Utilizing the Attention Economy to Drive Healthier Choices

Hi everyone, it's Richard Zwicky with High and Healthy. The show where we know there's no one path to healthy so we explore them all. You know, one aspect of mental health and how we approach treatment of it is the psychology of persuasion. And can we use attention and the attention economy for good? Because everybody's people are used to behavioral targeting and behavioral systems interjecting into our lives to present this information at the time the advertiser or the company thinks is the most opportune. But can the same psychological mechanisms that keep us scrolling and keeping us looking at things also get us to take our medication, exercise, get a cancer screening. And when does this behavior become manipulation? What we're joining us today is Amy Butcher, who's the chief behavioral officer at Lyrio and is going to discuss a lot of these issues. Welcome for the show Amy. Thanks Richard. I'm excited to be here with you. So you know, we just mentioned off air that Lyrio has about 80 employees. And I think I noted started in 2023. How was the product shifted between when it started and where you are now? Because technology shifted a lot and the opportunities have shifted. And were you targeting the same thing? No pun intended? Or has that thing shifted as that markets evolved? Yeah, that's a great question. We're slightly older than that. I actually joined in 2021 because we really launched in 2019. And it was a strange time to be getting into what we do because we nudge people to take action around their health. And our first use cases were really situated in the clinical system. So one of our first nudge products was getting women who were overdue for mammograms to don't have that screening. Well, we all know what happened in 2020. And so people weren't going into clinics to get mammograms at that point in time. And it gave the company an opportunity to, you know, sort of sit back and rethink some of the approach and the technology. You know, and I joined the company during that time, just as we were starting to resume some of those clinical activities. So I joined at this time where some of the behavioral science and content had been refined while we were, you know, quote unquote, off the air, awaiting out the pandemic. And we were really testing out, you know, can we get people to come in? We also had launched a COVID-19 vaccination intervention around that time. Right when that vaccine was rolling out in the United States in a really challenging environment, actually, because our client for that effort was in the South and a Catholic health system, you know, in a population that was maybe less predisposed to be excited about the vaccine. So there's lots of information and disinformation floating around that create a confusion on the market. Absolutely. And you know, just politically speaking, because that particular, you know, disease and vaccination have been so politicized like that's part of the environment in the United States that tended to be a little bit more politically against. But, you know, that's the environment that I joined in. And we were very much focused on these sorts of singular behaviors, you know, coming in for one appointment or, you know, two vaccines that were six weeks apart. One of the biggest ways that we've changed and I think it's, it's both a function of the fact that we've had success as a company that has allowed us to get in more deeply with client organizations and sort of own more of the patient journey. But also evolving of the technology is that now we have much more complex use cases that are multiple behaviors over time. So we might be nudging a patient around all of their preventive care, inclusive of vaccinations and cancer screenings and wellness visits. And that means that it's not just about getting the person in for that one visit, but understanding how do these things fall together in a pattern, you know, what where the place is where there might be universal barriers or, you know, facilitators for health behaviors. You might see something like a person has really great health insurance and that makes it easier for them to get all of their health care or the converse. But then there's also behavior specific issues. And so, you know, our technology as we interact with people is capable of learning about what might cause a person to act or not act around specific behaviors. And what are some of those more global factors that we want to bring into every conversation. So we've definitely seen a lot of growth and I'd say the evolution is really more towards that complexity. And so, you know, you spend your career designing systems that change behavior. When you look at a social app, what do you see at the rest of us don't? Oh, man. You know, one of the things that I notice a lot on social apps is the infinite scroll and what that does to your attention and to the sort of need to have willpower to stop it. I fall victim to this as much as anyone else. So even though I understand how it works. It's effective, even those of us who know it so fall for it. Absolutely. I mean, I will tell you, I canceled my Facebook account in part because I was falling susceptible to it. And I was seeing some of the effects that it had on my mood and my attention. I do have TikTok. It's my guilty pleasure. And I on purpose only look at it if I'm willing to lose a little bit of time. But that that infinite nature, you know, I think as human beings, we are looking for markers in our environment about our progress. And you know, what does success look like and when are we done? And social apps are designed to never be done. So, you know, we're missing those sorts of guideposts, those milestones that we look for in other places. Like, you know, think about your education, you have your unit tests or your end of semester exam or, you know, you've mastered a certain topic and you know, you can move on to the next. There is no next when it comes to your Facebook or Instagram feed. It's it's just more and more and more. And then the other thing that is very interesting to me that I think is a little bit newer in its intensity. The amount that they add in accounts that