Therapy Apps
train AI therapists, or otherwise fodder for automation. “It almost seemed as if there was a bot responding to me,” one former Talkspace user said of his therapist’s “boilerplate” responses (things like “Tell me more” and “How did that feel?”). But the problem with app therapists isn’t that they’re robotic; it’s that they have human limitations and needs.
Robot therapists do exist, at least sort of. At this stage, they’re more like interactive workbooks. Woebot, developed by a team at Stanford in 2017, is a chatbot that walks users through the basics of CBT. Woebot asks how you’re feeling, and you respond by selecting among emoji moods. Woebot asks you to write out your anxious or unhappy thoughts, then quizzes you as to which cognitive distortions (“Black-andwhite thinking”? “Mind-reading”?) your thinking may evince. “I don’t always understand everything you write,” Woebot says, “but sometimes the act of sharing is just as good.” The act of sharing, though, is only half of the responsive give-and-take that clients expect from therapy.
Stephen Schueller is a clinical psychologist and an associate professor of psychological sciences and informatics at UC Irvine, where he studies digital interventions in mental health. (He also runs a nonprofit site called the One Mind PsyberGuide, where he and his team evaluate mental-health apps based on scientific credibility, user experience, and transparency.) “They’re not good replacements for therapists yet,” said Schueller of today’s therapy chatbots. “Try to book an airplane ticket through an automated system, and now try to think about doing therapy— like, we’ve got a ways to go with these things.” His research focuses on improving access to care through technology because “we’re never going to train enough professionals to solve and tackle the mentalhealth challenges we have.” At the same time, he said, “technology alone is not going to solve this problem. We need to think about how we use technologies to make the work of people more efficient, wider-reaching—both geographically and culturally. But I think we see humans need to be involved for the most benefit.”
Across approaches to mentalhealth treatment, the relationship between patient and therapist is generally understood to be essential to therapy’s success. Even for more datadriven forms of therapy, when questions of personality would seem to matter less, the therapist and patient still need to be united in a belief that all those numbers are meaningful and that working together to change them will alter the way the patient feels. That shared belief rests on trust. Tech companies are, at the moment, having trouble in this regard, the companies behind therapy apps among them. A New York Times report last year raised concerns about Talkspace’s privacy practices and marketing; companies like BetterHelp and AbleTo have also faced questions about their use of data, sponsorship deals, and treatment of workers. And Talkspace has taken a disruptive stance toward obstacles like regulation. Last fall, CNBC reported that Talkspace had sent around a memo offering to indemnify therapists willing to practice in states where they lacked a license, as the company worked to meet soaring demand. (Licensing requirements were waived during the pandemic in some states, but the emergency orders are expected to eventually expire.) “They’re saying, ‘Well, if we don’t like these health-care rules, we’ll just violate them, and we’ll see what the consequences are,’” Torous said. “When are they going to be trying to rewrite rules that impact patient care directly?”
When Hillary Schieve made her proposal to bring Talkspace to Reno, she heard plenty of tech-wary concerns. A group of local therapists and psychiatrists organized the Reno Mental Health Consortium to propose local alternatives to the Talkspace deal. They were worried about start-up interlopers, about Talkspace’s rumored pay rates, and about the limitations of fully remote care. Kat Geiger, a therapist who runs a group practice called Thrive Wellness of Reno, said she was glad to have a mayor who embraced mental health as a cause—but, she said, “when you’re not a therapist, you don’t know what you don’t know.”
While Schieve hopes to work with local therapists on their own mental-health initiatives, she was ultimately unconvinced by their proposals. She would like to do more eventually, but for now, Talkspace offers the relief of a ready-made solution: The challenge of recruiting enough therapists and promoting the program would be Talkspace’s to undertake. By late March, 1,357 Reno residents had signed up, including Ethan Clift. He had come to the service with low expectations (“I thought it was probably going to be one step above a suicide hotline”) and been pleasantly surprised. He appreciated the speed and convenience and liked the therapist he had matched with, though he had seen some people on Facebook complain that their therapists had ghosted them. “If we stop one person from committing suicide because we did this initiative, it was well worth it,” Schieve told me.
The psychologist and tech commentator Sherry Turkle has observed that technological solutions often start out being regarded as “better than nothing” only to supplant the alternatives and come to be treated as “better than anything.” Yet the pandemic has been a potent reminder of exactly what tech can and can’t replace. Nobody thinks a Zoom happy hour is a satisfying alternative to anything. As it stands, research suggests the apps that work best at providing therapy are those that most closely approximate the in-person experience. “Those are the ones that are least scalable,” Torous said. “Those are the most expensive. So those ones aren’t really going to increase access.”
Adrian Aguilera is a clinical psychologist and a professor at Berkeley, where he directs the Digital Health Equity and Access Lab. He says the issue with looking to tech start-ups to solve problems of access—to provide “Therapy for All,” as the Talkspace motto has it—is that app designers tend to solve problems for people like themselves, and they tend not to be the people with the greatest need. Indeed, the more elaborate the technology (if it requires a new smartphone, highspeed internet, a good camera), the more likely it is to exacerbate existing inequalities rather than remedy them.
In a study Aguilera and Schueller published in 2017, they described a grouptherapy program Aguilera had run at a public hospital in San Francisco offering CBT for depression and focusing on Spanish speakers. The group met for weekly sessions and, in between, received automated daily SMS text check-ins. The group who received daily texts stuck with therapy nine weeks longer than a control group who had received none.
Automated SMS text messages and group therapy are not the kind of technology likely to part a venture-capital firm from its cash. They lack the flashy consumer appeal of an on-call feelings concierge. Aguilera’s program didn’t promise comfort at the push of a button, nor did it use technology to distract from overburdened human care; it appeared unlikely to yield any Rube Goldberg workaround whereby inventing a new healthcare business somehow solves the problems of health-care businesses as they exist now. The texts—like much technology in the past year—were a reminder of something forged in the real world. They were a way of looking forward to the time when it would be possible to meet again.