Software might know if you're depressed

Israeli researchers develop a tool called Pedesis that spots words, phrases, and even metaphors that indicate depression in online text.

Leslie Katz Former Culture Editor
Leslie Katz led a team that explored the intersection of tech and culture, plus all manner of awe-inspiring science, from space to AI and archaeology. When she's not smithing words, she's probably playing online word games, tending to her garden or referring to herself in the third person.
  • Third place film critic, 2021 LA Press Club National Arts & Entertainment Journalism Awards
Leslie Katz
3 min read

depression keywords
The semantic field of depression as identified by the Pedesis system. Blog posts like "I feel so sad and I can't figure out why help" scored high on the depression scale. Yair Neuman/Ben Gurion University of the Negev

A software program under development in Israel can supposedly detect depression in online communication, and not just through obvious indicators like "I'm sitting here alone in the dark mulling how much my sorry life sucks."

Instead, it purportedly can identify depressive meaning hidden in language that doesn't necessarily include glaring terms like "depression" or "suicide." Yair Neuman, an associate professor in the department of education at Ben-Gurion University of the Negev and leader of the interdisciplinary team that developed the software, suggests the program could be used to conduct initial screenings of text penned by people who may not even realize they have a problem, thereby raising self-awareness and hopefully leading to medical help.

woman at computer appearing depressed
Ben Gurion University of the Negev

The program spots words, phrases, and even metaphors, to detect possible signs of depression (anxiety, sadness, preoccupation with self and with death). For example, words like "black," combined with terms such as "sleep deprivation" or "loneliness," will be recognized by the software as "depressive" texts.

To understand similarities in the way people describe the blues, the researchers conducted searches using Microsoft's Bing and extensively analyzed the word pattern "depression is like..."

They then tested the program, called Pedesis, by scanning more than 350,000 English-language texts from 17,031 bloggers (with the permission of the writers), as well as 1,600 online queries addressed to mental health experts at sites like MentalHealth.net. Once the program identified texts as depressive, a panel of four clinical psychologists reviewed 200 examples from that category. The verdict of the computer program and the analysis of the human panel correlated 78 percent of the time, according to the researchers.

"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,'" said Neuman, who specializes in semiotics and psychology and will present his team's work at the 2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agency Technology in Toronto later this summer.

Despite the preliminary nature of the system, the idea is that it could eventually serve as an additional avenue for identifying individuals in need of treatment. It could, for example, be installed by Web sites focused on consumer mental health, with a pop-up tool indicating if user comments post indicate a depressive pattern.

"No one can actually replace excellent human judgment," Neuman noted. "The problem is that most people are not aware of their situation and they will never get to an expert psychologist."

The tool can also analyze online text for word patterns indicating love and vengeance, according to Neuman. But for now the focus remains on depression.

Of course, it seems like the software could present a sizable potential for false positives. What if a blogger were to pen a sentence like, "My friends hate me because I can't stop talking about how much I'd die for the band The Black Sleep Deprivation of Loneliness"?

Remember that sarcasm-detecting algorithm we told you about not too long ago? That might come in handy here.