you don't follow and/or advertisements and, you know, mixing that in with the people that you follow on purpose. And it's really interesting to me because, first of all, I think they do a very good job at understanding what you might be interested in. So, they're, you know, grabbing the themes from the things that you have deliberately decided to interact with and showing you more of the same. But I also find it interesting that they're they're doing a good job. I think with that Explore, let's exploit where they also interject novel topics, novel accounts and see if that's maybe an additional interest that you have. I think TikTok is especially good at that. I've actually been kind of shocked at how well it can present new sorts of topics to me that actually it does turn out I have an interest in. I got very into watching hoof cleaning videos. There is a, yeah, Nate the hoof guy. I am a city girl, like born and raised. And I'm watching this guy, you know, clean cow hooves. And I'm fascinated. But there's something about the patterns of interactions that I have had that told the system that I might be interested in this. And then, of course, I confirm it with my own behavior. So that that's something as well that the technology has gotten so much better. And now social apps are really taking advantage of it to make sure there's variety there that keeps our attention peaked over a longer period of time. So it's harder and harder to tear away. And which of those mechanisms do you use at your own work in Lyrio and which of you rolled out? Well, that Explore exploit bit is I think a key part of any successful AI system. So we send content to people around their healthcare. And we use behavioral science in that content trying to address the reasons where they may, you know, be skipping their cancer screening or their vaccination or, you know, their medication. And we might know things that tell us with a very high degree of certainty that this particular message is likely to resonate and be relevant for a person. But on purpose, we will explore other topics, you know, other reasons. And part of that is because in healthcare too, it's very rarely just, you know, one simple reason why I didn't do this thing. Like maybe I've been avoiding it because I'm afraid it will be expensive and it's going to hurt and it's going to be embarrassing. And I'm busy with other stuff. And addressing just one of those barriers isn't going to get me all the way there. So that exploration bit is definitely something that we do. And of course, then we learn whether or not a new technique is effective in addressing this person. And we can incorporate that into our large behavior model that we use to kind of understand that individual in context. We do not do anything along the lines of an infinite scroll. It's hard with healthcare. I do say sometimes that if we nudged everybody about everything they had to do around their health, you know, we'd be nudging them all day long, thousands and thousands of messages. And of course, we don't want to do that. We have, we use the analogy of an air traffic control. Like we try to prioritize the things that are most important, most likely to lead to positive outcomes for people. So we're actually trying to pay attention to not overstaying our welcome and people's attention. And, you know, we know that especially for something that's as unexciting to most people as healthcare, if we really try to take too much of their attention, they're going to unsubscribe, they're going to delete, you know, whatever app is giving them notifications and we don't want to find ourselves in that place. No, no, that's counterproductive. And, you know, the whole behavioral part, I mean, you wouldn't be aware, but my patents are actually the foundation of the system that most of the tech companies use in that area. And there's always been the question behind that, which is, is there anything inherently wrong for a healthcare platform to design around the reward? dopamine cycle. Yeah, and you know, I don't think there's anything inherently wrong with it, but I do think that it's incumbent upon the people who work for these companies. And especially people with training like mine, I have a PhD in psychology and I lead a team of other people with their, you know, doctorates in advanced degrees in the social sciences. Our training teaches us to respect human autonomy. And if we don't bring that value forward into our work, I think that's where the problem happens. And I have butted heads many times over the course of my career with people in other roles, typically in client organizations, because one of the principles that I try to uphold in my design work is called volitional non-adherence. So basically a person says, I don't want to do that. You have to respect that. And you know, you can continue to engage with them, you can present them with evidence or counter arguments or incentives. But at the end of the day, you can't and shouldn't force somebody to do something around their healthcare. They don't want to do even if it is to their own detriment. And we know from studying motivation that when we overstep that line and we get people into a forcing situation, then they are unlikely to develop any real interest in doing that behavior again. So when you think about healthcare, it's very rarely a one and done behavior. Even something that happens rarely, like a colonoscopy for many people is once a decade, it's still not one and done. And so we don't want to basically waste the chance to help a person form a meaningful reason why they might be interested in doing something. So, you know, for me, a really important part of this technology, this, you know, and using persuasive techniques is being respectful of those signals that a person is not interested, giving them an opportunity to exit the experience. And, you know, some of that is simple, like, legally required stuff, like allowing them to unsubscribe from a campaign. But, you know, even features like letting people set really, you know, detailed preferences for what you communicate to them about, letting them snooze something or step away from an experienced for a period of time. You know, we let people