Does Twitter know you better than you know yourself? That’s the question a group of scientists asked while studying what people tweet to determine if they’re at risk for depression.
Time spoke with Eric Horvitz, co-director of Microsoft Research Redmond, who believes that artificial intelligence might one day be sophisticated enough to scan your tweets for signs that you’re depressed.
Horvitz and his team have created a model that scans Twitter and predicts depression among users with 70 percent accuracy. However, the model is still in its infancy – it misses a significant number of users who may be depressed, and it produces false positives about ten percent of the time.
The study examined 476 Twitter users, and discovered that 171 were severely depressed. The team looked into each user’s Twitter history to uncover any potential hints – such as mentions of types of medication or their language choice – that they were suffering from depression. By comparing these results with the Twitter feeds of non-depressed users, they built the model that can predict depression before it is diagnosed in 70 percent of cases.
There are a variety of factors that indicate someone may be depressed on Twitter. The researchers looked at habits like when someone tweets, how often they interact with others and their language choice. And specific words, like “anxiety,” “nausea,” “nervousness” and “episodes” were used more frequently by depressed people.
The fact that Twitter is a public forum could lead to a surge of similar health-related research. The several billion tweets sent per week offer unprecedented insight into trends not only in depression, but also tracking flu outbreaks, vaccine research and fighting obesity.
(Stressed businessman image via Shutterstock)
- Facebook Outperforms Twitter, Google+ For Social Brand Interactions [STUDY]
- Want Your Tweets Indexed On Google? You Better Be Really Popular On Twitter [STUDY]
- Facebook, Google+, Snapchat, LinkedIn All Bigger Than Twitter In Norway [STUDY]
- 32% Of B2B Marketers Want Automated Social Media Integration [STUDY]