“How are you doing today?” Ellie asks with a friendly smile. She tilts her head as she waits for a response. She wears black slacks, a blue shirt and a tan cardigan. She looks and behaves as expected for any therapist.
“Where are you from?” she asks as she sits in her pink armchair with her hands resting on her lap. Her brown hair is pulled back into a neat ponytail.
Ellie is not just any therapist. In fact, she is not even a therapist.
Ellie is a virtual human who lives in a computer at USC's Institute for Creative Technologies (ICT). USC Viterbi research assistant professor Louis-Philippe Morency, psychologist Albert “Skip” Rizzo and their ICT team created her as part of a project called SimSensei. The project is funded by DARPA as a way to help veterans dealing with depression and post-traumatic stress disorder, or PTSD, by trying to abolish the stigma around therapy and help clinicians as a decision support tool.
Of the nearly 900,000 veterans seeking treatment, about 30 percent suffer from PTSD, according to a study by the Department of Veterans Affairs. Those 300,000 people are just a fraction of those dealing with PTSD. The National Institutes of Health reports that PTSD affects nearly 8 million Americans.
All these people may not be receiving the attention they need. Signs of PTSD can often go undetected, both by clinicians and patients themselves. Through SimSensei technology, Ellie can detect subtle signs of psychological distress that might indicate PTSD such as tone of voice or shifting gaze. With this information, clinicians can better assess and treat patients. Morency compares this technology to a nurse checking a patient’s vital signs before the patient sees the doctor.
A patient may seem to be smiling frequently and report feeling happy, but the technology can reveal that that patient’s smile is tense, one sign of psychological distress that could indicate PTSD or depression.
“I believe that as a psychologist we can advance mental health, clinical care, and rehabilitation dramatically by applying engineering technology only dreamed of a few years ago,” said Rizzo, who leads ICT's Medical Virtual Reality Lab and began the project in the hopes of revolutionizing psychology in the 21st century.
“Are you a veteran?” Ellie asks as her questions grow in intensity. A “yes” tells Ellie to focus on this subject to gather more information.
The actual words may not reveal as much as the patient’s behavior does. Through three different sensors, a microphone, webcam and Kinect sensor, Morency has developed SimSensei's underlying technology to detect nonverbal cues and understand their meaning. His MultiComp lab at ICT explores the recognition and analyses of visual cues, such as head nods and eye shifts, to facilitate more human-computer interaction.
Morency's research identified a few key signs someone may be suffering from depression or PTSD. These people often smile for shorter periods of time and speak in a more monotone voice.
Analyzing nonverbal communication is not a new concept. Clinicians and non-clinicians alike analyze nonverbal cues every time they interact with a patient. However, these cues can be subtle and clinicians may not notice them immediately. The SimSensei technology assists the assessment process. Rather than replacing clinicians, it is designed to complement their skills in picking up these nonverbal cues.
“Sometimes people don’t even realize that it would be helpful for them to seek treatment,” Morency said. “These tools could help identify some of the indicators and allow for a better diagnosis or screening.”
Ellie has not always been this advanced. She emerged from a project called SimCoach started at ICT about five years ago. With SimCoach, patients typed their answers rather than verbalized them. Ellie’s ability to carry a real conversation enhanced the experience for the user.
“It was really good to be able to just talk, especially about things that don't usually come up,” said one male veteran who requested anonymity. “Her voice was comforting.”
Added a female veteran: “I liked talking to Ellie because it made me feel like I wasn't being judged.”
SimSensei is in its last round of testing at ICT. Then, the team will move on to testing larger groups of veterans for extended periods of time. The end goal is for this technology to be used in clinician’s offices to help support clinicians in assessing psychological status to improve diagnostic accuracy.
“If I can help just one person have a better life, that would be big,” Morency said.