with our vaccination campaigns tell us, like, please don't message me again until XYZ date, like, those kinds of things that put some control in the hands of the person and respect their ability to say, no, are really important. And if we don't do that, that's where I start to feel very uncomfortable. So, where does that line cross when there's a treatment which is required for a person? Or, so let's go back to vaccines. In, if more than 3% of a population doesn't uphold the schedule with certain vaccine, it increases the chance of the mutation, which then reduces, well, increases the risk for everybody else, even if they've been vaccinated, because back diseases morph, they change, they evolve. There's an ethical question. And there's the personal question. Where do you draw the line and how do you decide how to draw that line? Yeah. Well, first of all, say I'm lucky that I don't have to make that decision, because it's a really, really tough question. As a person, I personally feel an ethical obligation to take action in that case. And, you know, I have always, as a person, been like the very adherent patient, I've always gotten my vaccines. And I think about those things specifically, like the herd immunity, and so forth. In terms of how we deal with it at a societal or policy level, I honestly don't know the right way to do it. I tend to, like, again, as an individual lean on that side of we should do this because it's the right thing to do for the community, but I also understand that when you create requirements and restrictions for people, you are getting yourself into this place of psychological reactants where, you know, people who feel forced into the action are not likely to repeat it. And I do wonder sometimes, you know, we see very low rates, at least in the United States, where most of my work is of COVID vaccination now that the sort of peak danger pandemic has passed. And I wonder sometimes they lag behind flu vaccination rates. Would we see more similar behavior in those two conditions if we had treated them from a policy perspective more similarly? So it's also interesting now that the data is out that proves the the RNA vaccines actually tie into helping reduce cancer rates. How that will shift people's perspective. But, you know, that also, the whole aspect of it and what we're talking about here is one thing we've seen since the dawn of the PCH is that people end up adapting themselves to systems and technology rather than systems adapting to people. Does personalization reverse that or does it make the adaptation visible? Yeah, that's that's also very interesting. I think good personalization should adapt to the person. But realistically, there are so many restrictions in how personalization takes place across technology that it still is very shaped by what exists in the system today. And, you know, just use the example of Lyrio. Our precision nudging is, I'll say form factor agnostic. I use that phrase a lot when I'm talking to prospects like we can integrate with, you know, your EMR, your CRM. We can do email or text message or, you know, voice. But what we end up doing is whatever our clients are already doing to communicate with their patients or their consumers. It's like they're sort of restricted by the technology choices that they bring to bear already in the situation. And we very rarely work with an organization that says, you know, let's take a close look at what our audience is actually using or like what makes the most sense for this type of communication and kind of start from scratch and build it in the way that, you know, is best personalized to that behavior in this person's life. So we do find ourselves really restricted by the technology that's already in place. One of the things that's also sort of interesting to me, we do a lot of text message, which feels very unsexy in a lot of ways. But if you look at what people are actually doing day-to-day and, you know, how they might react if you are messaging them to, you know, hey, go to the pharmacy for this. It's not always vaccineing. Sometimes it's a medication conversation or a lot of pharmacies will have clinics where they can do some basic primary care things. We find that people are very positively reactive to text message nudges because it's something where if you're out and about and, you know, you get it on a Friday afternoon, like, heck yeah, I can swing by the pharmacy, you get my flu shot on my way home from work. It all sorts really well for anybody with ADHD because it's an interruption that causes, right? Yes, yes. As I know myself. And funny enough, I am a very, like, list checklist-oriented person. So I love emails because I use my inbox like a to-do list. So if you send me an email about something to do, it stays in that inbox until it's done and it itches my brain unless I check it off. But that's another area, too, where personalization, like, different form factors for different people. But, you know, that's a technology that I think when people are talking about personalization and tech, they're thinking about, you know, hey, let's do virtual reality. And sometimes it's as simple as a text message. That's really what a lot of people still use to communicate and manage their days. So, let's say a patient's put on a treatment protocol or they've subscribed for something. And their own goals differ from the health systems. Who's objective and function wins? Yeah. Well, I maybe naively believe that part of the role if someone like me is to find the overlap in the Venn diagram of the patient's goals and what we'll call the system's goals. Because I do believe that there should be common ground. And if there isn't, we have a real problem. But at the very least, most people want to feel good. Patients do not be sick. They want to be able to live their lives, enjoy the things that they care about. And even if the system or the provider is framing their goals in terms of, you know, I need you to be adherent to this treatment or show up for this appointment. Ultimately, those things should be in the service of removing illness for storing function. And so, I believe there's a role for behavioral scientists to, you know, help retell the story to