Modelling Online Misinformation Susceptibility Using Hybrid AI–Statistical Techniques

Abstract

The rapid expansion of social media platforms has significantly increased the spread of misinformation, influencing public opinion, political behaviour, and social decision-making. Identifying individuals who are more susceptible to online misinformation has therefore become an important problem in computational social science. This study proposes a hybrid AI–statistical framework for modelling misinformation susceptibility using logistic regression integrated with artificial intelligence-based analytical techniques. Socio-demographic, behavioural, and psychological variables are combined with AI-derived indicators obtained through natural language processing and user interaction analysis. The proposed methodology employs logistic regression as an interpretable statistical foundation while incorporating machine learning-assisted feature extraction and variable selection to improve predictive performance. Explainable AI tools are further utilised to assess variable importance and model transparency. The framework enables both accurate classification and interpretable inference regarding factors associated with misinformation vulnerability. Simulation studies and empirical analysis using social media datasets demonstrate that the hybrid approach outperforms conventional logistic regression models in predictive accuracy and robustness. The proposed framework contributes to computational social science by combining statistical interpretability with modern AI-enhanced learning techniques for understanding online misinformation behaviour.;