How a computer program Ellie learned to recognize depression

Ellie developed to diagnose post-traumatic stress and depression. When you enter the booth, she first asks questions - about the family, about feelings, about the greatest regrets.

It may seem that these emotions are really confusing and difficult for the program to understand. But Skip Rizzo, the psychologist who helped develop Ellie, thinks differently.

When you answer Ellie's questions, she listens. But the program does not process words. She analyzes the voice. The camera in the smallest detail tracks facial expressions.
')

“Contrary to popular belief, depressed people smile as often as those who do not suffer from depression,” says Rizzo. “But their smiles are less cheerful and not so long. Rather, it’s a courtesy smile, not a real one coming from the inside.” .

Ellie compares my smile with a database of soldiers returning from battle. Is it real or forced? Ellie also tracks pauses. She watches as I look sideways or down. If I lean, she also notices.

This whole analysis seems to work: Ellie was able to detect signs of post-traumatic stress disorder and depression in affected people in the course of research.



Jodi Mitic served in the Canadian Forces in Afghanistan. He lost both legs due to a bomb blast. Mitic tells how robot Ellie helped him open up. “It seemed that Ellie was just listening,” he says. “When you start talking about some terrible things that you may have seen or done, many therapists — you can see it in their eyes — show the appropriate response.”

With Ellie this does not happen.

At the moment, the development is used strictly for diagnosis . The idea is that Ellie first identifies the soldiers who have problems, and then the doctors worked with them.