each party so that common ground becomes more apparent. And it's not, you know, it's not just a wordplay game, although that is part of it. Because another piece of it is, especially on the patient side, because so many times the things that they're asked to do are bringing short-term discomfort or cost. And I mean, costs there both literally and figuratively. Helping them to really understand how those things support their goal. And I think we missed the mark on that a lot. We just tell people you should do this and we don't tell them why, or show them progress, or teach them skills where they may even be able to monitor for themselves that, you know, they're experiencing some improvement. Are you feeling more energetic? Are you sleeping better? You know, is your mood improved at all? We don't give people the tools to watch these sorts of soft signals that there may be positive progress around treatment. And that is not equipping them to feel like there's a goal alignment. Right. And we like to assume that people act when they know what they should do. That it's the best thing for them. But behavioral science tells us otherwise. Yep. Should people have the right to know when an algorithm is deliberately trying to change the behavior? I think yes, they should. You know, in general, I am in favor of more disclosure rather than less. And we actually also know from behavioral science that people's reactions, including their emotional reactions, don't necessarily change because they know it's an algorithm and not like a life human being or a generic system. There's really great research actually around loneliness and social connection that found that even when people knew that characters in a video game were totally automated, they could still feel that like almost depression like sense of rejection if the characters wouldn't pass the ball to them, right? So, you know, I don't think that we lose the effectiveness of what we're doing by disclosing to people and, you know, in general, to go back to my whole idea of non-volutional non-adherence and supporting people's autonomy, like that's part of supporting people's autonomy, you don't want to trick them. So, yeah, I think they do have the right to know that. And again, I think they have the right to opt out of it as well if that's something that they feel strongly about. And do you design the system for somebody who's avoiding a behavior or a fear versus some people who simply forget? And how do you change the mechanics if you do? Yeah, the content of the system is different based on why somebody's avoiding the behavior but the mechanics not necessarily. So, we use a multi-agent architecture where there's different AI agents responsible for different decisions. My team builds libraries of content that correspond to those different reasons why people may not be taking actions. So, we would have, you know, messaging that helps people think about their fear differently, maybe way pros and cons or relative risk, you know, hopefully help them come to a place where they're not feeling that fear anymore. As well as ones that are more geared around forgetfulness, there's techniques like action planning, for example, that can help people keep something front of mind on a to-do list. That content would exist separately in our system and there would be an agent that's responsible for choosing whichever version of it seems to be most relevant for that person based on what we know about them. And there's other agents that are making decisions about, you know, is it a text message or an email or, you know, a prompt within an app, does this come on Tuesday or Friday? Does it come in the morning or the evening? And so those decisions can come together to create a more personal experience for somebody and what you might find is that if it's somebody who's forgetful and, you know, over time, we're noticing patterns that indicate that this is sort of forgetfulness. If you remind this person at a moment in time when they can take action, they tend to do it. You might find that they are getting, you know, quick text messages on that Friday afternoon when they can stop into the pharmacy on their way home from work, where the fear-based person might be getting, you know, emails on a Tuesday morning where they have time to really read through it. Read through it and, and ruminate and figure out, is this something I want to do? And we do find that people with more complex barriers tend to respond better to the longer sort of email form factor compared to a quick text message. So we know that there's some differences there. And, you know, clinicians have come on the show before have, I was spoken, you know, I'm a strong proponent. I have, you know, my work deals with it on a day-to-day basis. Some vaunt the power of AI to support their work, to be an adjunct and to be something that people should use between sessions. Other warn about AI becoming an attachment figure that constantly validates us and actually reinforces the bad behavior. One of the dangers is one of the lines to draw is where does the health app feel like a companion or and where should it feel like a tool? And how do we, how do we balance that line to make sure it's effective and people engage without feeling it's an invasion of privacy? Yeah, and I do just want to be clear I'm not a clinician. Oh, no, I can't speak to the clinical. I totally hear the question and of course I work in the healthcare space. So adjacent to it. We don't do a lot, we do something with generative AI, but for the most part our content is created by human beings. And we use AI at certain points in our process, you know, for quality checks and brainstorming and, you know, sometimes we'll see like what happens if we use AI to transform the same bit of content between, you know, a long email to a short text message to a WhatsApp message. My chart messages are actually constructed differently. So can be really useful tool to sort of do versioning and, you know, then have the human quality check come in. We've used conversational AI and sort of cut and dried use cases where we don't feel like there's a ton of danger around, you know, creating this false sense of alliance or, you know, leading