cated with other classifiers, more geographic regions, and/or
a more explicit group of military users but it is at least sug-
gestive of the sort of population-level analysis enabled by
this data and these techniques.
There remain several important open questions. Do users
who self-report diagnoses differ from other diagnosed indi-
viduals, perhaps sharing more relevant mental health infor-
mation? What other mental health conditions can be stud-
ied using our approach of identifying self-diagnoses? Fi-
nally, what opportunities exist for interventions with identi-
fied users? What linguistic signals are present in social me-
diabutnotcapturedbyLIWC?Pursuingtheanswerstothese
questions will provide many exciting opportunities for men-
tal health research and help to address this serious public
health concern.