Computer Networks that Help People Stay Sober
As a member of the World Well-Being Project, a research group at Penn that uses machine learning to enable computers to better understand people’s personalities and emotions, as well as their mental and physical health, Lyle Ungar is interested the way that users express themselves on social networks. The specific words that people employ in social media posts may hold deeper insights into what they are feeling or doing, beyond what those messages say on the surface.
Ungar, a professor in the Department of Computer and Information Science, is currently collaborating with Brenda Curtis of the Perelman School of Medicine and Sober Grid, a social network for people recovering from addiction, on a study investigating whether language can predict relapse.
Starting with anonymized Sober Grid posts, along with self-reported timelines of those users’ sobriety, the team is using machine learning to determine if trends in the emotional language of posts preceding a relapse can be used to predict them in other Sober Grid members. Ideally, a members showing signs of a potential relapse could received tailored messages or interventions that would help them cope.
Starre Vartan reported on the study in The Daily Beast:
There’s more than just predicting relapse and addressing it in situ: “For me, the interesting thing is… identifying the potential drivers of relapse for different people. Can we use statistical models to identify, for a given person, what is happening in their lives that warns about possible relapse — or that suggests healthy recovery?” Unger asked.
Continue reading at The Daily Beast.