We study the effect of cues---normatively irrelevant messages that may nonetheless bring certain possibilities to min---on beliefs in a simple controlled experiment conducted on Prolific.
Experimental design
Treatment variation consists in randomizing whether and what types of cues---example outcomes from the data-generating process---that participants see while they form their beliefs.
Design details
Randomization method
Done within qualtrics using javascript.
Randomization unit
Individual-level.
Clusters
We plan to collect data from 3190 participants.
Observations
3190 participants.
Arms
200 each in the following treatments:
1. Control
2. One baseline positive cue
3. One baseline negative cue
4. One extreme positive cue
5. One extreme negative cue
6. One close positive cue
7. One close negative cue
8. Five baseline positive cues
9. Five baseline negative cues
10. Ten baseline positive cues
11. Ten baseline negative cues
12. One impossible positive cue
13. One impossible negative cue
In addition, we will recruit 10 participants to see one cue with a final score equal to each of the 59 possible values not covered by any of the other treatments.
This leads to 13*200 + 10*59 = 3,190 participants.
Primary outcomes
Our primary variable of interest is participants' beliefs about how often the process ends with a positive score. We are also collecting data on how often they believe it ends with an even/odd score. In addition, we collect similarity data to generate predictions from our model.
Secondary outcomes
We also elicit from participants one outcome that they felt came to mind for them while they were forming their beliefs. Finally we collect a measure of cognitive uncertainty for each belief question, though our model does not generate predictions about cognitive uncertainty.
Cited by
No citation data available for this registration yet.