The Future of Healthcare and Conversational UI
A summary of thoughts on the intersection of the future, healthcare, and conversational UIs.
As technology is integrated with our healthcare, the products we offer often ignore the fundamental truth that healthcare is personal. Many tools are insensitive to the vagaries of a disease, some product designs can’t get anything with true empathy past their regulatory committees, others just don’t focus on the user enough to make things personal. But a new design paradigm, the conversational UI, could improve how people use techonology to stay healthy.
The future, healthcare and conversational UIs are all enormous topics. This is a summary of thoughts on the impending intersection between them.
Health and Humanity | What it Means to Design in Healthcare
Humanity is about connections. Since the inception of the Web we’ve looked for two things from technology: information and connectivity. We want the digital world to lend us a hand when we need information; find us an answer without making us work too hard. Then we want to take what we’ve learned and share that with others. For this daisy chain of digital connections to work, the process, and the people in it, need to feel authentic. They need to feel human, even when it’s technology, not humans, they’re connecting with.
Healthcare, or the vision of many people who build tools in healthcare, addresses the same core needs, and adds compassion as an umbrella concept. Tools in healthcare must meet our basic technology expectations, informing and connecting, but they must also comfort us in our time of need. They must make things bearable. Hopeful. If health is personal, successful tools must have a personal, approachable tone, must exude empathy and encouragement, as they connect and educate.
Therefore, when you start building tools in health and wellness, whether they’re for health care professionals (HCPs) or patients, you have to think about the deeper implications of what you’re building. You need to consider how it connects people to people, and people to information, compassionately.
Take this patient example: as a rule of thumb, there’s a steep drop off on drug adherence (taking medication according to schedule) for patients in the first 90 days. There are a number of reasons for this, it’s hard to integrate something new into your daily routine, new side effects can be hard to manage, etc., but if you want to help a patient start (and continue) taking medication, you need to figure out what hooks to design into a system to encourage adherence-friendly behavior change. You need to figure out the patient’s wants, needs, and fears, and design to the right goals. You learn what patients value and try to connect to it, to educate about it, and to meet it with compassion.
The elements you’ll find in most healthcare tools, therefore, address
- Compassion
- Education
- Adherence, or some element of overall behavior change
- Reporting, so that the doctor, family or community can weigh in
The UX of a healthcare project, from the content strategy down to the interaction pattern paradigms, always takes these elements into account.
Things shift slightly when designing healthcare tools for doctors and other HCPs. Obviously it’s not so much about what a doctor fears, but instead about how to
- Tap into her empathy to create the right tone so she can show that compassion
- Give her tools to properly educate a patient
- Connect her to data, to improve her monitoring of adherence
- Help her encourage any positive change in behaviors
It’s not exact parity with design for patients, but it’s close. While keeping all those factors in mind for HCPs, we also must maintain those at the right levels while designing for efficiency. Granted, those are a lot of high-level requirements to design for, particularly in a regulated industry, where the goal of multiple legal reviews is much more singular: to design against risk, rather than to design for empathy. But at its heart we always design around information, connections and compassion.
Every patient’s journey through illness and recovery (or even improvements in wellness) is unique and significant to them. The right tone makes all the difference.
The Experience and The Empathy
Lately the excitement in the user-experience-design community has been around the concept of conversational UIs. The conversational user interface paradigm is a text-based interaction where the system (we can call it a Bot, short for Robot, in all its futuristic goodness), asks the user questions and learns to predict his needs, leading him through a product. It’s interactivity via conversation, and it usually goes something like this:
BOT Hello! I’m your new Software. I can create reminders to help you take your prescription. Can I get you started with a reminder right now?
USR [Y] [N]
BOT Great. I’m starting to set that up right now. I just need to create an account for you. Feel free to type [Help] if you have any questions. What would you like your username to be?
USR talber
BOT Hello, talber! It looks like you’re in NYC. Should I assume we’re on East Coast Time?
USR [Y] [N]
BOT Cool. Noted. It looks like it’s almost 6 PM. Do you want to give me more information to set up your account and record your dose, or should I check back tomorrow morning?
Compare that to a traditional input form, an empty column of boxes waiting for someone to type. It’s markedly different. It’s warmer. It’s an onboarding interaction that seems to understand the patient. Of course, all good forms, as well as design in general, should be approached with empathy, but there’s still only so much humanity you can inject into the process of typing in boxes. In fact, many conversational UIs integrate pauses; humans take a second to type an answer, so it makes sense to built a typing indicator in for a bot. The conversational UI adds a semblance of humanity back into the exchange.
