I Didn’t Know Either - the Dynamics of Strategic Ignorance

This experiment relies on information provision about the behavior of others.

2024-12-02

2025-02-28

Using the between subjects design, our main outcome of interest is the ignorance rate and elicited belief regarding the ignorance rates across the different treatment arms. Ignorance rate is calculated as a proportion of subjects within a treatment arm who did not reveal information regarding the recipient's payoffs. Elicited beliefs on ignorance will be quantified between 0 to 100 percent.

Subjects will be asked to evaluate the social appropriateness of revealing versus not revealing on a 4-point Likert scale, describing the behavior as being socially inappropriate to socially appropriate.

We may classify social appropriateness by grouping the ratings on the 4-point Likert scale into two groupings: inappropriate and appropriate and derive a ratio for the degree of overall social appropriateness.

In addition, subjects will evaluate themselves on The Conformity Scale (TCS) followed by a questionnaire on general demographics.

We will develop a measure of social conformity using the The Conformity Scale by averaging subject's responses across the 11 questions.

The experiment is divided into three distinct sections with 6 treatment arms in a 2 by 3 experimental design, varying conditions of whether or not subjects know that they will be playing the moral-wiggle room game and varying conditions of the low versus high ignorance norms and a control condition without any norm information.

The first section elicits subjects' beliefs about the behavior of other subjects. The second section asks subjects to participate in the moral-wiggle room game. The third section asks subjects to rate the social appropriateness of revealing and not revealing the payoff tables.

The experiment is divided into three distinct sections with 6 treatment arms in a 2 by 3 experimental design. In the first section, subjects are briefed about the moral wiggle room game (Dana et al. 2007). Upon successfully answering a series of comprehension tests, subjects are then first asked to elicit their beliefs on the proportion of subjects in a similar, but separate experiment who had chosen not to reveal. Then subjects are asked to elicit their beliefs on what proportion of other subjects believe that it is socially appropriate to reveal.

The first set of experimental manipulation involves the framing of the belief elicitation. In three treatment arms, we tell subjects that they will actively play the moral wiggle room game, while in the remaining three treatments, we surprise the subjects by asking them to play the game after eliciting their beliefs about how other subjects had behaved in the game. Subjects are incentivized for accuracy, earning a bonus if their answer is more accurate than the majority of the other predictions.

In the second section, subjects are then tasked to play the role of the dictator in a modified moral wiggle room game. Before making their decision, in the low and high ignorance norm treatments, dictators are informed about the reveal rate that they had been asked to predict. I exogenously vary the real reveal rate by re-using the results from a separate study (currently a working paper). I re-use the ignorance rate from another experiment in which the choice architecture drastically altered the ignorance rate. The norm treatments consist of low ignorance (30%) and high ignorance rates (70%).

Lastly, in the third section, subjects are then asked to evaluate the social appropriateness for revealing and not revealing followed by a questionnaire battery consisting of The Conformity Scale and a general demographics questionnaire.

Subjects in the known (unknown) treatments will know (not know) if they will be participating as a dictator in the moral wiggle room game. Across both known and unknown conditions, dictators will make a decision under: no norm, low ignorance norm, and high ignorance norm information, with no information about the reveal rate, a 70% reveal rate, and a 30% reveal rate. The 6 treatments are thus:

Known - No Info
Unknown - No Info
Known - 30% Ignorance
Unknown - 30% Ignorance
Known - 70% Ignorance
Unknown - 70% Ignorance

Low and high ignorance rates of 30% and 70% are selected as they provide a clear majoritarian norm. Low and high ignorance treatments for both Known and Unknown conditions serve to examine how self-serving biases may be collapsed once information is revealed. However, depending on the actual beliefs of dictators, if such beliefs drastically understate ignorance rates, it may be necessary to run a 10% ignorance environment using results from Lind, Nyborg, and Pauls, 2019. A 10% ignorance environment will be considered if less than 20% of subjects report a belief that ignorance rates are higher than 30%. This is necessary for testing how subjects would behave if their belief about ignorance rates are higher than the actual rate as both 30% and 70% ignorance environments would be considered relatively high ignorance rates relative to beliefs.

Computer randomization and balanced assignment for treatment arms

Individual

No

N/A

At least 600 online participants; higher N depends on budget constraints. 800 if including the 10% ignorance rate treatment arms

At least 100 subjects per treatment arm; higher N depends on budget constraints

In line with (Dana et al. 2007), we look to discern differences of 0.2 or 80% power. For a difference of proportions for a binomial distribution, to obtain a power of 80%, a sample of 97 subjects per treatment is needed. To simplify, I aim for 100 subjects in each group in each comparison.

I_Didn_t_Know_Either___the_Dynamics_of_Strategic_Ignorance.pdf