Codes and data for paper "Environment Design for Biased Decision Makers".
In simulation
folder, Experiment.py
is main function for simulation test. Use following command to run experiments. MDPenvironment.py
, Agent.py
, Principal.py
, are for environments, agent models and principal models. Simulation results is stored in simulation/results
folder.
python Experiment.py baseline # baseline test
python Experiment.py reward_modify # reward function modification test
python Experiment.py action_nudge # action nudge test
We recruit 300 individual workers from Amazon Mechanical Turk. Experiment data is stored in human-subject-experiment/mturk-experiment-data.csv
:
uid
represents anonymous workers, from 0 to 299treatment
treatment of the worker, from 0-2 (0-baseline, 1-reward function modification, 2-action nudge)test_idx
test order of games, 6 integers from 0 to 5data
actions (worker moves) in each game, first is in test game (not included intest_idx
)scores
scores in each game, first is in test game (not included intest_idx
)
Experiment setting is stored in human-subject-experiment/experiment-setting
in forms of .json
.