Reducing Effort in Health Data Interpretation Through Ethical Personalization
Conceptual UX redesign
AI-assisted personalization
Problem
High cognitive effort and low clarity when interpreting personal health data.
Context
Mobile health tracking application focused on menstrual and reproductive health.
Health apps don’t fail because users lack data, they fail when users have to work too hard to understand what the data means for them.
Problem Definition
The app provided detailed health data and algorithmic insights, but meaningful information was difficult to find and required excessive navigation and manual setup. As a result, users lacked confidence interpreting their own health patterns.
Core Problem
How might we help users quickly understand what their health data means without requiring expertise or extensive effort?
Affected Users
Brittney
A busy user who wanted quick, high-level insights to feel reassured about her cycle without reading through dense data.
A busy user undergoing hormone therapy who wanted to understand how treatment affected her overall well-being through clear patterns and visuals.
The core issue wasn’t a result of missing functionality. It was low confidence driven by cognitive load, fragmentation, and unclear system behavior.
Context and Constraints
Why the Problem Existed
-
The app already provided AI/ML generated insights, but neither the users nor I initially recognized them as such
-
Participants didn’t associate existing features with AI and often assumed the system was entirely rules-based
-
Insights were buried deep in the experience and framed passively, reducing both visibility and perceived value
-
Personalization relied on manual customization that users rarely completed
AI should not be invisible or defaulted. Trust needs to be intentionally designed through explanation, consent, and control.
Constraints
-
Conceptual redesign without access to production data
-
Sensitive health information requires explicit consent and clear communication
-
Limited scope meant focusing on comprehension, effort, and trust, instead of long-term behavior change or clinical outcomes
A lack of transparency around AI creates confusion, incorrect assumptions, missed value, and mistrust.
Both the context and constraints directly shaped a design approach focused on clarity, cognitive ease, and explicitly communicated AI behavior.
Key Decisions
Surface Insights Earlier in the Experience
Initial Assumption
Users would be willing to navigate to the Analysis section to find AI-generated insights.
Usability Testing Round 1
Participants consistently described insights as “helpful but hard to find” and expressed frustration with the number of screens required to reach them.
The issue was cognitive overload caused by irrelevant options and poor prioritization.
Iteration
Moved personalized insights to the home screen and introduced feature prompting to preserve depth without overwhelming users.
Usability Testing Round 2
Participants were able to explain what the insights meant without additional prompting and reported feeling more confident interpreting their health trends.
Outcome
Reduced navigation effort while preserving access to detailed data for users who wanted deeper exploration.
Make AI Explicit, Transparent, and Optional
Identified Risk
Because users were unaware that AI-generated insights already existed, introducing new personalization features without explanation risked increasing distrust rather than value.
Usability Testing Round 1
Participants did not reject AI itself, but repeatedly raised concerns about:
-
Not knowing when AI was being used
-
How their sensitive data is handled
-
Whether they had control over participation
Users are uncomfortable with invisible decision-making.
Iteration
Designed an explicit opt-in flow that clearly communicated:
-
What data would be used
-
How insights were generated
-
How users could opt in/out at any time
Usability Testing Round 2
Participants described the AI feature as “helpful” rather than intrusive and expressed greater willingness to engage with personalized insights.
Outcome
Transparency shifted AI from a perceived risk to a supportive, user-controlled feature.
Reframe the Problem for Hormone Therapy Tracking
Initial Assumption
The problem was difficulty logging hormone therapy dosage.
Usability Testing Round 1
Participants successfully logged dosage but struggled to locate relevant traits during daily tracking and often overlooked customization controls.
Low discoverability undermined perceived usefulness, regardless the quality of the insights.
Iteration
Introduced AI assisted trait prioritization that surfaced the most relevant traits based on user behavior.
Usability Testing Round 2
Participants completed tracking faster and described the experience as “more tailored” without requiring manual setup.
Outcome
Reduced cognitive load while preserving flexibility for users who preferred detailed tracking.