Dr. Alison Darcy is a clinical research psychologist, and the founder and President of Woebot Health, a mental health chatbot supported by clinical research trials. Prior to Woebot, Ali was an academic faculty member for almost ten years at leading universities, researching and designing treatments for people living with eating disorders. She holds a PhD, MLitt and BA in Psychology from University College Dublin, and has spent her career building scalable technology solutions to address the pervasive mental health crisis. You can find her on LinkedIn here and on Twitter here.
While working for a government-funded charity in Ireland called Bodywhys, Alison saw firsthand how resource-intensive it was to provide in-person support groups for people due to logistics, stigmas and demand, which was outpacing available services at the time. Recognizing a tremendous need for better preventative services, Alison set out to translate in-person mental health care into technology models.
While earning her PhD at Stanford, Alison worked on treatment development, but couldn’t shake the idea that the work didn’t matter if most people didn’t have access to treatment. To address this challenge, she turned to computational psychiatry, cycling through various technologies that brought new perspectives and capacities to the table. However, it wasn’t until she prototyped an AI-powered chatbot that she understood the potential of big data to solve a fundamental human problem and bridge gaps in the mental health care system.
To this point, emotional and intellectual accessibility have been fairly absent from psychotherapeutic technology design, but are just as important as location and timing. According to Alison’s research, Cognitive Behavioral Therapy (CBT) – an approach which aims to change thinking patterns and, in turn, shift behavior – lends itself well to digital translation, but is best delivered with a guide. Alison also knew that people like to share their experiences in natural language when they’re upset, so she set out to create a conversational interface accessible to all people.
Thus, Woebot was born—an app that invites people to check in with themselves once a day about their mood. If a user is happy, Woebot uses the opportunity to teach them new practices. If a user is down or upset, Woebot offers guidance through evidence-based tools, addressing them in a natural way. Ultimately, Woebot’s conversational dialogue, which acknowledges feelings and responds appropriately, makes it easy for people to engage when they need support the most—a key challenge for most behavioral health technology which struggle to maintain engagement over a long period of time.
As the world faced a global crisis and shared emotional uncertainty, Alison and her team looked at what people shared with Woebot and recognized similar anxieties across the board. To meet the moment, they launched Perspectives on March 17th, 2020, offering interpersonal psychotherapy to help people process loss and role transition.
Alison felt it was important that Woebot didn’t lean into the negative circumstances. Instead, the chatbot provided perspective, encouraging quotes, grounding techniques and insight into the shared experience. Perspectives ultimately reminded people of what others have experienced and how they got through it to let app users know they aren’t alone and that there’s a light at the end of the tunnel. Ultimately, Woebot’s user base doubled over the course of the pandemic.
In May, Woebot Health published a study exploring how Woebot’s relationships with users compared to human-delivered care. The team administered a validated questionnaire, called the Working Alliance Inventory, which asked people to evaluate Woebot within 3 to 5 days of their first conversation. From there, they plotted the scores against those from human-delivered CBT modalities, including outpatient care, inpatient sessions and group-based CBT.
The study found that Woebot’s scores were indistinguishable from those of the human-delivered care. More striking is that Woebot achieved that bond within days—earlier than the other studies which all took about 2 to 6 weeks. Alison explains this is likely due to the fact that Woebot is with a user at all times, whereas a human therapist typically only sees a patient once a week. Even at eight weeks, the scores remained almost identical, indicating that the rapport held up over time.
According to Alison, applications and digital therapeutics that target an individual’s specific experiences and allow them to respond naturally will see the most success. She notes that we can see this in precision medicine already, but not in psychotherapy or psychiatry.
With Woebot, in particular, Alison is focused on improving its natural language processing capabilities, while delivering quality care. Her hope is that Woebot will keep some people out of the clinic who can make use of its tools earlier, and allow human therapists to focus on people who need in-person care the most. Ultimately, Alison’s innovation has provided services for hundreds of thousands of people. Today, Woebot exchanges approximately 4.7 million messages with people every week.
Jim Freeze Hi! And welcome. I’m Jim Freeze, and this is The ConversAItion, a podcast airing viewpoints on the impact of artificial intelligence on business and society.
On today’s episode, I’m joined by Alison Darcy, Founder and President of the mental health care app Woebot Health. Through a combination of AI and digital therapeutics, Alison and her team have built a relational agent, named Woebot, that engages with patients via chat to provide real-time support, adapting to their emotional responses and delivering the right intervention at the right time.
