Data and material for the experiments and the modeling reported in Rothe, Deverett, Mayrhofer, & Kemp (in press). Successful structure learning from observational data. [preprint]

---

data_by_subject.csv contains the causal structures that participants inferred in Experiments 1 to 5.

Columns:

• exp_paper Experiment number, same ordering as reported in the paper
• subject Participant ID (note, subject 1 in Experiment 1 is not the same participant as subject 1 in Experiment 2, etc.)
• block Block number, same ordering as reported in the paper
• repetition Either 1 or 2 (only relevant for Experiment 1, where each block was shown two times)
• graph ID of the causal structure drawn by the participant (for all 64 causal structures and their IDs, see graphs_overview.pdf shown below or the identical Figure C1 in the Appendix)

---

data_model.csv contains the aggregated responses from participants and the model predictions

Columns:

• exp_paper
• block
• graph In this csv file, for each block in each experiment, graph ranges from 1 to 64
• subject_rel The percentage of participants that drew this graph in this block (i.e., sums to 1 within each block)
• BSL ... LSL Model predictions = model posteriors over graphs (i.e., each posterior sums to 1 within each block)

---

instructions_outline.pdf outlines the instructions for Experiment 1 (the footnote "Original" was not shown in the experiment). The instructions for Experiments 2 to 5 were built upon these core instructions (see Table A.1 in the Appendix for an overview and descriptions in the main text for details).

---

graphs_overview.pdf