Same Stimulus, Different Ideas - Analysis Code and Derived Data

Supporting materials for the manuscript Same Stimulus, Different Ideas: Prior Knowledge and the Integration of AI-Generated Stimuli in Creative Ideation, under review at the Journal of Intelligence (MDPI). This is version 2 of the deposit. It accompanies the revised manuscript and replaces the analysis of version 1. It contains the analysis code (semantic and conceptual similarity, topic modelling, idea-level uncertainty estimates and mixed models), the derived data needed to check every data-based table and figure of the revised manuscript, and the plotting scripts and figures. The complete folder structure is in ideation_stimulus_integration_v2.zip; README.md, CODEBOOK.md and the derived CSV files are also provided individually. Restricted material. The verbatim textual corpus of ideas, external stimuli, design-problem statements and traditional-knowledge inventories is not included. It embeds traditional knowledge of the participating Indigenous and Afro-descendant communities, and its unrestricted publication would breach the collective consent agreements established with their traditional authorities. The only idea texts in the deposit are those already published in the manuscript. The corpus is available on request from the corresponding author, subject to authorisation by the participating communities. Changes from version 1. All indicators were recomputed on the twenty stimuli presented to each group, with a corrected conceptual pipeline and corrected transcription errors; topic modelling is now reproducible, with coherence, stability across 30 models and a chance baseline; new sensitivity, robustness and uncertainty analyses were added. Version 1 listed the semantic model incorrectly as all-MiniLM-L6-v2; the model used is paraphrase-MiniLM-L6-v2. See README.md for details.