'MoSa - Info-sampling, Bayesian-learning, feedback & externalities'
(AsPredicted #104844)
Created: 08/17/2022 04:47 AM (PT)
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.
1) Have any data been collected for this study already?No, no data have been collected for this study yet.
2) What's the main question being asked or hypothesis being tested in this study? How is Bayesian-learning and info-sampling for a binary decision affected by feedback and externalities in the outcome of the decision?
Specifically:
- Do externalities create an asymmetric sampling behaviour?
- How do different types of externalities affect decision behaviour?
- How does the decision threshold change due to externalities and feedback?
- How are prior and posterior beliefs impacted by feedback?
- And how does the feedback influence the success in the actual sampling and decision task?
3) Describe the key dependent variable(s) specifying how they will be measured. We simply measure the following variables
- Amount of sampled information per round and amount of sampled information per option in every round,
- Which option the subjects chose,
- The prior and posterior beliefs the subjects state.
Furthermore, we calculate with Bayesian statistics the objective certainty level after every sampled information and when the participants make their decision.
4) How many and which conditions will participants be assigned to? In total there are 9 conditions in a 3x3 design along the dimensions "Feedback" and "Externality":
In the feedback dimension there are the following 3 conditions:
- No feedback: at the end of a round subjects receive no feedback.
- Correct/Incorrect feedback: They receive information whether the option they chose was correct or incorrect.
- Posterior Feedback: Subjects receive feedback on the objective probability given the information they sampled and can compare it to their stated posterior belief.
In the externality dimension there are the following three conditions:
- No externality in any of the two option
- Randomly assigned "positive externality" to one of the two options: if subjects choose this option and it is correct, 1 point goes to an organisation subjects stated they are most likely to contribute money to.
- Randomly assigned "negative externality" to one of the two options: if subjects choose this option and it is wrong, 1 point goes to an organisation subjects stated they are least likely to contribute money to.
The 9 conditions are then all combinations on the two dimensions.
5) Specify exactly which analyses you will conduct to examine the main question/hypothesis. Through regression analysis we will analyse:
The impact of no / positive / negative externality and no / correctness / precise-posterior feedback on:
- Amount of sampled information per round
- The balance of sampling between the two options
- The objective decision threshold at which subjects make their decision
- The posterior belief they state
- The relation between objective decision threshold and posterior belief (over / under confidence)
6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations. We will exclude the posterior belief data point if it is <50%.
7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined. 50 per condition.
8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?) Additionally, we will do agent-based simulations of this experiment where we vary parameters such as prior beliefs and use learning models. We will compare the performance of the agents using the simulated learning models with the data of the human participants. We do this to infer the performance of different learning models and to show what kind of strategies human agents actions where comparable to.
We will compare them to the human subject data and will always clearly state which data are human subject data and which are agent-based simulated data.
