| Popis: |
Because we had to change the model due to overfitting, and because the effect sizes were small, we decided to pre-register a replication of study 2 (https://osf.io/eh7rt) with small modifications in the experimental design. We focus on one theme only, because the variance in the response due to the different thematic sets was low: it means that what is found in one domain is reproducible in the two other domains. We chose the domain for which the proportion of variance explained in Study 1 is the highest (i.e., the domain in which the difficulty of the questions best explains the actual performance of the people who got the questions right). This is the ‘superheroes’ set. To moderate the problem of overfitting, we will ask each participant to evaluate a greater number of questions and we will split the overall experiment in two: (1) one experiment will be about evaluating virtual individuals who are said to assert a widespread information or respond to an easy answer (according to the condition) and (2) the other experiment is about evaluating virtual individuals who are said to assert a rare information or respond to a difficult answer (according to the condition). The reason for conducting separate experiments is that the hypothesized mechanism for discounting the knowledge of people who make assertions, by contrast with the knowledge of people who answer questions, is different. When someone asserts something that is widely known, they might reveal that they are not aware the information is indeed widely known, and so its assertion isn’t terribly relevant. When someone asserts something that only few people know, they might be suspected of using that to brag, without necessarily knowing much more about the domain. |
