This is an anonymized version of the pre-registration. It was created by the author(s) to use during peer-review.

A non-anonymized version (containing author names) should be made available by the authors when the work it supports is made public.

No, no data have been collected for this study yet.

Q1: I ask whether subjects behave as by the standard models of dynamic choice with learning. My theoretical results link these models with linear indifference curves. Which I collect and measure whether subjects display parallel curves.

Q2: I ask whether subjects use a proportional heuristics and whether they display belief uncertainty. These are all defined in the theoretical section of the paper. Proportional heuristic is defined as choosing the sample that a higher proportion of good draws - irrespective of sample size. Belief uncertainty is measured using a novel design which allows subject to learn the correct choice at a cost.

Q3: I ask how much the shape of the indifference curves can be explained by the proportion heuristic and belief uncertainty. As well as whether belief uncertainty is correlated with proportion heuristic use.

For each subject, I collect indifference curves as well as a measure of belief uncertainty, each choice (or pairs depending on question) gives one indifference curve and one measure. My three key variables are:

1) Angle of indifference curves and their standard deviation.

2) For each choice, whether it is consistent with heuristic use, and the average number of times this happens. This is defined as consistent if the choices lie within a 5% interval of each other.

3) The number of times the subject displays belief uncertainty as elicited via my measure.

There are 3 treatments - between subjects, so each subject only sees 1. They are about information being of symmetric, asymmetric, ambiguous accuracy.

Similarly, all subject makes 10 sample choices (2 blocks) and 6 belief elicitation choices (1 block). The order between these three blocks are randomized while within the blocks the order is the same for all subjects.

Correlation between the three key variables.

T-test of angle equality of indifference curves.

Linear regression of the heuristic use and belief uncertainty on standard deviation on indifference curves.

I will conduct two regressions, both with the standard deviation of indifference curves as the dependent variable. The independent variable of interest are heuristic use and belief uncertainty measures for the two regressions. Both will have treatment controls and the interaction terms.

I will present two sets of results. With the full sample and a sub-sample where I drop subjects who violate a monotonicity condition. The condition is that they must understand that an extra good signal in the sample over a bad one implies that the object is more likely to be good. This condition is taught to them during the comprehension tests.

No need to justify decision, but be precise about

I will collect 400 subjects who passes a comprehension test from Prolific. They will be screened as follows. They must have completed at least 100 studies and have an approval rate of 97-99% as well as be residing in the United States.

(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)

Analysis of differences between treatments.

Analysis of order effects of the different treatments.