To test our hypothesis H1 and H2, we carry out a modified version of a dictator game. We use a manipulation experiment in which the dictator game comes under the disguise of an on-line survey. We invited over a 150 students from within Taiwan National University to join one of the 8 survey sessions available. Sessions are held in a computer room of the
Department of Information management of the National Taiwan University. Students are not limited from within the department. Students from a wide range of departments signed up for their preferred session. Funding for the survey has been generously provided by the National Science Council under budget number 79-2752-H-002-008-PAE.
In the survey each participant is matched with a virtual other participant to form a team. To augment the manipulation, we have printed the instructions for participants on one side in Dutch, and on the other side in Chinese. On the Dutch side of the instruction page, we have put a little sticker with a login ID and password, supposedly put there by the other participant.
Hence, the participant is made to believe he is the second and final participant of his/her group. At the end of the survey, each team gets rewarded a fixed amount for the team and the ratio for each team member has to be decided by the participant, who by deciding, becomes and instant dictator. The dictator and the responder are both guaranteed anonymity with respect to each other and to the experimenter. This is important because we want participants to feel free to choose any amount they see as appropriate and not merely show socially acceptable behavior. Although the other agent’s identity is usually hidden in dictator games, factors like location may provide some basic clues about the other person (e.g. that he is a student too), or at least that he is a person.
There are 9 possible routes, scenarios, through the survey which are carefully controlled.
The difference between scenarios can be found in Table 1, and are discussed in chapter 4.1.
Because in dictator games the responder has no decision power, we have made his role is a virtual one. (The dictator, of course, is not aware of that)
1 2 3 4 5 6 7 8 9
Virtual team member's performance N.A.
Social recognition (MSN help account) No No Yes No Yes No Yes No Yes Online search questions None All All All All All All All All Distribution Question Yes Yes Yes Yes Yes Yes Yes Yes Yes Equalitarian outcome ( NT$ ) 125 125 125 155 155 125 125 95 95
Scenario
No info. 38 % 50 % 62 %
4.1 The survey
The topic of the survey, “information search, a cultural component” is in itself not
important. It only serves the purpose of being a task for which the team gets rewarded. For the manipulation to be successful, we have tried to design a survey that makes the fact that
participants are teamed up, seem logical to them. The only question important for the experiment is the final one, in which the dictator is asked to decide a ratio for the reward.
Table 1 Scenarios
4.1.1 Scenario 1, The control group
Participants that are selected for scenario one, are not required to fill out the survey. Their ID number brings them directly to the ratio question at the end of the survey. In effect, scenario 1 is a simple dictator game.
This is the only scenario that does not have its own session. We felt it would be more natural if each of the other sessions had two or three participants with this very short and specific task. A message inside the survey, but after the decision question, asks these participants to stay at least until 20 minutes into the session, in order not to raise questions from the other participants.
4.1.2 Scenario 2, Shared task, no information
The ID number of a participant in scenario 2 brings him virtually to somewhere halfway the survey. The participant has no information about the absolute position or progress throughout the survey, but is manipulated to believe that his partner has completed part of a collective question list. Participants can not go back to check the questions or answers that their team member has completed.
4.1.3 Scenario 3, Social interaction
The ID number of a participant in group 3 brings them to the same position in the survey as the participant in scenario 2. He also does not know how much his (again, virtual) partner has done. The difference with scenario 2 is that participants get the opportunity to ask questions or communicate on MSN with a research assistant. During the session for scenario 3 (and 5, 7 and 9), Each PC has a preconfigured MSN account. Sometime at the beginning of the session an assistant sends a scripted MSN message. In the message the assistant thanks the participant for helping the experiment by doing this survey, and offers help in return in case it is needed.
The MSN accounts deliver social recognition outside the context of the survey, which makes it more likely to the dictator that he is socially recognized and leaves ambiguity about social recognition of his teammate. A negative effect may be that having an MSN account open during the survey raises the suspicion that the dictator is observed and no longer anonymous.
We need to test for this effect.
4.1.4 Scenario 4 to 9, different levels of relative performance
To test if participants use information about the performance of team members, starting from scenario 4, we offer a progress indicator on top of the survey page that indicates the percentage of questions that the virtual partner has completed (38 – 50 – 62 percent respectively, as in Table 1). Progress indicators are common in online surveys. The
participant is not specifically told to use this information in rewarding his/her team member.
From a pilot survey we know we need to be specific about the different colors in the progress bar.
Figure 3 Previous version of progress bar, participants failed to notice.
Figure 4 Final version of progress bar.
4.2 The decision question.
The final page of the survey tells participants in all scenarios that their team has earned 250 NT$. We tell them they have to decide how much of that amount they will take for
themselves and how much they leave to the teammate.
Bardsley (2008) indicates that the range of the choice set that is available to subjects in a game carries some clues toward acceptable test-behavior.
By starting the slider at nothing for participants and 250NT$ for the team member, rather than at 125 NTD for each, we achieve two things.
- We make sure participants do make a conscious choice.
- We avoid providing a reasonable choice set.
Figure 5 Slider before making a choice
Figure 6 Slider after making a choice