load_model_based_model_free#

load_model_based_model_free()[source]#

Load the two-step task dataset from BrainExplorer.

Returns:
pd.DataFrame

A pandas DataFrame with one row per trial and participant.

See also

cpm.applications.reinforcement_learning.HybridMBMF

hybrid model-based / model-free reinforcement learning model of the two-step task.

load_two_step_data

the two-step task data of Smid et al. (2022).

Notes

The data are from 93 participants who played the two-step task in the BrainExplorer app (https://brainexplorer.net/), the game called “Goblin Heist”, in up to two sessions of 80 trials. On each trial, participants choose one of two spaceships, which takes them to one of two planets, where they receive a reward. Like the task of Kool et al. (2016), on which it is based, it has two starting states with two spaceships each, deterministic transitions, and a stake of 1 or 5 on each trial. The task uses the original coding, so the columns need converting before they can be passed to HybridMBMF.

The dataset includes the following columns:

  • userID: the participant number.

  • trial: the trial number.

  • block: the block number.

  • stimuli: the two spaceships shown on the trial, numbered 1 to 4 across both starting states, in the order they were displayed, such as "[1, 2]".

  • s: the starting state and the planet reached, such as "[1, 2]".

  • rews: the reward values of the two planets recorded by the game on the trial, from 0 to 100, such as "[40, 50]". They are not always the reward received; that is points.

  • choice: the chosen spaceship, numbered 1 to 4.

  • stake: the stake of the trial (1 or 5).

  • points: the reward received, from 0 to 9.

  • points_real: the points shown to the participant, points times 10 times stake.

  • planet: the planet reached (1 or 2).

  • theor_best_choice: the spaceship that leads to the planet with the higher reward.

  • RT: the response time of the choice, in ms.

References

Kool, W., Cushman, F. A., & Gershman, S. J. (2016). When does model-based control pay off? PLoS Computational Biology, 12(8), e1005090.

Examples

>>> data = load_model_based_model_free()
>>> print(data.head())