does not affect persons that are selfish to begin with, but it is related to decreasing cooperation for prosocial persons that gain positive utility from outcomes of others and score high on the related general personality trait honesty/humility. Our results demonstrate that the spontaneous cooperation hypothesis has to be qualified in that it is limited to persons with a specific personality and social values. Furthermore, they allow reconciling conflicting previous findings by identifying an important moderator for the effect.


Sample characteristics
The first study was conducted on Amazon Mechanical Turk (MTurk) to stick as closely as possible to the methods used by Rand, Greene, and Nowak (2012) 1 . The sample consists of 134 MTurk workers (51 female) with a mean age of 30.5 years (range 18-62 years), half of them from the United States and the remaining from other western countries and India. The second study collected data of 105 participants (70 female participants), mainly German students with mean age of 26.97 years (range 19-62 years). The representative Panel sample for US and German population consists of 504 participants in total (258 female) and a mean age of 46.13 years (range 18-77 years). The subsample of US citizens consisted of 249 participants, the German subsample accordingly of 255 participants with similar age and gender distributions. We excluded 4 participants in Study 1, 4 participants in Study 2 and 11 participants in Study 3 due to too long reaction times (more than + 3 SD from the mean), lack of game comprehension (contributions > 100%), or as drop-out during the study. Including them did not change the results in the overall analysis and only slightly but unsystematically in the separate analyses (i.e. for the better in Study 1 and for worse in Study 3). Furthermore, participants who answered control questions about the game structure incorrectly were screened out and not allowed to take part in the studies.

Method and experimental instructions
We used the instructions of Rand et al. 1 in Study 1 and for the US subsample in Study 3. The same instruction was translated into German or the Study 2 and the German subsample in Study 3. In the first two studies we additionally manipulated the framing of the game as either 'Wall Street Game' or 'Community Game' 2 . This manipulation does not influence the presented results, which persist when including the framing as a covariate.
First participants were introduced to the setting before being familiarized with the instruction of the Public Good Game (PGG). The decision how much to contribute was displayed on a new page in order to disentangle reading time from decision time. Control questions assured the understanding of the game structure and were asked afterwards, in order not to prime any reflection beforehand. A different order was used in Study 2 (lab study), in which participants answered the questions at the very beginning of the experiment. This was done in order to avoid a costly dropout of participants. In the online studies, participants with incorrect understanding of the game structure were automatically screened out by a filter option.
We measured beliefs after the control questions, in order to avoid biased contributions by deliberation taken beforehand. The next screen asks for the cooperativeness of daily life interaction partners, identified as a moderator of spontaneous cooperation by . Cooperativeness is operationalized via trust in order to assure a uniform and more familiar understanding of cooperativeness 3 . Afterwards, we extended the replication of  by including the Social Value Orientation (SVO) Slider Measure of Murphy, Ackermann, and Handgraaf 4 . Participants were clearly instructed that either the PGG or the SVO Slider would randomly be chosen in order to calculate the bonus payment. Both tasks offered roughly the same expected value.
In the panel study, the HEXACO-6 factor model of personality was included afterwards, before an exploratory instructional manipulation check 5 that was mainly included to make inferences about the framing manipulation in the first to studies as either 'Wall Street Game' or 'Community Game'. We don't pursue this manipulation in this paper any further, but (as mentioned above) our results persist when including the framing as a covariate. Finally, previous experience with the experimental setting of social dilemma games and demographic variables was enquired.

Screen 1 (Introduction)
Thank you for taking part in this study. You will receive a basic incentive. In addition, you have the opportunity to receive additionally up to $3.75 for one of the two major tasks this study consists of. Both offer you about the same amount of bonus payment. It is important that you answer both tasks; one will be randomly selected to calculate your payment.

Screen 2 (PGG instructions)
You have been randomly assigned to interact with 3 other people. All of you receive this same set of instructions. You cannot participate in this study more than once. Each person in your group is given 150 cents for this interaction (in addition to the basic incentive you already received for participating). You decide how much of your 150 cents to keep for yourself, and how much (if any) to contribute to the group's common project (in increments of 2 units: 0, 2, 4, 6 cents etc). All money to the common project is doubled, and then split evenly among the 4 group members. Thus, for every 2 cents contributed to the common project, each group member receives 1 cent. If everyone contributes all of their 150 cents, everyone's money will double: each of you will earn 300 cents.
But if everyone contributes their 150 cents while you keep your 150 cents, you will earn 375 cents, while the others will earn only 225 cents. That is because for every 2 cents you contribute, you get only 1 cent back. Thus, you personally lose money on contributing. The other people are REAL and will really make a decision -there is no deception in this study. Once you and the other people have chosen how much to contribute, the interaction is over. Neither you nor the other people receive any bonus other than what comes out of this interaction (in case this task is chosen to calculate your bonus payment).

