It's kind of addressed in section 5: "We cannot be certain that the correlations shown in the data are evidence of causality, but we think it is appropriate fo

It's kind of addressed in section 5: "We cannot be certain that the correlations shown in the data are evidence of causality, but we think it is appropriate for those who care for children to act on the preponderance of the evidence (which is the standard in a civil trial) rather than waiting for evidence beyond a reasonable doubt (which is the standard used in a criminal trial." It's not that social (or medical) researchers are unaware that observational studies can't establish causation, but in most cases when studying human reactions nothing better is feasible. For a proper randomized controlled trials even voluntary enrollments are problematic: the researcher needs to choose and randomize the treatment. How can that ever be implemented for a sufficiently powered study of smartphones/social media use for children (or adults)?