Task-specific AI-reliance prediction reproducibility package

Version 2.0.0 reproducibility code and deidentified derived features for the manuscript Prediction of a Task-Specific AI-Reliance Choice from Facial Expression and Galvanic Skin Response Signals, by Hamid Shamszare and Avishek Choudhury. The package reproduces statistical results from the fixed N=60 cohort using repeated participant-level cross-validation, cross-fitted stacking, bootstrap intervals, and paired AUC comparisons. It includes reference outputs and verification instructions. Raw biometric recordings and raw-feature extraction are outside the shared package. The study is exploratory, and the results do not establish generalizable predictive ability or a general advantage or lack of value of multimodal fusion. Code is licensed under MIT; derived data and documentation are licensed under CC BY 4.0. Generative AI assistance and numerical validation are disclosed in the README.