Woebot exchanges approximately 4.7 million messages with people every week – more than twice as many as it did a year ago – and is now used in more than 130 countries around the world. In our conversation, Alison unpacks the science and technology behind Woebot, and shares how the service is making mental health care radically accessible for those in need during the pandemic and beyond.
Alison, welcome to The ConversAItion!
Alison Darcy Thank you very much.
Jim Freeze We’re thrilled to have you here. So to start, we want to talk a little bit about your impressive career. Translating in-person therapeutic models to technology platforms, which has been grounded in both scientific research and technology development. Can you walk us through your background and what led you to found Woebot Health?
Alison Darcy I mean, when I look back there have been several key things that have given this perspective, I think, that led to Woebot in the end. But I did an undergraduate in psychology, but I also volunteered on some helplines at the time and did the Samaritans training. Samaritans are a really great service for people who are in crisis over here in Ireland.
But I also grew up in a home where my father had been through AA, and so there was a sponsorship dynamic if you like. He was very successful there, but both my mom and dad were always very much there for people in their great need, at no matter what time or day or night. And I just remember that stance that they had being a child. And I think I reflect on that now as being fairly formative in what it means to be really there for somebody in a moment of crisis.
But basically with one of the helplines that I worked for, a great charity, a charitable organization, a support organization that’s government-funded here in Ireland called Bodywhys. I was talking with one of my friends who co-founded us, and we were discussing their support group model and how it was really hard and resource-intensive to provide in-person support groups for people. And yet they were a really important service.
So we started toying around with the idea of maybe you could make them online. And that was the first service that I, quote unquote, digitized and went through the thinking there of how do you take something that’s helpful in-person and translate it to a digital setting? Both solving some of the problems of the in-person space, like having to be somewhere at a certain time, and stigma, all of those things. But also how could you leverage the technology itself to create a better experience in some ways?
And that led me down a path. And then I did a PhD in an inpatient psychiatric setting, and there I really learned we really need to have better services at the preventative early intervention side. I mean, that was the biggest take-home for me. Went over to Stanford, did a postdoc there. Those years were really about, oh, I just learned to be great scientist. I mean, that was such an incredible period in my scientific career.
And we were doing great treatment development work, and evaluating them and NIH-funded trials and what have you. But of course, I always could not shake this idea that it doesn’t matter how beautiful these treatments are if most people will never be able to access them. And so for me the problem of access was growing and growing. And I saw technology as being part of that solution. And mostly my work brought me towards computational psychiatry and I became really intrigued by what big data I could bring to the table.
And of course at that point we had cycled through many different technologies, the internet and then mobile and then big data. And all of those of course we’re bringing a new perspective and a new potential capacity to the table. But I think it wasn’t until I made a prototype of a Woebot that I realized this was the technology that could really solve what was a human problem fundamentally, not a technology problem.
So there are some of the key things I think that led me to decide to found the company effectively, once we had made Woebot and I could see how people were reacting to it really differently. And it became this very natural way to deliver or get some of these great tools that we have from the clinic into people’s hands in a meaningful way.
Jim Freeze Yeah, it’s interesting. I believe you founded Woebot in 2017, and at that time, leveraging artificial intelligence in mental healthcare was a fairly novel concept, in any kind of healthcare. And you’ve already touched on some of the concepts of it, the importance of big data and how you can apply that.
I’m curious as to at what opportunity at that time did you say, “Okay, now is the time to actually create Woebot.” What was it about it that said it’s time to leverage AI in this space?
Alison Darcy Well, again, I was trying to solve this problem of access and scale. And access really broadly defined, which goes beyond having to be at a certain place at a time … Not just the fact that you have technology with you the whole time, but access in terms of being emotionally accessible, intellectually accessible, accessible for all kinds of people. This has been really so far absent from both our clinics, and to some extent the design of technology itself as well.
And what I noticed with Woebot was that … Well, it turns out, we’ve known this for a long time, that when you’re upset it really helps to talk about it, and to share what you’re experiencing in natural language. And so it was part of that, it was the interface.