Screen 3 (Contribution)
Please indicate the amount of money (in cents) you wish to contribute.

Screen 4 (Control question I)
What level of contribution (in cents) earns the highest payoff for the group as a whole?

Screen 5 (Control question II)
What level of contribution (in cents) earns the highest payoff for you personally?

Screen 6 (Beliefs)
What do you think will be the average contribution (in cents) of the others to the common project?

Screen 7 (Cooperativeness of interaction partners in daily life)
To what extent do you feel you can trust other people that you interact with in your daily life?

Screen 8 (SVO instructions)
This is the second task; it is not related to the first one. Here, you have been randomly paired with another person, one who has not been in the group before. We will refer to him or her as the other. This other person is someone you do not know and will remain mutually anonymous. All of your choices are completely confidential. You will be making a series of decisions about allocating resources between you and this other person. One choice is randomly selected and payed as bonus fee for you and the other. For each of the following questions, please indicate the distribution of money you prefer most by marking the respective position along the midline. You can only make one mark for each question. Your decisions will yield money for both yourself and the other person. The monetary unit of the displayed numbers is cent. In the example below, a person has chosen to distribute money so that he/she receives 50 cents, while the anonymous other person receives 40 cents.
There are no right or wrong answers; this is all about personal preferences. As you can see, your choices will influence both the amount of money you receive as well as the amount of money the other receives (in case this task is chosen to calculate your bonus payment).

Screen 11 (Instructional manipulation check)
Modern theories about decision making processes have shown that decisions are influenced by many different aspects. Individual preferences, prior knowledge and context variables influence the decision process. To improve our research on decision making, we are interested in getting to know some aspects about you, the decision maker. Especially, we are interested if you take enough time to read the instructions carefully. If you don't, some of our manipulations based on differences in the instructions are ineffective. This is why we ask you to ignore the question at the bottom of this page and to click on all response options instead in order to show us that you read the instructions. Afterwards, you can proceed to the next page. Thank you very much! Which of these countries have you already been to? Please tick in as appropriate.

Screen 12 (Previous experience)
To what extent have you participated in studies like this one before? (i.e. studies in which you choose how much to keep for yourself versus contributing to benefit others)

Results
Table S1 presents descriptive statistics, summarizing demographics and the main independent and dependent variables. Note that we observed considerable differences concerning the level of decision time between studies, with participants in Study 2 taking much longer than in the other studies (S1). This could be due to the fact that participants in the lab (Study 2) might generally take more time to decide in comparison to online samples (Study 1 and 3). The payment in the lab is usually higher than that of the online providers so that people might be used to work more carefully. Furthermore, it might be more tempting to increase hourly wages by shortening the time to work on a task in an anonymous online setting as compared to a lab involving some aspects of group norms.
Comparing the results of the Tobit regression presented in Table 1 to an OLS regression, the same general pattern of results holds true (Table S2). Note. t-values are presented in parentheses; predictors are centered. Overall analysis also includes two study dummies which are not reported. Figure S1 shows that the overall spontaneous cooperation effect replicating Rand et al., 2012. Figure S1: Replication of the spontaneous cooperation effect over all three studies (N=743). Regression line represents the effect of reaction times on contributions from an OLS regression. Whiskers denote the 95% confidence interval; the diameter of the grey dots is proportional to the sample size.
When splitting the sample in proselfs (SVO angle < 22.45°) versus prosocials (SVO angle > 22.45°), the spontaneous cooperation effect can only be found for prosocials (Table S3). Further analyses revealed that the moderating effect of SVO angle with regard to the spontaneous cooperation effect goes beyond the effects of previously observed moderators (Table S4). The coefficient of the SVO * time interaction remains stable also when including interactions of time with cooperativeness of daily life interaction partners (trust) and experience in related studies. Persons with a high personality scores in Honesty-Humility (HH; Cronbachs alpha = .78) show a similar effect of decision time on cooperation as prosocials do, which dovetails with recent findings from mouse-tracking studies. 7 Table S5. Tobit regression of Honesty-Humility and reaction times on contributions (Study 3)