Fig. 1: Choice engineering assignment.
From: Behavior engineering using quantitative reinforcement learning models
a In the experimental task, the participant repeatedly chooses between two alternatives (1 and 2). Following each choice, the participant is rewarded or not rewarded ($ or X, respectively) in accordance with a predefined binary reward schedule. Unbeknownst to the participant, each of the alternatives was associated with exactly 25 rewards (red-filled circles). b With the objective of maximizing bias in favor of alternative 1, a choice engineer can use a model of the participant’s learning strategy in order to construct an effective reward schedule. The schedule depicted here is the competition winner, a reward schedule optimized for the behavioral model CATIE. c Alternatively, a choice architect may use the principle of primacy and favor allocating as many rewards as possible to alternative 1 at the beginning of the task. Similarly, primacy may dictate deferring all rewards allocated to Alternative 2 to the end of the task.
