This recently published paper by Selim Sametoğlu, Dirk H. M. Pelt, and Meike Bartels, is based on a clever idea, namely to look at twin studies to see if heavy social media users have innate tendenciees toward lower social well-being.  Overall the results are not encouraging for seeing a strong causal connection here:

Meta-analyses report small to moderate effect sizes or inconsistent associations (usually around r = -0.10) between wellbeing (WB) and social media use (SMU) and between anxious-depressive symptoms (ADS) and SMU (also around r = 0.10). This study employs the classical twin design, utilizing data from 6492 individuals from the Netherlands Twin Register, including 3369 MZ twins (893 complete twin pairs, 1583 incomplete twin pairs) and 3123 DZ twins (445 complete, 2233 incomplete) to provide insights into the sources of overlap between WB/ADS and SMU. Both hedonic and eudaimonic WB scales were used. SMU was measured by (1) the time spent on