We also observe better performance of the model that
uses the linguistic style features alone. Results in prior literature
suggest that use of linguistic styles such as pronouns
and articles provide information about how individuals
respond to psychological triggers (Rude et al., 2004;
Ramirez-Esparza et al., 2008). Next, we note that, one of
the main characteristics of depression is disturbed cognitive
processing of information as indexed by disturbed startle
reflex modulation, as well as a reduced sense of interest
or motivation in day-to-day activities (Billings et al., 1984;
Oxman et al., 1982). Hence we observe better performance
of depression language features in the prediction task. Finally
the better performance of ego-network features
shows that the network in which depressed individuals are
embedded, serving as a proxy to their social and behavioral
environment, bears key information in light of their condition.
In essence, we conclude that social media activity
provides useful signals that can be utilized to classify and
predict whether an individual is likely to suffer from depression
in the future.