Explainable Modeling of Stakeholder Trust and Skepticism Toward AI-Generated Content Across Different Models
Computer and Decision Making: An International Journal · 02.11.2026 · 3 (6S) · 3270-3284
DOI: 10.1922/comdem.v3i6s.19066 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 02.11.2026. Verfasst von Abdulwahab M. Alsulami. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Given the proliferation of AI-generated content, understanding how users evaluate and trust these responses is crucial. A user’s trust or skepticism varies based on AI-response features, individual user traits, decision-relevant factors, and verification behavior. However, standard machine learning methods focus heavily on prediction accuracy, offering limited insight into the underlying factors driving user skepticism. To address this gap, this study proposes an Explainable Tabular Learning framework for Modeling Trust and Skepticism (XTL-MTS) toward AI-generated Content to analyze user skepticism toward AI-generated responses. Unlike traditional machine learning models that mainly focus on prediction accuracy, this integrated approach aims to make the reasoning behind the predictions more transparent. The proposed XTL-MTS framework considers AI-response characteristics, user characteristics, decision-related factors, and verification behavior to distinguish different user skepticism categories. HSIC Lasso is employed to select informative features based on their statistical dependence on the target while reducing redundant information. The selected features are then provided to the TabNet model for multi-class classification of Blind Trust, Moderate Trust, Skeptical, and Highly Skeptical users. To improve the training stability of the TabNet model, the Lookahead optimization strategy is applied during model training. Furthermore, SHAP is employed to explain the model predictions and identify the contribution of different AI-response and user-related characteristics to the predicted skepticism categories. The experimental evaluation of the proposed XTL-MTS framework is performed using the AI skepticism dataset, and the obtained results demonstrate its effectiveness in modeling user skepticism toward AI-generated content, achieving an accuracy of 96.67% and an F1-score of 93.66%.
Zitieren
APA
Alsulami, A. M. (2026). Explainable Modeling of Stakeholder Trust and Skepticism Toward AI-Generated Content Across Different Models. Computer and Decision Making: An International Journal, 3(6S), 3270-3284. https://doi.org/10.1922/comdem.v3i6s.19066
Vancouver
Alsulami AM. Explainable Modeling of Stakeholder Trust and Skepticism Toward AI-Generated Content Across Different Models. Computer and Decision Making: An International Journal. 2026;3(6S):3270-3284. doi:10.1922/comdem.v3i6s.19066
BibTeX
@article{Alsulami2026,
author = {Alsulami, Abdulwahab M.},
title = {Explainable Modeling of Stakeholder Trust and Skepticism Toward AI-Generated Content Across Different Models},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {3270--3284},
doi = {10.1922/comdem.v3i6s.19066}
}
EUVIMED