Two common practices that minimize evidence of risks associated with the use of social media

Why do researchers analyzing the same datasets reach different conclusions about whether social media use poses risks to adolescent mental health? We argue that two common practices can minimize evidence of harm and lead to a systematic divergence in conclusions. First, many studies emphasize correlation coefficients without evaluating risk elevation, thus obscuring concentrated risk to heavy users. Second, researchers often rely on highly blended analyses that combine exposures (e.g., multiple digital activities), outcomes (e.g., well-being composites), and populations (e.g., girls with boys). Such blending can dilute the core issue in the public health debate—the association between heavy social media use and internalizing symptoms in adolescent girls. We illustrate how sequential blending can transform a clinically relevant association into a negligible one, which helps explain the persistent disagreement in the literature. We conclude by highlighting two further limitations in the literature: limited platform-specific measurement and the undercounting of harms beyond internalizing outcomes.