And then the other piece, of course, is that when you’re engaging people in a much more naturalistic way, you’re actually engaging people. And that solves another major issue, which is with this technology and all of the behavioral health technology that I think had come before, was really challenged by this engagement problem. Only the most motivated people would, quote unquote, stick with the programs. And whereas something that was conversational, something that paid attention to how people were in any given moment, including very, very upset, and could respond appropriately. Make it simple for people to really engage in a therapeutic technique at that moment when they need it most.
Well, now you have a really different starting place. And one in which I think that could fulfill the promise of, as you said, data-driven innovation. And because I think the other big piece of this is how are we going to further the field itself? We have great psychological approaches and psychotherapeutic approaches, but we also are very early as a field. And if we’re going to progress the field, what we probably need to do is address individuals a little bit better, really understand for whom what works and under what circumstances. And that is the potential of platforms like Woebot.
Jim Freeze Can you explain how Woebot works today and who it’s designed for? And it sounds like, based on what you’ve just said too, it’s constantly evolving.
Alison Darcy Absolutely. By definition, it’s constantly evolving. So Woebot is, it’s just a conversation really, but Woebot is actually a character effectively, a archetypal, almost retro robot character, that invites people to check in with themselves once a day. So there’s a mood tracking element. And then also offers guidance through some of the best evidence-based tools that we have, and if people are feeling upset or a bit down. And if people are feeling well, then Woebot could use that opportunity to teach a construct from a core curriculum of psychoeducational constructs.
So this is really digitized cognitive behavioral therapy. That is the primary approach that Woebot takes, although there are elements from different approaches as well. But I think the experience from a regular person, I don’t want to say user, it’s such a funny phrase, but it’s just a conversation with an entity that gets to know you over time and checks in with you. And then just offers really good help if and when you need us.
Jim Freeze Yeah. You know what’s really interesting, is the language you’re using, some of it, I’ve just written some of it down, is so similar to … We use artificial intelligence to develop intelligent virtual assistants, which are used by large enterprises as the front door to their customer service operations.
So you talk about Woebot as a character, we talk with our customers about developing a persona that is consistent with their brand. You talk about conversation and the space we operate in, it’s conversational AI because it’s truly a conversation. And I love that you talk about engagement because we do the exact same thing. In our space, that legacy technology deflects customers.
Alison Darcy Right, right.
Jim Freeze Who wants to be deflected? Customers want to engage. So it’s so interesting to hear that same terminology applied in a mental health way.
Alison Darcy Look, absolutely. It’s about leaning into and addressing people as humans, because absolutely, I just don’t understand why we would not want to build technology to interface with us in a much more natural way, the way in which we’re used to.
And it turns out when you’re talking about problems, you don’t swipe and click through problems, you talk about them. And actually, when you’re really upset, you don’t actually have the cognitive capacity to engage in things that are very abstract or try and solve complex visual problems. Or frankly, try and figure out how to use your phone or … And so it’s just, this is just about tapping into the way we are as humans.
Jim Freeze Yeah, exactly. That’s exactly right, I love the language. So what’s interesting, and this is, I guess, isn’t surprising, Woebot actually engages with twice as many people as it did just a year ago. And I suspect that’s in part due to the rising need for mental healthcare during the pandemic.
To meet the moment you added coronavirus support, something called Perspectives, which leverages interpersonal psychotherapy to help people process loss and role transition. Can you share a little bit about how this service worked and how it helped address the unique stress of this past year?
Alison Darcy Absolutely. Yeah, my pleasure. I mean, we launched Perspectives on March 17th, 2020. And it was amidst … I mean, I remember that moment very clearly because while all of us were packing up our offices and preparing to work from home, it was very unusual obviously, but something that was shared globally. And almost the emotional temperature was something that everyone in the world, in every region, was going through at the same time.
And we started looking at the kinds of things that people were saying and sharing with Woebot. The anxieties were so similar it was striking. I mean, down to the trauma calls by the first time people saw lines outside the grocery market. The things that we took for granted were suddenly at risk, and there was a sense of this just impending fear and anxiety. I mean, which of course we understood, we were going through it at the same time.
So when we published Perspectives, the reaction was just so great, but I think it was some of the best content that our clinical writers have ever come through. Because of course it was a way for us to deal with our own anxiety as well. To be able to funnel those into something that was genuinely providing people with the ability to get grounded, it was just such a unique perspective for us.
