MedPal
Helping users consume health information online
01 | Overview
MedPal is an AI chatbot that helps users verify, contextualize, and summarize online content, enhancing the experience of researching health-specific information.
For this project, I owned:
User research
Led evaluative & generative research initiatives and analyzed key findings
Information architecture
Created IA diagrams
Interaction design
Created user flows and interactions
Role
Team lead, designer, researcher
Team
3 designers, 1 researcher
Context
Academic project
Duration
Sept - Dec 2023 (3 months)
Skills
User interviews, diary studies, affinity diagramming, rapid prototyping, evaluative testing
Problem Statement
How might we support people in staying accurately informed about their health conditions?
There is a vast amount of online sources to navigate through, each full of lengthy walls of text and medical jargon. Sources also vary in quality and accuracy, and fail to provide contextual information. Fact-checking can be time consuming, and you end up with too many tabs to keep track of. We wanted to create a solution that would empower users to fact check their information on their own accord.
02 | Solution overview
Customizable information flagging
During the onboarding stage, users decide what type of information they want MedPal to flag.
Chatbot
When prompted, the MedPal chatbot offers summaries, suggested prompts, and citations to help users better contextualize health-specific content
Saved conversations
Users can bookmark specific conversations to refer back to. Based on what is saved, MedPal will adjust its answers to provide users with information specific to different health situations.
03 | Research summary
Research questions
We wanted to investigate how people look for, receive, and process health-related information online.
We conducted:
3
Semi-structured interviews
We interviewed 3 participants over Zoom about how they navigate health information
3
3-day diary studies via text/email
We followed up with 3 of our participants to observe and explore what kinds of health-related info they encountered that day.
1
SME interview
We spoke with a doctor about her observations when assessing patients' information browsing habits.
Findings & Analysis
We color-coded and categorized our notes into 3 major themes:
learning & information, family & relationships, and relationship with health/diseases
Color coding interview notes into Figjam
Insights:
01
Participants actively consume health-related information.
Participants constantly seek out or unintentionally come across health-related information online. Browsing health topics is part of their every day lives and is a personal experience for everyone.
02
Participants value trust in the information they're consuming.
Participants validate the information they learn by cross-referencing different sources. They expressed that they’re more likely to trust information that comes from reliable sources.
03
Participants try to manage their own health.
Maintaining good health is a priority for participants, and they often try to achieve that without needing to see a medical provider.
05 | Ideation
Desired outcomes
We want our solution to:
Prevent consequences that come with misleading health information
Make information related to health more digestible
Integrate with its users' existing internet browsing habits
Brainstorming, aligning, and dot voting
We brainstormed individually and as a group, then categorized and voted on the ideas to select the one we would explore further.
Brainstormed ideas
Categorized similar ideas together
Organized ideas along 2 axes: user- vs. system-initiated, and personal vs. community-focused
Design principles
We voted for the idea we wanted to focus on based on the design principles we agreed upon below:
Create a feeling of control
Our solution will give users a sense of agency and independence. It will help, not force, users to achieve their goals in a way that doesn’t disturb their overall lifestyle/ established habits.
Acknowledge user differences
Our solution will be tailored so that it supports people with different conditions, severity of conditions, and/or goals. It will be easy to understand and provide concise information upfront.
Create trust
Our solution will establish a sense of trust. It will create a judgement-free environment that will not make the user feel penalized or guilty in any way.
Selected idea
We explored our idea of a "browser buddy" that automatically fact-checks health-related information on a webpage. We wanted users to be able to follow up on their own as well.
User flow diagram
06 | Concept testing
Testing with paper prototypes
To test this idea, we tasked 3 participants to go through our imagined solution's onboarding flow and chatbot interactions, asking them to think out loud as they progressed.
Onboarding flow
What do participants expect when shown the personalization options?
We walked each participant through MedPal's onboarding flow to see what they would use MedPal for.
Receiving a notification
What is the best way for MedPal to notify users of inaccurate health information?
While participants read through the webpage, we presented 3 versions of notification types to assess their reactions to the level of intrusion and usefulness.
Slide-in
Chatbot
Pop-up alerts
Interacting with MedPal
Will participants trust the information MedPal provides?
Finally, we presented participants with the MedPal notification. We tested to see what their thoughts are if they tap on the notification to expand the chatbot and interact with it.
Notification
Sample conversation w/ chatbot
Insights:
01
Providing relevant context of the health-related information is valuable.
“It's hard for MedPal to give me a good answer without the context, it could mean anything.” - P5
02
Communicating the way MedPal works help users decide the validity of information.
“I am wondering how it fact-checks. Are doctors approving it? Is AI doing it?” - P3
03
Displaying information in more digestible ways helps users understand and learn about misinformation.
“I'm generally not that interested in the details but if you give it to me in a nice way I'll read it.” - P4
04
Presenting the browser widget in the form of a chatbot balances intrusiveness and effectiveness.
“I would prefer this more because it's short enough I can digest it immediately.” - P3
08 | Final design
Saved conversations
Users can save conversations with MedPal to refer back to. Saving conversations lets MedPal learn more about its user and what information is important to them.
Design system
We wanted to emphasize a feeling of friendliness, so we chose teal as the main color. To convey trustworthiness and professionalism, we used simple line graphics and UI elements.
09 | Next steps
If I were to continue this project, I would:
Conduct another interview session to specifically test how successful MedPal is in facilitating trust.
Conduct further usability tests with high-fidelity prototypes to see MedPal in real-life contexts.
Expand content by considering other kinds of sources and adding more personalization features.
10 | Reflection
Expanding initial target audience
We initially set out to focus on older adults over the age of 65, recognizing their increased likelihood of experiencing health conditions and lower levels of digital literacy. However, as our research progressed—through team workshops, user interviews, and feedback sessions—we began to see opportunities for broader impact, leading us to revisit and redefine our original design challenge.
Privacy, risks, and concerns
Throughout this project, we remained mindful of the potential risks associated with MedPal. Key concerns—such as the ethical implications of AI, data privacy, and the risk of overreliance or misuse—sparked important discussions within our team. These conversations not only shaped our understanding but also informed design decisions, many of which continue to evolve through ongoing iterations