PROBLEM STATEMENT
There is a lack of accessible, understandable, and interactive resources people can use to understand and manage their health concerns. Often people don’t fully comprehend what’s happening to them when they’re sick, which prevents them from taking control of their health.
In INFO-H 541-Interaction Design Practice, I decided to tackle this problem through my class UX project.
DURATION
One Semester
SOLUTION
An application that allows users to “chat” with an AI chatbot to assess their symptoms and receive a report with possible causes and recommended action. It includes an easy to understand medical dictionary that users can refer to better describe symptoms/illnesses.
TOOLS USED
Whiteboard
Miro
Xtensio
Figma
MY CONTRIBUTION
User Observations
Interviews
Affinity Mapping
Persona CreationUser Storyboarding
Prototyping
User Testing
PROCESS
RESEARCH
To understand the problem space of healthcare further we conducted observations and interviews. We aimed to understand the mindset of both patients and healthcare professionals. We also conducted research on technology and its impact on the healthcare process, specifically AI.
We spoke to 6 individuals in the healthcare field, ranging from medical researchers, bioinformatics students, nurses, and doctors to gain a better understanding of the space between patients and doctors, and where AI can help. The interviews lasted 30 - 45 minutes and took place over zoom.
A lot of patients come in having knowledge about their symptoms and wanting to fix them, but don't understand how to take care of themselves, or how to resolve the problem. We need to help them take control of that aspect.
- Nurse Practitioner From IU Hospital
KEY OBSERVATIONS
- Patients have anxiety about consulting doctors for smaller issues.
- There is a lack of reliable sources patients can utilize to be more aware of their symptoms and condition, and manage their health.
- There is a communication gap between the patient and doctor due to the patient’s lack of understanding of medical terminology.
- Although AI is reliable, patients and doctors felt that it has a long way to go in understanding distress or anxiety in patients.
PAIN POINTS
1. Lack of accessible and reliable resources for patients to assess their health.
2. Patient anxiety in asking questions and consulting doctor for smaller issues that become dangerous
3. Communication gap due to medical terminology and assets like x-rays, medical reports, etc, especially through telehealth.
DEFINE
AFFINITY MAPPING
Once we were satisfied with the information we received in the research, we decided to organize the information into an affinity map to understand the problem space better!
One of the things that stood out to us was the "personal contract" between the patient and doctor, which was initiated through nonverbal language, and would be hard for AI to simulate.
PERSONAS
Once we conceptualized the problem space using affinity mapping, we understood the target users we wanted to focus on.
Silvia: The graduate student who has no time, especially for her health.
Carol: Mother of 2 kids, busy with work but wants to prioritize good health.
IDEATE
BRAINSTORMING AND STORYBOARDING
Once we understood the key user requirements, we decided to brainstorm ideas to help solve the problem. Our mission was to help patients through treatment somehow.
IDEAS FROM BRAINSTORMING
Idea #1: Patient Notetaking application that helps patients take notes with the help of an AI dictionary.
Idea #2: AI fashioned Chatbot for easy access to resources that are customized per user's needs and history with the application.
Idea #3: Patient Reminders to take care of themselves: mental health checkpoints and medicine reminders.
Idea #4: AI Doggo that follows you around to check your health and reminds you of medicines and to stay hydrated.
STORYBOARDING
To explore the solutions, we used the personas to storyboard the ideas and see if they would fit our target users. We decided to storyboard the ideas above, and I was in charge of the AI chatbot, which I created here.
DESIGN DIRECTION
After ideation and scenario building, we decided to pursue the idea of creating a medically coded AI chatbot that would help users conduct symptoms assessments, and gain a better understanding of their health. We decided this since this would be one of the more accessible options and would reduce patient hesitation because "it's a chat". Some features I wanted to incorporate were:-
- An AI chatbot system that users can “talk to” to assess their symptoms and receive an informal report with possible causes and treatment.
- An easily accessible dictionary to understand symptoms and illnesses so they can be better described.
- Easy to access information that would empower a user to take control of their health.
PROTOTYPING
With the design direction, we decided to move forward with creating low-fidelity prototypes. Since I was more experienced with design, I took lead in creating low fidelity and high fidelity mockups, with help from my teammates.
With the aim to understand the conversational flow, we created a mock-up wireframe in Figma, with the use case scenario of a person experiencing cough and cold symptoms. We focused on wireframes and information placement.
DESIGN
When we spoke to the professor and 4 peers about the design, the main feedback that we received was to research thoroughly into the conversational UX and to create deeper conversations with the user.
To visualize the conversation, we created a user flow on the whiteboard.
The scenario we picked was a user who has symptoms of a cough and cold.
HI FIDELITY MOCKUPS
Getting Started
In the first phase of the application, the chatbot asks for the user's name and gives a brief introduction to the application.
Select Symptoms
From here, the user types the problem and Suzy detects the potential symptoms the user may be having and asks the user's input through suggestion chips for quicker and easier interaction.
AI Dictionary Interaction
Users can search through the knowledge base of PhysAI to better understanding and convey their medical concerns.
Novel Interactions
The user can interact through novel UI elements that allow them to convey the details of their problems (duration, intensity, and change) in a quantifiable way.
PhysAI Report
With the user's information confirmed, Suzy provides an initial diagnosis of the problem, with potential causes, and informs the user about some actions they can take to relieve symptoms.
Try it yourself!
Testing
For further evaluation of the application, we conducted user testing with 6 users in a one-hour session on Zoom in the form of cognitive walkthroughs and self-heuristic evaluations. Some feedback we received was:-
- More personalized questions like age or gender would help the application pinpoint symptoms and treatment better.
- The application currently follows one use case scenario and must be broadened into multiple to truly understand the impact of the app.
- Adding more interactivity features with the AI dictionary would increase its use.
- Conversational UX should be more sensitive depending on the symptom or question.
With this feedback in mind, we hope to further progress this application by creating more use case scenarios to overcome the complexities of healthcare.
FUTURE SCOPE
I think this application can be created so it’s specialized into different medical fields so it provides more personalized and accurate information. It can also be expanded into different fields like education, administration, and commerce, where there is a lot of knowledge and questions but not as many resources or people to answer them.
Thanks for looking through my work! 
Connect with me on LinkedIn or email me at sowmyaa.chandra@gmail.com. I would love to talk 
design, 
plants, or anything you have in mind!