An online ‘depression detector’ developed in Israel can identify depression 78 percent of the time, according to a team of clinical psychologists.
You know that the Internet is a goldmine of information, but did you know that the Web is able to diagnose your emotional states and determine whether you are depressed, in love, feeling vengeful or aggressive? A new algorithm developed by Israeli researchers can pinpoint human emotions and intentions by studying vast amounts of publicly available data on blogs and other social media outlets.
Lead researcher Dr. Yair Neuman from Israel’s Bar-Ilan University has especially promising results for identifying depression. The data provided by his application could help human experts to home in on who might be at risk for suicide. And the application, which skims the Internet automatically, can be adapted to any sort of data he says, presenting intriguing possibilities for crime fighters, pollsters or Homeland Security.
Currently enjoying a short summer sabbatical at the University of Toronto, Canada, Neuman will present his research in late August at the 2010 International Conference on Web Intelligence. He tells ISRAEL21c that he doesn’t imagine that his invention will replace the human element, but rather that he views it as an aid to help sift through and make sense of the billions of bits of data about the human experience that are available online.
Neuman has used his application, which is funded by Israel’s Defense Ministry, to scan more than 300,000 English language blogs linked to mental health websites. The software was programmed to identify the 100 most depressed, and the 100 least depressed individuals, by analyzing their responses to metaphors and questions.
The resulting data was handed over to a team of clinical psychologists, who confirmed a high correlation to what they would diagnose in the clinic, reporting that the software was able to identify depression 78 percent of the time.
New software could pinpoint would-be terrorists
“The software program was designed to find depressive content hidden in language that did not mention the obvious terms like ‘depression or ‘suicide’,” Neuman relates. “A psychologist knows how to spot various emotional states through intuition. Here, we have a program that does this methodically through the innovative use of ‘Web intelligence’.”
“I emphasize that the tool cannot substitute for an expert. It can provide a powerful way to screen for depression through blogs and Facebook. It analyzes text – the written language – and it can help us to identify people who are presenting signs of depression,” he tells ISRAEL21c.
“Most who suffer from depression won’t commit suicide. But this is a powerful way of screening for depression. It has psychosocial applications as well for Homeland Security – let’s say monitoring depression levels of the population in a political circumstance to understand the population’s sentiment.”
The software can also “excavate” the meaning of context Neuman continues, testing for love, revenge and happiness among people, or even the way a society feels about its minorities. To date, the standard way to gain insight into these issues is through costly and time-consuming public opinion polls. The identification of people prone to vengeful behavior might also help to pinpoint would-be terrorists or criminals, he adds.
Another Israeli researcher, from the Hebrew University of Jerusalem, recently developed an online sarcasm detector, but Neuman says his approach is different: “I am analyzing the metaphor by sending queries to the web as questions, and understanding the way people metaphorically describe the answer.
“We could ask: What is love like? What is depression like? And look for how people describe the personal experience. We are harvesting a web of metaphors,” he explains.
Combining the wisdom of people with IT
For now his tool won’t be open source [meaning that for now the public does not have access to the end product's source material] and with no immediate plans for commercialization, Neuman isn’t commenting on which, if any, agencies may be interested in the application. Still, his “no comment” suggests that it may be of interest to the US defense department. “It’s not artificial intelligence, but intelligence augmentation, combining the wisdom of people with IT to mine a goldmine of data,” he says.
The Mayo Clinic for example, offers a list of clues that someone may be suicidal and includes in its watch list “statements about hopelessness, helplessness or worthlessness, sudden changes in mood, direct or indirect statements that reference death or dying and isolating oneself from friends or family.” These are signs loved ones can look out for, but what if the signs aren’t obvious? This is where the Israeli application could help.
In the US alone, about 35,000 people commit suicide each year and the problem of diagnostics remains a challenge, since people identified as suicidal are those who are already seeking help, or those who have been ‘diagnosed’ through an online questionnaire. With increasing software advances, in the future diagnosticians will be able to scan the online world for more proactive signs of depression and suicide, for better intervention, monitoring and prevention.
While developed for academic purposes, the new Israeli findings could be used to screen for suicidal teens and bloggers, mining the Internet for the clues they may leave about their intentions, perhaps enabling friends and relatives to intervene in time, if someone they care about is at risk.