Data and analysis code for: Cohen, A.O.*, Nussenbaum, K*, Dorfman, H.M., Gershman, S.J., & Hartley, C.A. (2020). The rational use of causal inference to guide reinforcement learning strengthens with age. npj Science of Learning.
Anonymized trial-wise data for all participants are provided in anonymized_mining_data.csv. A key for the variable names in the header of this csv file is provided below. The data were minimally processed for flexible use in behavioral analyses and model-fitting. Both the Rmarkdown code and the Matlab code take this file as input. Summarized data from reported simulations can be found in accTable_realParams_sims.txt and are used in the Rmarkdown file to generate figures. Model-fitting analyses require the mfit package. Data were loaded into Matlab using load_data.m and separated by age groups using sepDataAgeGroup.m prior to model-fitting.
subject: randomly generated subject ID
usable: subjects to be included in analyses (1 = include)
age: exact age of participant
age_group: participant’s age group designation
gender: 0 = male, 1 = female
version: the territory and trial presentation order
block_num: 1 = first learning block, 2 = second learning block, 3 = third learning block
condition: condition (1 = robber, 2 = millionaire, or 3 = sheriff)
trial_num: trial number
trial_in_block: trial number within each learning block
mine_prob_win_left: probability of a positive outcome for the stimulus on the left side
mine_prob_win_right: probability of a positive outcome for the stimulus on the right side
subj_choice: button press (0 = right, 1 = left)
feedback: reward feedback received (0 = negative outcome, 1 = positive outcome)
latent_guess: button press for subject guess about latent agent intervention (0 = no, agent did not intervene, 1 = yes, agent did intervene)
optimal_choice: whether the subject chose the better mine (0 = no, 1 = yes)
choice_RT: choice reaction time