Social media platforms cater to a fundamental human need for social connections, but, surprisingly, their use may be associated with lower levels of individual well-being and increased mental health issues. When we analyze the valence levels of individuals' online language before and after they express having a strong emotion, we can observe emotional levels rapidly returning to baseline levels within minutes after their expression. These results reveal the possible beneficial effects of putting one's feelings into words, otherwise known as affect labeling, even in an online setting. However, analyzing longitudinal indicators of subjective well-being in online language and comparing the results to an individual's social relations, we observe a strong homophily of subjective well-being in social networks as well as a significant
friendship and happiness paradox. The latter indicates that many social media users may experience a situation in which their friends on average seem more more popular and happier than they are. These results may shed light how social media use can negatively affect a population's emotional and social health. I will conclude with a brief overview of our present efforts to model the morbidity, online correlates, and longitudinal dynamics of mental health issues from large-scale social media data. If time allows I will also briefly discuss a proposal that could greatly enhance the mental health of scientists looking for funding.