decisions_from_experience

replication and extensions of Hertwig et al. decisions-by-experience experiments

Potential Experiments

1. straight replication of Hertwig et al., without distribution information (mostly for comparison to other conditions, and to show that things aren't wildly different on MTurk).

2. Replication with knowledge about what the potential rewards are on each option to create well-specified task. Within this, we can have multiple sample-cost conditions:

   • no explicit sample cost
   • low explicit sample costs
   • high explicit sample costs
   • non-monetary sample costs?

3. (Speculative) If people tend to over(under)-sample, can we do some kind of training to get them to sample less(more)? Taking inspiration from Anna's project, this could involve first doing a game with certain problems, certain sample costs, etc...

   • the pipe dream would be if explicitly charging people for samples the right amount or only at the beginning actually caused subjects to make more money than if they weren't charged for samples.