a person down a wrong road. And so an example might be a frequently asked questions type of scenario where it's sort of pure informational content and we're using, you know, a rag model to restrict the AI to approve sources and the topic is not one where people are likely to make a really poor decision because of something the AI said that wasn't intended to say. We've been actually doing work with Vanderbilt University Medical Center under an ARPA age grant to figure out, you know, what is a system that we can put in place in the mental health space to create some of the guardrails because, you know, and even in the day that we generated as part of that project, my, like, my lord, the AI will say things that are really, you know, off base. Oh, it takes a lot of work to build the right guardrails and to maintain them. Yes, and so, you know, part of what we're working on there is a system that would allow, you know, people to evaluate is this an adequate system for others to use, you know, so if you imagine you are a clinician and you want to tell your patient to use a tool or not, this would be something that you could use to evaluate and say, yeah, this one has good guardrails in place, this one does not. But yeah, I think in general, right now it's really erring toward the side of caution until we get to a place, if we get to a place where it's better. But I mean, there's a lot of promise in AI as well, and so I feel more excited than afraid in general about it. It's how we, the framework we give the system for how they should interact with us is key and the tools we give the system to know how and when to interact with us is, you know, and that's a whole other discussion I can go on, we don't have enough time. But, you know, the attention economy is what it is, right, we, even if we approach it from the perspective of there's good ways to use that, there's some who take advantage of it just to sell more shoes. And that's less desirable and the systems can go in that way. But one of the things that we need to look at is success doesn't just have to be measured by what happens in the app, it's what happens in person's life after they leave it. If you were trying to do that, how would you do it? No, I love that you said that I completely agree with it. And early in my career, I worked for a startup called health media that had spun out of the University of Michigan School of Public Health and founder, Dr. Rick Stracker used to always say, basically, you know, if if we get you to a place where you don't need our product anymore, we had digital health coaching, so not even an app, because this was so long ago. We're successful, right? Like you, we're not measuring how many minutes you spend on screen. We want to measure, did you go out in the world and successfully manage your health because we gave you the right tools and empowerment to do that. And that, you know, philosophy, I think has really shaped the way that I think about it as well. One of the things that we do at Lerio that works well, you know, we call ourselves a drug-driven AI, and when we are first implementing with a client, we're sitting down with them and saying, you know, what do you really want to achieve here? What does success look like in terms of the things that we could nudge people to? And an example I'm really proud of, and this just they just published an article on this in the joint commission journal that we work with cone health in North Carolina, we built a nudging program for them for people with high risk hypertension. And we were really serving as like the safety net because they went through an incredible organizational effort to revamp all of their hypertension care and make sure that their black patients were getting to goal as frequently as patients of other races, which in the past wasn't always true. And Lerio sort of came in and nudged the people who hadn't responded to clinician outreach or my chart messages and, you know, followed up with with emails and were able to get a lot of them to come in and complete their clinical visits, but what we were really measuring, what was important was what happened to their blood pressure. That's outside of the system, that doesn't happen because they went in the doctor's office, that happens because of all the other things that happened, you know, after they left that doctor's office and, you know, changed their lifestyle took their medication. So we, because we have access to that data in our technology connection to cone. Our algorithms are basically looking at, you know, as I'm nudging Amy to do things, is this actually making a difference in her clinical measurements when she comes in and if not, you know, it's going to make different decisions. The trade-off is really based on that clinical measurement, not on, did she do it, did she click on the message, those, those are weighted at a much, much lower reward function for the algorithm. And I think that kind of setup prevents us from being like a Facebook newsfeed because we could absolutely reward the algorithms on, you know, Amy clicks on every message seven times. And then spends a long time on whatever webpage it leads her to, but we don't really care about that so much. We don't even care, honestly, if you click the message or not, there's a phone number in there, pick up the phone and call, that's fine. We don't need the digital engagement signal. What we need to know is, are you getting healthier? And that really is the most important thing. And what the technologies all should be should focus on and most do is, is the person getting healthier. Amy, we are out of time for today. This has been a great discussion. Love to have me on to them one of these days. People want to learn more about lyrio, where can they visit? And more about you? Yeah, sure, sure. You can definitely visit lyrio.com, our website. And we have, you know, quite a bit of our research and publications on there. Also on LinkedIn. And you can find me in both of those places as well. Wonderful, thanks for joining us down high and healthy Amy. Thanks for having me. And thanks to everybody for listening. I'm Richeswick with High and Healthy and we back again with you soon. [BLANK_AUDIO]