Character Strengths as Universal Predictors of Health? Using Machine Learning to Examine the Predictive Validity of Character Strengths Across Cultures

Abstract

ObjectiveCharacter strengths are positive personality traits that not only define our core identity, but also yield positive effects for us and those around us. Psychological research has often been one-sidedly focused on tackling health risk factors or maladaptive traits, disregarding the potential of fostering positive resources such as character strengths when aiming to influence health trajectories. We examine the predictive validity of character strengths for health-related outcomes using machine learning algorithms.MethodsUsing a sample of 4,830 adults from five countries, we examined the validity of character strengths for the prediction of 12 health-related indicators (e.g., sleep quality, feeling anxious, or healthy dieting) across two levels of measurement (items vs. scales), modeling approaches (multiple regression vs. three machine learning algorithms), and cultural contexts.ResultsThe outcomes could be predicted by character strengths with R² values ranging from .02 for the prediction of poor physical health to .28 for poor mental health. Character strength items rarely out-predicted their overarching scales. Machine learning algorithms were able to enhance predictive performance by means of regularization, but the results did not point to meaningful non-linear or interaction effects. The largest differences in predictive performance were found when evaluating models across culturally dissimilar countries.ConclusionsCultural context proved an important moderator of the association between character strengths and mental as well as physical health indicators. In contrast, the incremental value of character strengths at the item level or including complex relationships in the modeling compared to simpler modeling approaches is negligible.