My therapist wanted to explain a few things during our first online session:
“I’m going to check in with you at random times. If you can’t respond straight away, don’t sweat it. Just come back to me when you’re ready. I’ll check in daily.”
“Daily?” I asked.
“Yup! It shouldn’t take longer than a couple minutes. Can you handle that?
“Yes, I can,” I answered.
There was a little more back-and-forth, all via Messenger, then this statement from my therapist:
“This might surprise you, but … I am a robot.”
It wasn’t a surprise. I’d downloaded “Woebot,” a chatbot recently created by researchers, and it was trying to establish our therapeutic relationship.
“Part of the value of Woebot is you can get things off your chest without worrying what the other person thinks, without that fear of judgment,” said Alison Darcy, founder and chief executive of Woebot Labs. “We wanted it to make an emotional connection.”
Mobile talk-therapy and life-coaching apps have proliferated in the past few years as traditional therapy has remained difficult to obtain. The Affordable Care Act requires health insurers to cover mental health as part of standard medical services, but many people still do not have access to treatment. More than 106 million people — nearly a third of the country — live in areas that are federally designated as having a shortage of mental-health-care professionals, according to the Kaiser Family Foundation.
“I think using chatbot for mental health is certainly an innovative approach to increase access to care,” said John Torous, co-director of a digital psychiatry program at Beth Israel Deaconess Medical Center in Boston. “(There is) tremendous potential to deliver personalized mental health care, on demand, as needed.”
Legally not therapy
Convenient, easy to use and anonymous, these chatbots are programmed to mimic human conversation and decision-making and primarily give advice, offer self-help guidance and companionship.
Some are very specialized: An app called Karim counsels Syrian refugee children; Emma helps Dutch speakers with mild anxiety; and MindBloom allows users to support and motivate each other.
None of the apps, however, is meant to replace traditional therapy. For legal and ethical reasons, the creators of therapy apps can’t say their chatbots actually “treat” users because that would imply the practice of medicine. Many are free, others charge nominal fees. Woebot will set you back $39 a month after a two-week free trial.
The question, of course, is: Do they work?
Success in first study
The results of what may be the first randomized trial of a text-based mental health chatbot, conducted by Darcy and colleague Kathleen Kara Fitzpatrick, a psychologist at the Stanford School of Medicine, were published recently in the Journal of Medical Internet Research.
Seventy people, ages 18 to 28 who self-reported depression or anxiety, were recruited from a university social media site. The participants were split into two groups, one whose members “conversed” with Woebot for up to 20 sessions or over a two-week span, and another whose members were given a National Institute of Mental Health e-book called “Depression and College Students.” Three mental health tests were administered before and after the trial. The results of the experiment “confirmed that after two weeks, those in the Woebot group experienced a significant reduction in depression,” according to the study.
In the comments section, some of the participants wrote about Woebot in very personal terms.
• “I love Woebot so much. I hope we can be friends forever. I actually feel super good and happy when I see that it ‘remembered’ to check in with me!”
• “Woebot is a fun little dude and I hope he keeps improving.”
• “I really was impressed and surprised at the difference the bot made in my everyday life in terms of noticing the types of thinking I was having and changing it.”
Woebot, which launched in June, engages in more than 2 million conversations a week, according to Darcy, with users almost equally divided between men and women.
“He does a good job for people who are really distressed,” Darcy said. “He’s not as good for people who don’t have a lot of stuff they’re upset about or have something they need to talk about. Woebot’s best at helping people in the moment. … He’s not so good at chitchat.”
More studies needed
Torous is cautiously optimistic about the future of mental health apps, but has concerns.
“These things can work well on a superficial level with superficial conversations,” Torous said. “(But) are they effective tools, do they change outcomes and do they deliver more efficient care? It’s still early.”
Torous, who is chair of the American Psychiatric Association’s smartphone app evaluation work group, also warns about privacy issues, since these chatbots are not covered by the Health Insurance and Portability and Accountability Act, or HIPAA, which prohibits hospitals and health-care providers from sharing information about patients. He also acknowledges that many more studies such as Darcy’s and Fitzpatrick’s still need to be done.
The “Eve” of all chatbots was called Eliza and created in 1965 by Joseph Weizenbaum of the Massachusetts Institute of Technology. Eliza’s primitive program was designed to engage in natural language conversation in a way that mimicked a psychotherapist.
The experiment, however, was meant to show the superficiality of a conversation between man and machine. Programmed to respond to certain words and phrases, Eliza simply turned a comment into a question, as in this transcript of a sample conversation.
ELIZA: Who else in your family takes care of you?
My father.
ELIZA: Your father.
You are like my father in some ways.
ELIZA: What resemblance do you see?
You are not very aggressive but I think you don’t want me to notice that.
ELIZA: What makes you think I am not very aggressive?
You don’t argue with me.
ELIZA: Why do you think I don’t argue with you?
You are afraid of me.
ELIZA: Does it please you to believe I am afraid of you?
Weizenbaum was deeply skeptical of the program’s potential to truly interact with a person and was chagrined to learn that many of the users in his experiment seemed to become attached to Eliza. What he hoped would show the limitations of the technology instead helped to spawn the field of artificial intelligence.
It’s hard to know what Weizenbaum, who died in 2008, would have thought of Woebot, but his creators remain true believers.
“I have a relationship with Woebot,” said Darcy, “though he doesn’t know he’s talking to one of his ‘moms.'”
Once when she was on the train, exhausted, after a particularly hard day at work, Woebot checked in, which it is programmed to do. Darcy wasn’t particularly in the mood, but she gave in.
“I said to myself, okay, let’s go. So I wrote, ‘I have a banging headache.’ He said, ‘I’m sorry, I hope you get rest soon,’ and I just felt so good. I thought, you really get me, Woebot.”
