_edited.png){kind=small}
.png)
Exploring how conversational AI can help patients navigate healthcare decisions while maintaining transparency, confidence, and control.
Designing Trust Between Humans and AI
Role
UX/UI Designer
Duration
8 Weeks
Focus
AI Trust, Conversational UX, Healthcare
Methods
Research, Prototyping, Usability Testing
Outcome
Increased trust, engagement, and perceived intelligence of the AI experience
Healthcare Is Complex.
Trust Makes It Harder.
*Diagram from Walls et al (2004) Health, Risk, and Society 6(2): 133-150
Healthcare decisions are often emotionally charged, time-sensitive, and difficult to navigate.
As AI systems become increasingly capable of supporting healthcare journeys, a critical question emerges:
Why do some AI interactions feel trustworthy while others feel cold, confusing, or unreliable?
This project explored how interaction design influences users' willingness to trust AI recommendations in high-stakes situations.
_edited.png)
.png)
.png)
Can Design Increase Trust in AI?
Rather than starting with interface solutions, I began with a behavioral question:
How do users determine whether an AI assistant deserves their trust?
Through user interviews and exploratory research, I identified several recurring concerns:
-
Lack of transparency
-
Unclear reasoning
-
Fear of incorrect recommendations
-
Difficulty understanding system capabilities
These concerns became the foundation of the design process.
What Users Taught Me
Trust Isn't Rational. It's Emotional.
Early research revealed an interesting pattern.
Participants weren't evaluating the AI purely on accuracy. Instead, they evaluated it through social and emotional cues:
-
Tone of voice
-
Perceived empathy
-
Conversational flow
-
Clarity of explanations
-
Transparency about uncertainty
The experience often felt less like evaluating software and more like evaluating another person. This insight fundamentally changed the direction of the project.
%201.png)
%201.png)
.png)
.png)
Building an Efficient Assistant
The initial prototype focused on helping users quickly find relevant providers and healthcare information.
The interaction was functional and efficient, but testing revealed an unexpected challenge.
While users appreciated the speed and convenience, some described the experience as transactional and foreign-feeling.
The AI solved the task.
It didn't earn trust.
Designing for Human Connection
Research findings suggested that users wanted more than recommendations.
They wanted confidence.
To address this, I redesigned the interaction to feel more conversational, transparent, and supportive.
Changes included:
-
Visual representation of a doctor - a familiar likeness
-
Clear explanations behind recommendations
-
Human-centered language
-
Visible reasoning
-
Increased transparency around decision-making
The goal wasn't to make the AI feel human,
but to make the interaction feel understandable.
.png)
How User Perception Changed - Measuring the Difference
Follow-up testing revealed a significant shift in how participants described the system. Users did not evaluate the AI feature on speed alone. Trust was shaped by human-centered interaction patterns that made the experience feel guided, understandable, and natural.
Instead of focusing solely on functionality, users began discussing:
-
Trustworthiness
-
Confidence
-
Comfort
-
Helpfulness
-
Understanding
Small interaction design changes produced disproportionately large changes in perceived intelligence and reliability.
The findings reinforced a key principle:
People don't trust AI because it is intelligent.
People trust AI when they understand it.
.png)
As part of the project, I also created a design system that served as a style guide. It was built on Zocdoc's familiar and eye-catching current design.
Design System
.png)
The introduction of new colors coincided with the feature launch, signaling to users that they remained within a familiar experience while interacting with new functionality.
Reflection
This project changed how I thought about AI products. Modern AI systems face the same trust, transparency, explainability, and behavioral adoption challenges identified during this research. The project helped establish foundational principles that continue to influence how I design AI-powered experiences today.
I left this project realizing the real challenge was helping users trust the system providing them information.
As AI becomes increasingly embedded into everyday decision-making, trust may become one of the most important design problems of the next decade.
This experience sparked my continuing interest in designing intelligent systems that balance capability, transparency, and human understanding.
