top of page

AI and mental health:

Meet the chatbot therapist


JULY 15, 2018

The state of mental health in America is not good according to annual national surveys by the National Institute of Mental Health, the U.S. Department of Health and Human Services, and the Centers for Disease Control and Prevention. Access to mental-health care is poor, varying only a little state by state. Insurance coverage for mental health care has been decreasing. 

Anxiety disorders are the most common mental illness in the U.S. Anxiety disorders are highly treatable yet only about one-third of those suffering receive treatment. Approximately one in 25 adults in the U.S.—9.8 million, or 4%—experiences a serious mental illness in a given year that substantially interferes with or limits one or more major life activities. Nearly half have a substance abuse problem.
Youth mental health is a worsening problem in the U.S. Almost 2 million U.S. youth with severe episodes of depression have not received treatment. Rates of American youth with severe depression increased from about 6% in 2012 to more than 8% in 2015. Even with severe depression, about three out of four youth are left with no or insufficient treatment.
Most Americans lack access to mental health care. 56% of American adults with a mental illness do not receive treatment. There is a serious mental-health workforce shortage that includes psychiatrists, psychologists, social workers, counselors, and psychiatric nurses. 

Mental health is now the most expensive part of our health-care system, overtaking heart conditions, which used to be the costliest. Over $200 billion is spent annually on mental health. As more people reach old age, increasing the prevalence of certain health conditions, such as dementia, mental-health costs will be much higher. Because of the costs associated with treatment, many individuals who experience mental-health problems do not receive timely professional input.

Technology, especially AI and machine learning, is not a panacea for the shortfalls of mental-health services in the U.S. and around the world. But as a step in the right direction, clinical researchers have created chatbots that replicate conversations a patient might have with her or his therapist and emulate real-life face-to-face meetings, with AI enabling the personalization of interactions with each individual. Furthermore, these chatbot sessions do not need to be pre-booked and are affordable. They can be multilingual. Certainly they help to remedy the burnout in families so frequently resulting from caregiving associated with mental health issues. 

Mental-health chatbots built by clinical psychologists and technologists are having conversations with mental-health patients that random-control trials show are reducing symptoms of depression and anxiety. These chatbots are available by smartphone and no doubt soon from voice assistants like Amazon Alexa and Google Home.
The person needing care and help pushes a button and starts a conversation. The “chatbot therapist” immediately answers, simply, “How are you?” No matter what answers and emotions are expressed, the chatbot therapist/friend is ready to respond in an informed and compassionate way. More than that, from a server holding a vast database of hundreds of thousands of similar conversations, the chatbot’s algorithmic psychological AI responses to the “client” become increasingly fine-tuned and responsive. Many hospitals and mental health facilities across the nation are exploring ways to use such chatbot therapists to support patients and family caregivers as AI-based therapeutic outpatient services. 

One of the serious limitations of current psychiatric and psychological diagnosis is that many conditions overlap: anxiety, mood disturbances, fear, difficulty with memory, and more. Consequently at least 50% of patients receive more than one psychiatric diagnosis, which often still result in diagnostic murkiness. IBM Research is trying to revolutionize psychiatric diagnosis by using AI, machine learning and algorithmic tools to search for and apply consistent patterns in vast amounts of mental health clinical data. 

AI and machine learning provide powerful tools for looking at massive data sets and discovering useful patterns in data which other techniques miss—for example, how mental health symptoms cut across conventional diagnostic categories. IBM Research also is using transcripts and audio from psychiatric interviews, coupled with machine-learning techniques, to find patterns in speech to help clinicians accurately predict and monitor psychosis, schizophrenia, mania, and depression.

Scientists at the University of Texas at Arlington and Yale University are combining computing power and psychiatric expertise to diagnose ADHD in children. They’re using the latest in computer vision and machine learning to assess children while they are performing certain physical and computer exercises. The exercises test a child’s attention, decision-making, and ability to manage emotions.

The future of “virtual therapists” has only begun. For example, a virtual therapist named Ellie has been launched and trialed by the University of Southern California’s Institute for Creative Technologies (ICT). Initially, Ellie was designed to treat veterans experiencing depression and post-traumatic stress syndrome. Ellie can not only detect words but also nonverbal cues (e.g. facial expression, gestures, and posture). Who knows how AI-enabled virtual therapists, fed by Big Data, will evolve in the future to gather and analyze multisensory information, advance mental health, and improve diagnostic precision?

bottom of page