When Joy Feels Wrong: Identifying Key Dampening Features Predicting Depressive Symptoms Using Machine Learning and Network Analysis

Abstract

Dampening of positivity is implicated in increased depression risk, yet research traditionally overlooks dampening’s and depression’s multifaceted natures and differential relations between dampening features and depressive symptoms. In this preregistered study, we pooled data from 13 studies, yielding four cross-sectional (N = 4,015; 13–86 years) and four longitudinal (N = 1,457; 14–86 years) data sets grouped by measures. Random-forest (RF) and network analyses examined the predictive utility of individual dampening features for specific symptoms. Across both analytic approaches, dampening features most strongly predicted core cognitive-affective symptoms, such as negative self-perceptions, pessimism, pervasive negative emotions, and to a lesser extent, fearful feelings. Concurrently, both approaches showed that the features on not deserving positivity, positivity not being long-lasting, and positivity being likely to end soon had consistently high predictive utility. The latter two emerged as longitudinal predictors in the RF analyses. Findings refine the relation of dampening to depressive symptoms and highlight intervention targets.