Strategic Curiosity

It is an online experiment which is conducted on Amazon Mechanical Turk.
The complete task takes about 7 minutes, treatments are built on the outcome of a die-rolling task and how many times it can be rolled.

2019-07-19

2019-08-02

Information seeking behavior and dishonesty

How the reported values change based on the outcome of the first roll and the number of rolls. In addition, when the outcome changes and generates unrelated information dishonesty might change.

We employ a digital version of the die-under-the-cup task (Shalvi et al. ,2011) where subjects roll a virtual die and report the outcome for monetary rewards. We choose the die-rolling setting as it allows us to experimentally vary the demand for justifications. Before rolling, all participants will be shown the payoff structure for their report, i.e. higher reported numbers will result in higher payoffs. The digital die will be programmed to be fair: it will randomly display the numbers of 1 through 6. The outcomes of the die is recorded, which means that this is an observed game similar to games used in recent studies on dishonesty (Gneezy et al. 2018; Pittarello et al., 2015). This design enables us to investigate whether the outcome of the first roll affects the probability that participants would want to roll more than once. Thus, we can test directly whether the distance between the observed outcome and the wealth-maximizing outcome predicts information acquisition.

Participants in our experimental set-up are allocated to four different conditions. The four conditions are as follows:
C: Participants are only allowed to roll the digital die once, before they report their result on a subsequent page (roll-once condition).
T1: Participants are asked to first roll the die once, and then to roll it two more times. They will then be asked to report their outcome from the first roll on a subsequent page (three-rolls condition).
T2: Participants are asked to roll the die once, and will then be able to roll it for as many times they would like. They then continue to a subsequent page and are asked to report the outcome from the first roll (roll freely - justification potential condition).
T3: Participants are asked to roll the die once, and will then be able to roll a different type of die as many times as they would like. This different die will only display unordered and non-numeric symbols. After rolling as many times as they would like, participants continue to a subsequent page, and are asked to report the outcome from the first roll (roll-freely - no justification potential condition)

Participants self-report their roll, which provides them with an opportunity to cheat (misreport), and participants who receive a lower number than 6 will have a monetary incentive to cheat. We will pay participants according to their report, as we inform them in the instructions. All participants will have equal chances to receive a high number on the digital die on their first roll (1/6), and will therefore have equal chances to earn the maximum bonus without cheating. In practice, all participants can claim the maximum bonus by simply reporting the number 6. After reporting their outcomes, participants will be asked questions about the experiment and other demographic questions (Experiment materials can be found in the submitted pre-analysis plan).

Participants will be randomized by the python based program that we use for the design of the experiment: Otree. Each condition will have the same number of participants.

Individual level randomization.

No

4 conditions

1600 participants

400 participants in each of the 4 conditions: control, roll 3 times, roll freely, roll freely with figures.

We aim for 0.3 effect size with min. 0.8 power.

Pre-Analysis Plan-Detailed