Data and Code for "AI-Assisted Interactive Narrative as an Integrative-Learning Design Task: Student Appraisal in Preservice Teacher Education"

Published: 3 September 2026| Version 1 | DOI: 10.17632/kkxscxz8dh.1
Contributors:
,
,

Description

This data-and-code package supports the article “AI-Assisted Interactive Narrative as an Integrative-Learning Design Task: Student Appraisal in Preservice Teacher Education” published in Thinking Skills and Creativity (https://doi.org/10.1016/j.tsc.2026.102366). The quantitative materials cover three independent cohorts totaling N = 1,064: a December 2024 primary cohort (N = 480), an independent December 2025 shortened-process cohort (n = 327), and an independent June–July 2026 subsequent same-course cohort (n = 257). A separate anonymous June 2025 different-course follow-up contains 317 response records from the same Year-2022 class population represented in the primary cohort. These records are non-additive, cannot be linked to the December 2024 records at the individual level, and are used only for unmatched descriptive comparison. The package supports descriptive summaries, dimension-reduction checks, project-cluster-robust regression, ordinal WLSMV analyses, estimator triangulation, sparse-category diagnostics, and equivalent-model sensitivity analyses. The qualitative materials include a post hoc reconstructed 203-record primary working corpus, a separate retained 229-response June 2025 working corpus, a 12-theme codebook, two frozen human-coder matrices, and human-only agreement statistics. Final descriptive frequencies retain the shared human code when the coders agreed and use the corresponding author’s adopted AI-assisted resolution when they disagreed; item-level records and rationales are provided for all 607 disagreements. The original unfiltered June 2025 export and contemporaneous documentation of its retention rule are unavailable. Identifiable platform exports, student identities, and identifiable student-project materials are excluded. README files and SHA-256 manifests document execution, provenance, file integrity, and evidentiary boundaries. These materials support reproducible analyses of task appraisal and implementation conditions; they do not establish causal effects, objective learning, creativity, internalization, transfer, paired cross-course change, or independent qualitative replication.

Files

Steps to reproduce

Download and extract TSC_102366_Data_and_Code.zip. Work from a copy of the extracted TSC-D-26-00817R2_Data_and_Code directory so that the archived reference outputs remain unchanged. The Python analyses require Python 3 with NumPy, pandas, and SciPy. The R analyses require the lavaan package. Exact software versions are recorded in the included session-information files: the primary ordinal CFA was originally run with R 4.3.3 and lavaan 0.6-17, whereas the subsequent-cohort, multigroup, and robustness analyses were run with R 4.4.1 and lavaan 0.6-21. Run the following commands from the extracted directory: python AIIN_analysis.py python AIIN_analysis_full.py python AIIN_qualitative_coding_reproduce.py Rscript --vanilla AIIN_lavaan_WLSMV_script_final.R Rscript --vanilla AIIN_added_cohort_WLSMV_script.R AIIN_added_cohort_lavaan_input.csv reproduced_added_cohort Rscript --vanilla AIIN_cohort_invariance_WLSMV.R AIIN_lavaan_input.csv AIIN_added_cohort_lavaan_input.csv reproduced_multigroup For the post hoc ordinal-CFA robustness analyses reported in Supplementary Table S14, run the five scripts sequentially using the commands documented in robustness/README.md. Compare the regenerated results with reproduced_quantitative_results.json, the archived lavaan_WLSMV_*, added_WLSMV_*, cohort_invariance_*, and robustness/results/ files. File integrity can be checked using the included SHA256SUMS.txt. AIIN_qualitative_coding_reproduce.py recomputes agreement, Cohen’s kappa, disagreement counts, and final theme frequencies from the two frozen human-coder matrices and checks them against AIIN_qualitative_coding_summary.csv. It does not regenerate or independently validate the 607 corresponding-author-authorized AI-assisted disagreement resolutions; those item-level decisions and rationales are archived in AIIN_qualitative_ai_assisted_adjudications.csv and the accompanying workbook.

Institutions

Categories

Educational Psychology, Educational Technology, Preservice Teacher Education, Generative Artificial Intelligence

Funders

Licence