The Task

Human subjects will be tested in a 100-trial, two-alternative forced choice experiment with binary rewards, denoted here as a session.

• A live version of the experiment can be found here.
• The full code of the experiment can be found here.
  • Note: If you use the code to collect data for the competition, we kindly request that you submit this data together with your schedule. This data will not be used for comparing the submitted schedules but we will make it publicly available (with credit to those collected it), together with the data collected in the process of the competition. One goal of the competition is to generate a publically-available large dataset of human behavior in this task.

• A session consists of 100 trials, in which the subject chooses between the two alternatives.
• If the subject chooses a rewarded alternative then she receives a monetary reward in that trial.
• A reward may be assigned to one, two or none of the alternatives in each of the trials
• Feedback is partial - the subject is informed about the outcome of her choice but not about the outcome of the forgone choice.

Your challenge is to construct a reward schedule that maximally biases the choices of the subjects, complying with the global constraint:

• Assign a reward to exactly one-quarter of the trials (25 trials) of each alternative.

We offer two participation tracks:

• In the “static” track, choice designers are invited to propose a reward schedules in the form of a sequence of rewards (see for example here).
• In the “dynamic” track, choice designers are challenged to submit a code that allocates, in every trial t, the reward/s of the next trial (r_t^bias ,r_t^¬bias ∈{0, 1}), based on the history of choices (a₁, a₂,...a_t-1; a_i∈{bias, ¬bias}) and rewards ((r₁^bias ,r₁^¬bias ), (r₂^bias ,r₂^¬bias ),..., (r_t-1^bias ,r_t-1^¬bias ); r_i^bias ,r_i^¬bias ∈{0, 1}), see for example here.