Speaking directly to the user, and parsing the user’s answers, removes the layers of abstraction we’ve created with interface design patterns. There is no learning curve for talking
The nice thing about conversational UI is that conversations are how we’re used to communicating. So although it may be difficult to build, we know what characteristics it should have. For example, when we communicate, we base our reactions on our level of understanding about the other person, and our history of interactions throughout the relationship. To mirror that successfully, conversational UIs need to learn about user and remember past interactions.
The best bots (or content strategies, for that matter), usually have a “helper” element, as if your best friend, or, even better, your mother, is helping you complete a task. Booking a flight? If Mom was booking a flight for you, she’d remember what you enjoyed about your last flight and try to repeat it. She’d know whether you were a morning person, and what time you left the last time you flew. She’d suggest flying times that didn’t conflict with your workouts. She’d think about what seat locations you’ve chosen in the past and make some assumptions. She’d give you enough of a layover to grab a snack. That’s the essence of the bot-conversational UI movement, and the primary reason for its success. The predictive, considerate helper we all wish we had that makes navigating any task more manageable. Like a good conversation with a friend.
A good conversational UI, therefore, also has the proper tone, connecting directly to the user just like a human. This is true across the digital world, it’s not confined to healthcare tools. In fact, it’s been barely mentioned in healthcare tools. But that’s about to change.
Simplicity
Beyond the seemingly human aspect of communicating in the right tone, predicting, and leading the user through a task, an additional reason conversational UIs will work so well in healthcare is their simplicity, at least in terms of what the user sees. Speaking directly to the user, and parsing the user’s answers, removes the layers of abstraction we’ve created with interface design patterns. There is no learning curve for talking.
Of course, this is also often slower, and more thoughtful; it only works in the right situations. But if you’re designing for emotion, for consideration, perhaps even looking to add a little thoughful friction back into the exchange to make the user feel like someone’s listening, the conversational UI is pretty great.
Beyond Empathy and Simplicity
There are other key features that make conversational UIs ideal for healthcare:
- Seeing where to pick the thread back up is easy. Visually conversational UIs look like a typed conversation, like a text thread. The linear nature of these conversations lets a user remember what he did last. It’s comfortable to come back to, like you’re picking up where you left off.
- We want to feel like bots are human. This presents an overall people-centered approach rather than an app-centered approach. Using the conversational UI approach becomes about a type of relationship, not an interface. It simply feels good to get bot alerts, because it feels like the communication we’ve been conditioned to have with friends and family.
- Tone bubbles up to the top. The right tone becomes especially important when your interface is all about a conversation. A properly designed bot has a personality of sorts, one with the right personal touch. Every patient’s journey through illness and recovery (or even improvements in wellness) is unique and significant to them. The right tone makes all the difference.
The Reality | Kicking the Tires Today
In most systems, there’s a mixture of conversational interface and standard interface elements (like buttons). Consider the Lark app. Lark is a virtual health coach that keeps track of your exercise using your iPhone’s built in tracking capabilities. It knows how many steps you took, or what time you put your phone down to go to sleep, and uses those details in its conversation with you about your health. Those background habits become the backdrop of conversations about what you’re eating, and when. They become the foundation for lifestyle tips, just like they would in the real world.
Since some of the questions Lark asks are easier answered with a button, it gives you a button from time to time. For example, to learn more about the exercise graph Lark shows you, or to indicate a new branch of discussion, like when asks whether or not you want to log food, it takes the button shortcut. That’s faster — no one wants to ruminate on the phrase “Chat about lunch.” But, at other points, it asks you to type in what you ate, and then kicks off a conversation around whether it was a good choice. That’s something worthy of consideration.
The new Quartz News app has a similar design— the use of a conversational interface with buttons for guidance. News stories are summarized in a short sentence, with a button to see more details or the option to move on to the next story. The “more details” option rarely reads “more details,” however. Instead it adds its own brand of emotional commentary by using emojis that react to the story. Tapping the emojis leads you to more information.
It’s biased, sure, just like any conversation you might have with a person. But it’s also engaging. This also adds visual variation, and gives the app more personality. Emojis stand in for some of the tone you’d have in a real conversation. It pushes you forward, into the content, sometimes for no other reason then to see how clever Quartz can be with its gifs and reactions.
Operator is another example of a GUI-aided chat interface. It’s designed for e-commerce, where someone can help you find the right gift based on what you tell the app. It’s friendly and simple, and your “operator,” which is usually a real person, starts giving you suggestions as soon as you tell it what you’re looking for. You can tap to order anything Operator posts in the feed. Granted, the initial setup is key, (for e-commerce, the success of a conversational UI is contingent on removing the hurdles, i. e. adding credit card information); once the system knows who you are, the conversational UI can seem pretty slick.
