In this repository you will find:

• Python code to load levels, visualize levels, create levels, and simulate actions within levels (environment/)
• Level jsons used in the PNAS paper (in environment/Trials)
• Human and model data for these level jsons (in data/)

If you are interested in playing around with the code, or building machine agents for this environment, check out the instructions here.

After installing needed dependencies, a good place to start to understand how to use our API is make_basic_trial.py which shows you how to create a level, save it, load it, simulate an action inside it, and visualize the result of that action.

You can also play the game directly in your browser here: http://scripts.mit.edu/~k2smith/toolgames/