Four-armed (restaurant) bandit

Four-armed bandit in which participants choose between four restaurants and receive feedback. The most rewarding restaurant varies over the task. You can estimate learning rates and fit computational models to this task. We constructed a narrative in which bandits were represented by images of restaurants, and the fluctuating probabilities of different outcomes was communicated via the instruction that chefs could change randomly at any time, and that even good chefs would sometimes make a bad meal. There are 200 trials, and reward probabilities shift every 25 trials. Bandits A, B, C, and D remained in a fixed position at top left, top right, bottom right, and bottom left respectively. Each of the colors, Red (R), Green (G), Yellow (Y), and Blue (B) are used to represent each bandit for one in four participants, in a manner randomized between participants.

Creative Commons Attribution (CC BY)