The response on Twitter to the 2014 Australian budget was furious
Over-sharing on Twitter might prove to be a boon for mental health services. Grabbing 750 tweets a second, a new tool can read the emotional state of a region in real time. The idea is to figure out exactly what kinds of events affect people鈥檚 moods and tailor mental health treatments accordingly.
Researchers at Australia鈥檚 national science agency, the CSIRO, and the Black Dog Institute in Sydney, created an emotional vocabulary of about 600 words and confirmed their meaning by crowdsourcing responses from over 1200 people. They built an app that filters tweets by location and linguistically analyses their emotional content. The output is an interactive graph of the target region鈥檚 mood. It shows how much each of seven emotions are being expressed in that region.
Advertisement
鈥淚f it works then in the future we can monitor, and eventually predict, where services can be assigned,鈥� says C茅cile Paris, a computer scientist at the CSIRO. The plan is to make the app available for researchers anywhere.
Meaningful connections
The first tests of the system will be to show that the bumps in the graph correspond to meaningful emotions in the real world. To do that, the team will look at known correlations between events, regions and mental health and see if they are reproduced in the system.
鈥淔or example, we know that rates of suicide are higher in remote areas,鈥� says psychiatrist Helen Christensen from the Black Dog Institute in Sydney, Australia, which investigates mood disorders. 鈥淪o if we can show that, for example, there are more distressed tweets happening in those particular areas, that gives us some validation.鈥�
Coincidently, the week before the app was launched, a perfect test popped up: the Australian government announced the harshest budget cuts for decades. Twitter was awash with emotional outbursts. 鈥淪o much to be devastated and furious and ashamed about #budget2014鈥�, wrote one. She wasn鈥檛 alone (see image, top).
We examined emotions during the budget and compared this with the previous week, says Helen Christensen of the Black Dog Institute. 鈥淲e had initial spikes of fear in anticipation, a huge but rapid spike in surprise and a slow increase of sadness, which dissipated [by] about 11 pm.鈥�
Early warning systems
鈥淚f you can actually link this with real world outcomes, that鈥檚 very powerful,鈥� says , a psychiatrist at the Harvard Clinical and Translational Science Center who researches social networks. He says it could be a great tool for mental health services.
鈥淥ne would presume that there would be these triggering events that people could observe. Then you鈥檇 have a system in place to address these things that are happening. If there鈥檚 a storm coming in, you wear your rain coat or batten down the hatches,鈥� he says. 鈥淯sing social media you could see an emotional storm.鈥�
Ian Hickie, a psychiatrist at the University of Sydney in Australia, says that until now monitoring how events affect people鈥檚 mental health has been 鈥渂ackward looking鈥�. Researchers have spent months or years establishing connections after the event. Analysing social media in real time could make the process more like the monitoring of infectious diseases, with the eventual creation of early warning and predictive systems, he says.
鈥淭his is part of a big burgeoning world,鈥� says Rosenquist. 鈥淭hese technologies, if you build them right, they鈥檙e relatively costless so the potential is incredibly high and there鈥檚 a big need.鈥�
