Can AI Assist in Suicide Prevention?

There is a roughly one-in-five chance an individual will experience some form of mental illness in their lifetime. Moreover mental illness can strike anyone — even popular celebrities such as Ariana Grande, Michael Phelps, and Lady Gaga have spoken of their struggles.

Figures from nonprofit organization Mental Health America show the severe depression rate for US youth increased from 5.9 to 8.2 percent over the last five years. For reasons ranging from non-diagnosis to social stigmas, more than half of the 43 million Americans who suffer from some form of depression have received no treatment for their condition.

The numbers provide a sobering glimpse of the shortcomings in mental health care, and some researchers are suggesting AI might be able to address the problem.

AI Scientists and the Story of 662 Prevented Suicides

“I witnessed their pain and despair,” says AI scientist Zhisheng Huang, a tenured professor at the Free University of Amsterdam in the Netherlands and a professor at Beijing’s Capital Medical University. In March 2018 Huang invented a “tree hole bot” that eavesdrops on the dark corners of social media — discussion boards where the severely depressed post their suicidal thoughts and plans.

Huang says the tree hole bot captures suicidal confessions with 82 percent accuracy. It classifies potential suicide markers on a 1-10 grading scale and sends Huang and his volunteer team an alert when the severity exceeds level 6.

Over the past year the tree hole bot has flagged cases such as a woman who wrote “Anyone want to jump in the river with me?” and another who live-streamed herself swallowing 60 sleeping pills on her social media feed. Both were rescued through notification of police and/or family members.The Tree Hole Rescue Group says that as of August 2019 it had stopped 662 people from committing suicide.

“Of course you can do more research, publish more papers, or get more money…” Huang says, “ but if you don’t save someone dying before your eyes, it’s very painful.”

Huang is one among a growing breed of researchers seeking to leverage AI to tackle pressing social problems. Last year, Professor Fei-Fei Li’s Stanford University team devised an automated diagnostic system that combines 3D facial recognition and NLP technology that can detect depressed individuals with 83.3 percent accuracy. The training uses DAIC-WOZ data, including a questionnaire completed by 142 patients and 189 clinical interviews.

Growing Market for AI-Backed Mental Health Healing

Although AI products and services in the mental health field remain relatively immature, a number of companies and institutions are working to develop new and improved apps and devices. Real-time data collection, mental health predictive analytics and sentient consultation are focus areas for such research.

Synced has compiled a list of currently available solutions.

Psybot APP
Provides CBT cognitive behavior therapy and auxiliary training to help patients. Data suggests Psybot is capable of replacing 70 percent of the communication normally conducted by human mental health consultants.

Facial Expression Analysis Software EmotientAPI
An application capable of analyzing user facial expressions and sentiment index to measure mood (positive, negative or neutral), range of emotions (happiness, surprise, sadness, fear, disgust, contempt, anger) and additional variables such as frustration or confusion.

Wangli Tech AI Addiction Assessment System
Implemented in Chinese drug rehabilitation centers, this system collects objective physiological indicators such as EEG, galvanic skin response and heart rate, and uses algorithms to evaluate drug addiction severity. Accuracy on some tasks reaches 90 percent.

NYU Post-Traumatic Stress Disorder (PTSD) Recognition Algorithm
Diagnoses PTSD by analyzing the voices of veterans, reaching 89 percent accuracy. Researchers use third-party voice software to analyze recordings, capturing more than 40,000 speech features, including mumbled words and tones of voice associated with depression.

The above AI academic/industry analytics are from Synced’s “AI-powered Application Report Series,” focusing on smart industries and application scenarios.

---

Localization: Meghan Han | Editor: Michael Sarazen