For example, let’s say you can’t figure out what to buy your brother for Christmas, so you open Operator. When asked how Operator can help, you tell it you want to get something for your brother, for Christmas, for about $40, and that he’s into Atari, Russian literature, and Beethoven (he’s a cool guy). Give it a second, and your Operator friend starts posting options to the feed. If something costs too much but you like it, you just say that, and the operature will adjust his recommendations. It’s so much different than sorting a list of results from Amazon, more nuanced and reactive. Although the facilitator behind Operator is a real person, it feels very similar to the other examples. They’re all trying to meet the complex needs of a user through conversation, just like in real life.
Luka combines the conversational UI feature with a community approach. Luka began as a bot-based app best known for getting you the right restaurant recommendations. You just tell it what you’re looking for, i.e. the best taco spot in the neighborhood that’s open right now, and it starts giving you options. You can ask Luka to make the reservation for you if you see something you like. Or get additional options. Then you can share it with other friends from inside Luka, as if the bot is just another member of your group. It really feels like you’re talking about where to eat with a friend; tell Luka what you don’t like about something he recommends and he’ll listen, and recommend something else.
The app is built to be shared, so you can send the restaurant location or photos to a friend. Lately Luka has expanded into all types of other things you can do with your bot: news, quizzes, it even has a button labelled “I’m bored,” where it’ll serve you up some entertainment. It uses a mix of interface elements and a lot of personality. The more you use it, the better it can predict what you’ll like, and it becomes like the friend who can help you take care of a task, as well as someone to kill time with while you’re waiting for the train.
Predicting what you’ll want is another important feature of bots and conversational UI, it’s fundamental to understanding the user and, as I mentioned earlier, building a relationship. Digit.co is a financial tool that helps you save money by taking a tiny amount you don’t need and pushing it into a savings account. It starts out conservatively, only pushing a few dollars, but learns how you spend your money and how much cash you need to have on hand. Then it manages to save larger amounts over time, so you magically have a savings account with a decent balance. Digit has almost no interface. It’s all texting back and forth with the bot. Once set up, it’s super simple and surprisingly rewarding.
Facebook Messenger has begun integrating services that use conversational UIs, as well. Poncho, a weather app, lets me know what the weather is every day at 4:30 PM. I can ask it for details, or I can find extended forecasts (using a combination or buttons and card interface elements). It lives next to my other contacts, and may eventually integrate with those conversations.
Finally, we’re starting to see conversational UI appear in healthcare apps. Your.MD is a healthcare personal assistant that parses the symptoms you give it, asking appropriate follow-up questions throughout the process. After landing on a probable condition, it serves up doctor-reviewed content. Your.MD will also allow you to read about diseases or call a doctor for additional information (in English, French or Dutch). The tone is approachable and compassionate (you can see below that it guesses my gender incorrectly, which real people do all the time, and it recovers gracefully); it’s a good example of conversational UI.
The Healthcare Hook
Each of these examples illustrates something we would want in a conversational UI for healthcare. It should have the predictive nature of Digit, the social capabilities of Luka, the ability to reach out into the real world and accomplish things like Operator, the personality of Quartz and Poncho, the medical information of Your.MD and the deep understanding of Lark.
Moreover, all of the examples above have the following three things in common. They make you feel like
- They want to be your friend (connection)
- They have your best interests at heart (compassion)
- They want to learn about you and give you information that is more helpful over time (information)
These are the same things we look for in someone helping a patient. Compassion and connection. Information. The tools listed above learn to express preferences and solve problems through conversation, which is the base way to envision a future tool.
The Future | Conversations, In Both Text and Voice
Conversational UIs are laying the groundwork for future healthcare tools that operate on an interfaceless plane of existence: the voice. There are many of ways the future will be voice controlled. A conversation. Already looking at Siri on the new Apple TV (just ask who’s in the movie playing now), or Amazon Echo (what’s the weather?), begins to show us what’s possible. It will be even better when it can predict what I’d like to know, and how I’d like to hear it.
Some version of voice control will arrive in the next decade. After all, the best digital implementations work like things in the real world, and one of the first things we learn to do as humans is talk. We know the framework may not really be there yet for our Internet of Things, but the concept, via conversational UIs, has definitely arrived. The Internet of Things that reminds you to take a pill when you sit down at the table for breakfast will be much less invasive if it feels like it’s coming from your old confidant, someone you connect with, who can give you the information you need to make smart decisions, perhaps even remind you about all the things in life you’d like to live for, even if it’s not a someone at all. That’s the future of healthcare and Conversational UI.
