A context-aware timing optimization prototype that adapts micro-action prompts based on user feedback, health constraints, and response patterns.
A context-aware timing optimization prototype that adapts micro-action prompts based on user feedback, health constraints, and response patterns.
This project is a React-based digital health prototype that demonstrates an adaptive timing engine for low-burden health behavior prompts.
The core idea is:
The intervention content is pre-written and controlled. The algorithm personalizes when to deliver the prompt.
Rather than generating new health advice, this prototype focuses on learning when a user may be most likely to respond to a two-minute movement prompt.
Many digital health tools know what behavior to recommend, but not when to prompt the user.
A person may ignore a health behavior reminder not because the recommendation is wrong, but because the timing does not fit their context, fatigue, pain, routine, or availability.
This project explores how a simple adaptive algorithm can personalize prompt timing based on user feedback such as:
- Did it
- Partially did it
- Skip
- Later
- Too tired
- Pain/discomfort
This project is an early prototype of an Adaptive Prompt Timing Engine.
The long-term product idea is to help digital health apps deliver prompts at the moment each user is most likely to act.
Possible use cases include:
- walking interventions
- chronic disease self-management
- physical activity support
- wellness coaching
- remote patient engagement
- digital health engagement optimization
- User profile setup
- Daily context input
- Context-aware trigger timing recommendation
- Transparent adaptive scoring logic
- Pre-written micro-action prompts
- Safety-aware note
- Explanation of why a time window was selected
- Feedback buttons
- Timing score updates
- Feedback history
- Personalization dashboard
- localStorage-based demo persistence
- Reset demo data function
The prototype uses two types of inputs.
- preferred time windows
- main barrier
- baseline confidence
- preferred prompt frequency
- pain concern
- fatigue today
- pain today
- motivation today
- weather
- availability today
The engine recommends:
- a time window for the next micro-action prompt
- a pre-written two-minute movement prompt
- a safety-aware note
- an explanation of why the time window was selected
- a score snapshot showing how the decision was made
The prototype uses transparent adaptive scoring logic.
Each time window starts with a default score.
The current version includes four candidate time windows:
- morning
- afternoon
- evening
- after dinner
The algorithm updates timing scores based on feedback.
Examples:
- Did it → selected time window score increases
- Partially did it → selected time window score slightly increases
- Skip → selected time window score decreases
- Later → selected time window score slightly decreases, later time window increases
- Too tired → selected time window score decreases
- Pain/discomfort → selected time window score decreases and safety mode is activated
This prototype does not generate medical advice.
The micro-action content is pre-written and controlled. The algorithm only personalizes timing.
The prototype includes safety-aware logic such as:
- high pain activates a safety mode note
- high fatigue reduces the score for more demanding time windows
- pain/discomfort feedback lowers the score for the selected time window
- the system prioritizes low-burden two-minute movement prompts
The project is based on the idea that behavior change support should consider not only what to recommend, but also when to recommend it.
A reminder sent at the wrong time may be ignored.
A prompt sent at a feasible moment may help initiate a small action.
This prototype explores timing personalization as a practical pathway for improving engagement with digital health interventions.
This project connects to my broader interests in:
- digital health
- healthcare AI
- behavior change
- chronic disease self-management
- just-in-time adaptive interventions
- adaptive intervention design
- responsible AI in health
- personalized prompt timing
- implementation science
- human-centered health technology
The central research question is:
Can user feedback and daily context be used to adaptively personalize the timing of low-burden health behavior prompts?
This prototype also explores a potential product direction:
An adaptive prompt timing engine that can be integrated into digital health apps to improve user response and behavior initiation.
Rather than building another reminder app, the long-term goal is to build a reusable optimization layer for digital health engagement.
- React
- Vite
- JavaScript
- CSS
- localStorage
- Vercel
This demo stores data locally in the browser using localStorage.
Stored data include:
- user profile
- daily context
- timing scores
- feedback history
- last recommendation
This version does not collect real patient data, personal health information, or identifiable information.
This prototype supports behavior change timing personalization and is not medical advice.
Users with pain, medical conditions, or safety concerns should consult a qualified healthcare professional before changing their physical activity routine.
Future versions could include:
- CSV export of feedback history
- simulated user testing
- PWA support
- push notification support
- backend database storage
- comparison of fixed timing vs adaptive timing
- contextual bandit algorithm
- integration with wearable data
- integration with digital health coaching platforms
- pilot testing with users managing chronic conditions
- Clone this repository.
git clone https://github.com/Yeongseo-Lee/adaptive-micro-action-trigger-engine.git- Move into the project folder.
cd adaptive-micro-action-trigger-engine- Install dependencies.
npm install- Start the development server.
npm run dev- Open the local URL shown in the terminal. It is usually:
http://localhost:5173/This is an early-stage adaptive timing optimization prototype for digital health behavior prompts.