But yeah, we decided we didn’t want lean into how awful things were. We thought that what people needed was a little bit of perspective and just to feel your feet on the ground a bit, and to really feel the shared aspect of this. So that’s what we leaned into. And there were incredible things there with Leonard Cohen quotes and people … I mean, I shared an analogy from my own parents who were children during World War II, who described seeing German bombers flying overhead. And having to black out their windows at night in case they were bombed accidentally because they were in Ireland, which was a neutral country during World War II.
And just when you think about that, wow, the things that people have gone through before, this is just another one of those. We will get through this, as dramatic and as awful as it is. That’s what people I think needed to hear, and I think the response was well-received.
Jim Freeze Yeah, I’m sure it was. So one area that you’ve done some research on that we find really interesting is gauging consumer comfort with artificial intelligence. We’ve talked a lot about that concept on many different episodes on this podcast.
You recently published a study, which found that Woebot is capable of forming bonds with users comparable to human delivered care. Can you share a little bit more about the findings and what they implied for the broader mental health space?
Alison Darcy Absolutely. What we did was we administered a validated questionnaire called the Working Alliance Inventory. And we asked people to fill out the measure on Woebot within the first three to five days after their first conversation with Woebot. And we plotted the scores against other scores from other published studies across human delivered therapeutic modalities. So things like outpatient care, inpatient sessions, group-based CPT. Mostly in the realm of therapists helping with anxiety and depression, and on all of them were cognitive behavioral therapy like Woebot.
And what we find was really interesting, we find three major things. The first was that Woebot was definitely scoring in the human range. So the scores that users gave Woebot were indistinguishable from those of the human delivered therapeutic modalities.
Jim Freeze Wow.
Alison Darcy But the second piece was really interesting, was that again, that Woebot was actually getting those kinds of scores. So achieving that bond within three to five days of an initial conversation, which was much earlier than the comparison studies, which all start within two to six weeks. Which makes sense, because you’ve only seen your therapist twice in two weeks. Whereas, of course, it just shows when you’re talking to a Woebot, when it’s with you all the time on your phone, you can develop that rapport much faster.
And then the third piece was just as that, when we added scores from another cohort, which was Woebot for substance use disorder, we found that at eight weeks the scores are almost identical. So it’s not something that is a parlor trick that erodes over time or anything. This is the rapport that’s established and it is kept over time.
Jim Freeze That’s amazing. That’s really amazing. So one final question, where do you see room for development in your industry? What do you think is possible with AI in the mental healthcare space in the next five or maybe to 10 years?
Alison Darcy Yeah, well, I think following on from the bond studies, I think what’s interesting is that we find it … I mean, we were not surprised necessarily that Woebot was able to achieve those bond scores, but I think what is clear from it is that there are very specific design decisions that lead to that level of trust and the sense of mutual respect and what have you. And I think that’s going to be a key piece of developing this technology further as we go in the next few years.
But I think again, the real potential is developing applications and digital therapeutics that are meaningfully targeted at what a person is going through and what they respond to naturally. So this of course is, we’re seeing this in precision medicine already, but it hasn’t quite hit psychotherapy or psychiatry in a larger sense yet. And I think that’s the real potential of this technology. That’s where this can go. I think Woebots can get a lot better at English. Natural language processing is developing and making leaps and bounds at the moment.
Jim Freeze It sure is.
Alison Darcy But of course, you have to couple that with delivering just really, really good care. And I think hopefully we can keep some people out of the clinic, like those people who can make use of tools like Woebot, the idea is, can you reach those people earlier, get them well earlier. And so that the human therapists remain for people who really can get value out of seeing a person and who really need to see a human therapist.
Jim Freeze That’s terrific. Thank you very much, it’s been fascinating. I really appreciate you taking the time to walk us through the research and the sophisticated technology behind Woebot. After such a tumultuous year it’s fantastic to see how AI can help people access mental health care from wherever, whenever.
Alison Darcy Thank you.
Jim Freeze Thank you very much for your role in this too. Thank you.
Alison Darcy Thank you very much.
Jim Freeze That’s all for this episode, and this season, of The ConversAItion. A big thank you to all of our fascinating guests, and to all the listeners tuning in. We hope you’ll join us again in Season 5.
This episode of The ConversAItion podcast was produced by Interactions, a Boston-area conversational AI company. I’m Jim Freeze, signing off, and we’ll see you next season.