Space Treasure RL Task

Author: Paul A. Bloom

Date: Jan 2 2019

A gamified 2-stage reinforcement learning task built in Python

Task design based on Daw et al., 2011 and Decker et al., 2016. While virtually all task contingencies are identical to previous versions of the task, we make several modifications here to run this task with 6-13 year-old children:

• The transition matrix is explicitly spatial, and all on one screen (i.e. participants can always see both portals and planets)
• Trials are faster (usually under 3s), such that most participants complete the task (200 trials) in roughly 10 minutes
• We have added animations, sounds, and music in attempts to 'gamify' the task and engage children for the task duration

To run the game:

python main.py from command line

Python Dependencies:

pygame numpy pandas random datetime os textwrap