Validation of social science theories using machine learning models: a methodological perspective

There is a critical need to validate Social Trust Theory, Political Participation Theory, and Happiness and Well-being Theory using modern methodologies. This study employs machine learning models—Random Forest (RF) and Support Vector Machine (SVM)—applied to longitudinal data from 1972 to 2023 across six diverse countries. The findings reveal that Social Trust (24.5%) is the most significant predictor of societal cohesion, followed by Happiness Score (19%) and Income (16%), underscoring their central roles in shaping social outcomes. The results demonstrate the models' ability to capture complex, non-linear interactions among variables, surpassing traditional econometric approaches. Specifically, RF identified critical socio-demographic predictors of political participation, while SVM highlighted the interplay between cultural values and economic stability in determining well-being. These insights advance computational social science by enhancing the accuracy of theory validation and offering actionable recommendations for policymakers, such as targeting income inequality and fostering institutional trust. This research bridges computational and traditional methods, presenting a scalable framework for analyzing evolving social phenomena.