A Hybrid Structural Modelling and Explainable Machine Learning Framework for Human–AI Strategic Decision Support Systems
- 1Deanship of Development and Quality Assurance, King Faisal University, P.O. Box 380, Al-Ahsa 31982, Saudi Arabia
Abstract
Hybrid frameworks combine structural modelling, machine learning, and explainable artificial intelligence (AI) to evaluate human–AI Strategic Decision Support Systems (SDSSs). However, their methodological reliability requires systematic assessment. This study evaluates a hybrid framework integrating partial least squares path modelling (PLS-PM), predictive machine learning, and explainable AI. It uses a reproducible Monte Carlo simulation based on the Millennial Leader–AI Agent Resistance (MLAAR) model. The model includes six socio-technical antecedents, strategic decision-making effectiveness, and institutional resilience. The experiment uses a 3 × 2 × 2 factorial design. It examines three sample sizes (N = 200, 500, and 1000), two antecedent-correlation levels (ρ = 0.0 and 0.3), and two functional forms (linear and nonlinear). Each condition includes 100 replications, producing 1200 synthetic datasets. Structural recovery, bootstrap inference, predictive accuracy, and feature importance are evaluated against known population parameters. Mean relative coefficient bias increased in magnitude from −11.65% under independent antecedents to −29.97% under correlated antecedents. Correct identification of the strongest population antecedent increased from 70.0% to 96.3% as sample size increased. However, structural and SHAP predictor rankings agreed exactly only 22.3% of the time. Indirect-effect confidence-interval coverage declined from 65.8% to 9.8%, despite consistent statistical detection. Ordinary least squares regression achieved the highest average cross-validated predictive accuracy among the evaluated algorithms. These findings show that predictive accuracy, statistical detection, and predictor-ranking agreement do not guarantee accurate structural recovery. The study provides a reproducible framework for evaluating hybrid explanatory–predictive methods. Independent empirical validation remains necessary before applying the framework to organisational decision-making.
Related
- Explainable computational framework for subtype classification and biomarker identification in head and neck squamous cell carcinoma using machine learning
- Explainable machine learning framework with Granger-based temporal predictive validation for monitoring fiscal risk in national health insurance expenditure
- A Multimodal Explainable Deep Learning Framework for Alzheimer's Disease Diagnosis using 3D Magnetic Resonance Imaging and Clinical Data
- An explainable machine learning framework for cross-dataset chronic kidney disease risk prediction using hybrid data balancing
- Faithfulness Without Anchors: A Position Paper on the Structural Misalignments of Explainable AI Faithfulness Metrics for Feature Importance
BibTeX
@article{salem2026hybrid,
title = {{A Hybrid Structural Modelling and Explainable Machine Learning Framework for Human–AI Strategic Decision Support Systems}},
author = {Mostafa Aboulnour Salem},
journal = {Mathematics},
year = {2026},
month = oct,
doi = {10.3390/math14213885},
url = {https://www.mdpi.com/2227-7390/14/21/3885},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - A Hybrid Structural Modelling and Explainable Machine Learning Framework for Human–AI Strategic Decision Support Systems
AU - Mostafa Aboulnour Salem
PY - 2026
DA - 2026/10/24
JO - Mathematics
DO - 10.3390/math14213885
UR - https://www.mdpi.com/2227-7390/14/21/3885
AB - Hybrid frameworks combine structural modelling, machine learning, and explainable artificial intelligence (AI) to evaluate human–AI Strategic Decision Support Systems (SDSSs). However, their methodological reliability requires systematic assessment. This study evaluates a hybrid framework integrating partial least squares path modelling (PLS-PM), predictive machine learning, and explainable AI. It uses a reproducible Monte Carlo simulation based on the Millennial Leader–AI Agent Resistance (MLAAR) model. The model includes six socio-technical antecedents, strategic decision-making effectiveness, and institutional resilience. The experiment uses a 3 × 2 × 2 factorial design. It examines three sample sizes (N = 200, 500, and 1000), two antecedent-correlation levels (ρ = 0.0 and 0.3), and two functional forms (linear and nonlinear). Each condition includes 100 replications, producing 1200 synthetic datasets. Structural recovery, bootstrap inference, predictive accuracy, and feature importance are evaluated against known population parameters. Mean relative coefficient bias increased in magnitude from −11.65% under independent antecedents to −29.97% under correlated antecedents. Correct identification of the strongest population antecedent increased from 70.0% to 96.3% as sample size increased. However, structural and SHAP predictor rankings agreed exactly only 22.3% of the time. Indirect-effect confidence-interval coverage declined from 65.8% to 9.8%, despite consistent statistical detection. Ordinary least squares regression achieved the highest average cross-validated predictive accuracy among the evaluated algorithms. These findings show that predictive accuracy, statistical detection, and predictor-ranking agreement do not guarantee accurate structural recovery. The study provides a reproducible framework for evaluating hybrid explanatory–predictive methods. Independent empirical validation remains necessary before applying the framework to organisational decision-making.
SN - 2227-7390
ER -