Data from: Freedom through understanding: Instructed knowledge shapes voluntary action choices
Data files
Dec 06, 2025 version files 19.92 MB
-
autoreg_data_b1-2_grp1-3.xlsx
1.45 MB
-
autoreg_model.mat
1.37 KB
-
clmm_autoreg_data_b1-2_g1.xlsx
5.65 KB
-
clmm_autoreg_data_b1-2_g2.xlsx
5.65 KB
-
clmm_autoreg_data_b1-2_g3.xlsx
5.66 KB
-
data_fit_models_exclude_timeout.mat
9.17 MB
-
dataframe.mat
7.92 MB
-
gen_models_exto.m
3.96 KB
-
MLE_models_exto.m
7.01 KB
-
modelfitindices.m
190 B
-
myfigAI2.m
442 B
-
README.md
2.84 KB
-
Step1_choicebias.m
6.33 KB
-
Step2_fit_models_instruction.m
13.28 KB
-
Step3_LaggedRegression_data.m
1.71 KB
-
Step3_LaggedRegression_model.m
7.69 KB
-
Step4_LaggedRegression_analysis_saved_results.RData
1.29 MB
-
Step4_LaggedRegression_analysis.R
12.65 KB
-
Step5_LaggedRegression_plot.m
7.39 KB
-
Step6_model_plot.m
12.12 KB
Jan 08, 2026 version files 19.92 MB
-
autoreg_data_b1-2_grp1-3.xlsx
1.45 MB
-
autoreg_model.mat
1.37 KB
-
clmm_autoreg_data_b1-2_g1.xlsx
5.65 KB
-
clmm_autoreg_data_b1-2_g2.xlsx
5.65 KB
-
clmm_autoreg_data_b1-2_g3.xlsx
5.66 KB
-
data_fit_models_exclude_timeout.mat
9.17 MB
-
dataframe.mat
7.92 MB
-
gen_models_exto.m
4.09 KB
-
MLE_models_exto.m
7.01 KB
-
modelfitindices.m
190 B
-
myfigAI2.m
442 B
-
README.md
2.84 KB
-
Step1_choicebias.m
6.33 KB
-
Step2_fit_models_instruction.m
13.28 KB
-
Step3_LaggedRegression_data.m
1.71 KB
-
Step3_LaggedRegression_model.m
7.74 KB
-
Step4_LaggedRegression_analysis_saved_results.RData
1.29 MB
-
Step4_LaggedRegression_analysis.R
12.65 KB
-
Step5_LaggedRegression_plot.m
7.39 KB
-
Step6_model_plot.m
12.12 KB
Abstract
The capacity for voluntary action is a distinctive feature of human minds. However, experimental studies of volition struggled to capture defining features of human voluntariness. Here, we developed a competitive game which incentivised participants to innovate their action choices to find the right time to avoid a collision with an opponent who predicted the timing of the participant’s action choice. One group of participants received explicit information about the competitor’s action-selection rules, while a second group had no information about the competitor. Both groups showed increased behavioural stochasticity when adapting to a competitor who punished participants’ choice biases. However, the group that had no explicit information generated their action choices in a way that avoided the action that the competitor was likely to take. In contrast, the group that explicitly knew the competitor’s action-selection rules avoided the same action they took in preceding trials so that the competitor could not easily exploit the participant’s behavioural patterns. These findings suggest that people can develop beliefs about other agents in the social environment within which they work, and can adapt voluntary action choices accordingly. However, explicit explanations about the other agent facilitate model-based planning in the voluntary generation of novel action patterns.
https://doi.org/10.5061/dryad.47d7wm3rb
Description of the data and file structure
The file is stored in MATLAB (.mat) format. See dataframe.mat
There are 34 fields under the structure of "dataframe.mat"
Each field contains trial data for 219 participants (one participant per one cell). Each cell contains all trials data for the corresponding participant.
Important fields that were used to generate figures are:
- "iBlock": the number of blocks; 1 (baseline block) or block 2 (test block)
- "grp": group; 1 (control group), 2 (no instruction group) or 3 (instruction group)
- "point2": trial success; 1 (successfully avoided birds) or 0 (failed)
- "wt": wait time until a key response in second
- "wtbin2": the interval the participants selected; 1 (early), 2 (middle) or 3 (late interval)
- "bet": the interval the competitor selected; 1 (early), 2 (middle) or 3 (late interval)
Code/software
Follow Step1 to Step 6. These scripts generate a list of figures used in the manuscript, run statistical analyses and perform model fitting and comparison.
- myfigAI2.m: a function to make cosmetic changes on plots
- modelfitindices.m: a function to compute AICc values based on the results of the model fit
- Step1_choicebias.m: computes the success rates and choice bias scores across 2 blocks and 3 groups, which is used to generate Figure 2.
- Step2_fit_models_instruction.m: performs the maximum likelihood estimation to fit models described in the paper to the data (dataframe), using the function 'MLE_models_exto.m' and '
gen_models_exto.m'. The results of the fits are saved in a mat file "data_fit_models_exclude_timeout.mat".
- Step3_LaggedRegression_data.m: makes a dataframe which lists a trial-wise choice behaviour (selected intervals) in relation to choices the participants and the opponents made on the past trials and saves this dataframe as an xlsx file 'autoreg_data_b1-2_grp1-3.xlsx'. When choices are not registered due to timeout, the code puts 'na' in the corresponding cells.
- Step3_LaggedRegression_model.m: makes a trial-wise behaviour predicted by models and saves a dataframe as 'autoreg_model.mat'
- Step4_LaggedRegression_analysis.R: performs a generalised mixed-effects model on the data generated in Step 3 and save the results of GLM as 'Step4_LaggedRegression_analysis_saved_results.RData'. The fixed effects are tabulated in the files 'clmm_autoreg_data_b1-2_g1.xlsx', 'clmm_autoreg_data_b1-2_g2.xlsx' and 'clmm_autoreg_data_b1-2_g3.xlsx'.
-
Step5_LaggedRegression_plot.m: plots the fixed effects of GLM (Fig. 3A&B).
-
Step6_model_plot.m: performs the Bayesian model selection based on the model fit performed in Step 2 and plots the exceedance model probability (Fig 3C).
The behavioural data on healthy human participants was obtained through the online platform Prolific.
Changes after Dec 6, 2025: comments for parameters of models added to files "Step3_LaggedRegression_model.m" and "gen_models_exto.m"
