search

SpTB / mob_project

1 stars 0 forks source link

Task design (separating decisions from actions) #2

Open SpTB opened 3 years ago

SpTB commented 3 years ago

The new task will include drag-and-dropping items, to make it more engaging, I am still working on the details, but there are two general ways of implementing this:

  1. Similar to the previous design, people are presented with a screen and the two tasks are on two sides of the screen. They choose by performing one or the other.
  2. People are first presented with a choice: risky or safe task. This choice then redirects them to the screen with the task they chosen.

I personally like option 2 more. By separating decision from task we add ecological validity. Plus, the RTs now are choice-specific, so there would be less ambiguity in modelling them.

Obvious downside is time - the task will be longer, since participants need to perfrorm 2 actions (choice+task) in each timepoint.

Let me know your thoughts

pantelispa commented 3 years ago

We should definitely go with 2, and people should be presented with the risky task first since this is also the optimal strategy. The drag and drop task could work, but it could also become super repetitive after a while (and even more boring than simply clicking on safe).

We certainly need to pilot this and also think about what economists call incentive compatibility a bit harder. The more I think about it, a direct payout for the remaining rounds when people give up might not be a bad idea (especially when the bonus to safe rewards ratio is high).

SpTB commented 3 years ago

Well, doing an auto safe payout would certainly be the easiest from the coding perspective... My issue is 2-fold: 1) incentive confound
2) ecological validity. I have a feeling that the current bandit-task resemblance is to our detriment. Pushing a button and getting an instant (variable) gratification is a far cry away from completing a skill-based goal. As such, it is at risk of being affected by gambling fallacies, such as assuming that outcomes are dependent (.e.g expecting a high payoff following a few low ones or vice versa).

That being said, don't want to make the final call on this.

pantelispa commented 3 years ago

Here we concluded that when people terminate the risky task they will directly earn the money from the safe task. This will allow them to save time and move to another experiment earlier.

To account for that we can gauge the average pay of people in prolific (something like their hourly rate, and the amount of time saved when switching). We can redo this analysis using by also adding the opportunity cost to the earnings.