Your Next Therapist Might be a Robot: Artificial Intelligence Advances in Mental Healthcare


For many, the words “artificial intelligence” evoke a spectrum of ideas–everything from the adorable Wall-e and Eve to the scheming robots in the Matrix. While these science fiction, Hollywood portrayals of artificial intelligence have certainly captured the attention of audiences for years, it’s time to make room for a new side of artificial intelligence…one that saves lives.

With the advent of new technologies like Google’s Smart car, Amazon’s Alexa (an intelligent, interactive personal assistant), and Facebook’s facial recognition software, AI is becoming increasingly relevant in our daily lives. Based on the idea of harnessing computer systems to perform human tasks such as speech recognition, visual perception, and decision making, artificial intelligence has infiltrated the fields of medicine, linguistics, art, virtual reality, and more.[1] Particularly fascinating is its role in psychology and mental health care, where it offers a supplement to therapists and acts as a diagnosis tool for mental illness.

Born from Stanford psychologist Dr. Alison Darcy’s vision of creating smart, accessible, and scalable solutions in the mental healthcare space, Woebot is an interactive chatbot (available via Facebook Messenger) designed to aid those with anxiety or depression. The website introduces the technology as “your charming robot friend who is ready to listen, 24/7.”[2]

By asking daily questions about your mood, emotions, and energy (anyone feeling some Big Hero 6 Baymax vibes?), Woebot brings to the forefront the triggers in one’s life that may be sources of stress or sadness. Stanford researchers led a study confirming its effectiveness, finding that the program helped alleviate depression in patients ages 18-28 who used the program daily for two weeks, in comparison to control patients.[3] Since its launch in June, Woebot has received over a million messages a week from users in over 130 countries and has earned a significantly positive reaction, as users are particularly grateful that help is just a click away.
Picture1AnikaWhen asked about the inspiration behind the technology, the team responds, “there’s never been a more important time to address the mental health epidemic–social stigmas, high costs, and limited access to therapists often prevent people from getting the help they need.”[4] This interface caters to a large population of people who may feel too embarrassed to speak to another human about their internal struggles, or who simply cannot afford in-person therapy sessions—which can cost upward of $3000 a year.[5]

However, like with any internet-related activity, the issue of privacy arises. While the company cannot see the user’s profile, Woebot’s Privacy Policy states that Facebook has access to the content of the conversations and this online data can never be 100% guarded from theft.[2] The team also emphasizes that “Woebot isn’t designed to replace human therapists—the technology is meant to supplement existing mental health care resources.” It will not detect onset of a mental condition or play the role of anything more than “a friend checking in.”[6] In this sense, Woebot is not entirely beneficial to people who are in serious need of talk therapy, specific prescriptions, or a mental disorder diagnosis. However, when Woebot detects the user may be in real trouble—for example, contemplating hurting themselves—it presents hotline resources and real-world solutions.

As for Woebot’s future, the interdisciplinary company’s clinical, artificial intelligence, and core engineering teams hope to build an independent app and use new language processing to facilitate deeper conversations between the user and the robot.

Artificial intelligence is evidently valuable for disease intervention, but it is also making its mark on disease prevention as a diagnostic tool for schizophrenia. Schizophrenia is a chronic disorder characterized by difficulties distinguishing what is real versus imaginary, hallucinations, sometimes violent tendencies, and a withdrawn attitude.[7] One key symptom is trouble organizing thoughts into coherent sentences and “subtle disorganization in speech.” A group of researchers at  Columbia University, NY State Psychiatric Institute, and the IBM T. J. Watson Research Center honed in on this idea to develop an automated speech-analysis program that could predict psychosis.[8]

The algorithms built for this model simulate how the human brain interprets language, and were utilized to dissect spoken interviews from 34 at-risk children. Using a well-established technique known as Latent Semantic Analysis, the algorithms measured coherence, a language trait often lacking in those with schizophrenia.[9] Coherence refers to whether each word in a sentence is correctly related to the other words in the sentence in terms of meaning. For example, there is a  famous quote from a schizophrenic who identifies his family as, “mother, father, son, and Holy Ghost.”[10] The individual words are intelligible, but the phrase as a whole is incoherent.

Another sign of psychosis is known in the psychology realm as “poverty of speech.” Essentially, the subjects’ speech is very minimal and they only reply with 1-2 words when prompted.[11] The automated speech-analysis study found that these extremely short responses, coupled with a “decrease in the flow of meaning from one spoken phrase to the next,” were the markers that classified which of the 34 children would later develop schizophrenia. The computer program accurately identified 5 such individuals, outperforming assessments from actual psychologists.[9]

Artificial intelligence is establishing a strong presence in the world of mental healthcare, and its progress does not appear to be slowing down. The rise of machines capable of fulfilling human tasks will inevitably affect the job market, and many fear that this will limit career options. Self-driving cars could make taxis obsolete, chatbots like Woebot may be preferred over real therapists, and IBM’s Watson might destroy everyone at Jeopardy. However, the world is constantly changing and with each new invention comes adaptation: new fields will emerge, bringing along new jobs as well. By maximizing efficiency and decreasing costs of tasks like checking in on one’s mental and emotional wellbeing, having higher accuracy rates in disease detection, and transcending its application to diverse fields, artificial intelligence is ultimately a step in the right direction.

So next time you watch “Terminator” and freak out because Arnold Schwarzenegger almost started a nuclear holocaust, you can be reassured that not all robots are bad. Some just want to provide a platform for your thoughts and help you be your best self.

Works Cited

  1. English Oxford Living Dictionaries. (n.d.). Retrieved November 3, 2017, from
  2. Woebot. (2017). Retrieved November 10, 2017, from
  3. Fitzpatrick, K. K. (2017). Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial. Retrieved November 11, 2017, from
  4. The Woebot Team (2017, November 13). [Email interview]
  5. Nina Stoller-Lindsey. (2017). How Much Does Therapy Cost, and How Do You Pay For It? Retrieved November 5, 2017, from
  6. Mack, H. (2017). Woebot Labs debuts fully AI mental health chatbot via Facebook Messenger. Retrieved November 9, 2017, from
  7. Schizophrenia. (2017). Retrieved November 8, 2017, from
  8. Lafrance, A. (2015, August). Computers Can Predict Schizophrenia Based on How a Person Talks. The Atlantic. Retrieved from
  9. Gillinder Bedi, Facundo Carrillo, G. A. C. (2015). Automated analysis of free speech predicts psychosis onset in high-risk youths. Retrieved November 9, 2017, from
  10. Kuperberg, G. R. (2010). Language in schizophrenia Part 1: an Introduction. Retrieved November 9, 2017, from
  11. Poverty of Speech. (2015). Retrieved November 9, 2017, from

Cover Photo: “Exercise Plays Vital Role Maintaining Brain Health” by A Health Blog is licensed under CC BY-SA 2.0

Figure 1: Courtesy of Woebot/ photographer (no link)