AI Chatbots in Healthcare: UX Research

Designing conversational UI is a challenging task. I say this from my experience of designing a conversational AI chatbot in healthcare. From the moment I began working on it, I knew it wouldn’t be an easy feat. Questions like what kind of visual elements would I use, how can I reduce the user’s cognitive load in effort-intensive activities, were always on my mind. Prior to this, I had worked on UI design of many web and mobile apps. But none of them was as challenging as this one.How A Chatbot Can Help Your Healthcare Business | by Michelle Parayil | Chatbots LifeThe most challenging part for me was designing to handle the human-machine interaction. Each user is different. Unlike in websites/apps, where users can simply browse and leave, the chatbot users open the chat window to interact. They come with all sorts of questions- vague /smart /genuine /rogue /irrelevant and (sometimes) absurd. When they type a query, they expect the conversational UI to adapt to their needs–digest questions and construct intelligent answers/follow up questions.

The uncertainty of the usage makes the design process complex. Unlike websites/applications, there are no specific UI design principles for designing conversational interfaces. It might appear as a small thing but the limited knowledge on the UI design patterns for healthcare chatbots ultimately affects the customer experience.

So, what can a designer do to make sure that the conversational AI solution caters to most, if not every, user persona? I can share a few suggestions. Since I worked on a healthcare chatbot, most of my suggestions would be about best design practices for healthcare conversational user interface.

Chatbots are now making their way into healthcare solutions like patient engagement solutions, handling emergency situations or first aid, medication management, and so on. To create a great user experience, it’s crucial to pay attention to the process of designing the chatbot. One of the ways to do this is streamlining the process of UI design through UX research.

When I started working on UX research, I borrowed a few tried-and-tested methods applied in web and mobile apps. However, healthcare is a complex domain where data security is a major concern. Therefore, I modified a few of them to suit my needs.

Let’s talk about a few UX research methods which are important to conduct before deep-diving into UI design of a healthcare chatbot.

AI in healthcare | Artificial Intelligence in the healthcare industry

Discover users’ pain points

To discover users’ pain points, you must first know who your users are. So, the first step to any good UX research is defining your user persona. Your user persona should include demographics profiles, health profiles and task profiles.

The persona diagram must include needs, difficulties, frustrations, motivations, aspirations of your end users. For instance, a 56-year old woman’s persona should include pain points like– “I’m an old woman, I don’t have the patience to repeat the same thing over and over again” or “I am ageing towards dyslexia, I forget conversations I had 15 minutes ago.”

After you’ve defined your user persona, the next step is to figure out how they will interact with the chatbot. For that you can invest in any of the below attitudinal approaches-

  • Gather inputs by rolling out surveys to your target audience and asking them related questions.
  • Conduct interviews and listen to what users say.

Depending on the kind of healthcare chatbot, your choice of the method will change.

For example- if you’re doing UX research for a chatbot that guides people towards better mental health, then you can do anonymous email surveys as people don’t openly talk about issues like depression and anxiety.

How Chatbots Can Help Your Healthcare Industry? | BCC Healthcare

Observe users in their natural environment

Direct observation (a primary research method) is the key to understanding user needs and preferences for a new product. During observations, record verbal comments and the time spent on various tasks.

For example, if you’re designing a chatbot for surgeons, then observe them when they plan the pre and post surgical procedure. If the patient has a lot of complications, what dietary guidance do they offer? How do they perform the surgery? What instruments they use and how its usage differs from surgery to surgery?

By conducting interviews with people while they perform tasks, you can collect valuable inputs from them and use it to recreate a similar experience in a conversational chatbot.

You can also create interactive mock-ups to show artificial interactions. This gives sufficient scope to test the chatbot with users without requiring the engineers to actually build it. This helps in quick iterations with the users after validating the user experience and understanding their perception of the chatbot.

Research through complementary data gathering techniques

A quick and rewarding UX research method is secondary research. This is done by doing competitor analysis to see how others are solving the same problem. This approach isn’t useful if you’re developing something unique. But, if it’s something you are trying to improve, this method is easy and gives quick results.

You can experience real conversations, access failure threads and use data to improve your chatbot’s experience. Secondary research also includes searching through customer service logs, FAQs, online reviews or comments on blogs.

For example- if you’re building a patient registration chatbot, look at other chatbots available in the market. What is the tone and personality of chatbot? How much time does it take to book an appointment on the chatbot? What are the demographics that the chatbot covers?

Study the chatbot as a user and find out things that they are doing right or things where the experience could be improved. You will find some unsolved problems that could become a major feature in your chatbot.

Get help from stakeholders

There may be situations where you wouldn’t get time to do surveys and interviews with end users. In such cases, ask for help from people who want to build this chatbot. Get to the behind-the-scenes of the problem. Go to their sales and marketing calls. Understand the ‘why’ behind building a chatbot and what problems they are trying to solve.

Talk to the sales team to understand what kind of customer support calls/tickets they receive frequently. Interview their marketing team to understand what vision they have for their end users. What motivates their users? What do they need in order to be happy? What’s their idea of a good chatbot?

Gather real quotes/statements from the end users and use them to draw an empathy map and user journeys. This will help you identify major pitfalls and crucial moments.

The UX research phase is a crucial part of developing a great customer experience. When it’s given its fair share of time and effort, UX research helps discover important insights and reduces the number of iterations required to build the chatbot.

However, there’s no surety that your research findings will translate into a flawless user experience. All the research findings must be implemented and tested in real-time for you to discover if your research was right or not. Sometimes, a wrong method of research or the timing of the research may give you erroneous information.

I hope you figure out the right UX research method for your chatbot. If you have worked on a healthcare chatbot, I would love to hear the UX research methods you used.

Share:

Share on facebook
Facebook
Share on twitter
Twitter
Share on pinterest
Pinterest
Share on linkedin
LinkedIn
<b><strong>Karan Makan</strong></b>

Karan Makan

Technology Engineer and Entrepreneur. Currently working with International Clients and helping them scale their products through different ventures. With over 8 years of experience and strong background in Internet Product Management, Growth & Business Strategy.

On Key

Related Posts

Hook Up on Tinder

Since dating can be stressful, there is the possibility of humor to try to reduce tensions. In a new study published in the Proceedings of

error: Content is protected !!