📱 Screen Sense

Intelligent Behavioral Intervention for Digital Wellness

Screen Sense moves beyond static screen-time tracking. By analyzing app usage sequences in real-time, it identifies dopamine loops, predicts digital fatigue, and delivers proactive wellness nudges — before migraines and eye strain take hold.

---

🏗️ System Pipeline

┌─────────────────────────────────────────────────────────┐
│                    LAYER 1 — SENSORS                    │
│                                                         │
│  UsageStatsManager          ObservationService          │
│  (Historical Analysis)  +   (Real-Time Monitoring)      │
│  Nightly batch jobs         AccessibilityService hook   │
└────────────────┬───────────────────┬────────────────────┘
                 │                   │
                 ▼                   ▼
┌─────────────────────────────────────────────────────────┐
│                  LAYER 2 — DATA PIPELINE                │
│                                                         │
│  Kotlin Coroutines + StateFlow (non-blocking reactive)  │
│  WorkManager (nightly N-Gram mining + session builder)  │
│                                                         │
│  Room SQLite ──► AppUsageEvents                         │
│   (on-device)  ──► LoopFingerprints                     │
│                ──► UserProfile                          │
└────────────────────────┬────────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────────┐
│                LAYER 3 — SENSE ENGINE                   │
│                                                         │
│  ┌─────────────────┐     ┌──────────────────────────┐  │
│  │ N-Gram Analyzer │     │  Sleep Threat Scorer     │  │
│  │                 │     │                          │  │
│  │ • Gateway apps  │     │  Score: 0–100 (5min)     │  │
│  │ • Loop detect   │     │  Inputs:                 │  │
│  │   A→B→A chains  │     │  • Mins past sleep time  │  │
│  │ • Chain mining  │     │  • App-switch velocity   │  │
│  └────────┬────────┘     │  • Active loop flag      │  │
│           │              └─────────────┬────────────┘  │
│           └──────────────┬─────────────┘               │
└──────────────────────────┼─────────────────────────────┘
                           │
                           ▼
┌─────────────────────────────────────────────────────────┐
│             LAYER 4 — INTERVENTION MANAGER              │
│                                                         │
│  🔴 Sleep Alert      Score > threshold → High Priority  │
│  🔁 Loop Alert       Known chain triggered → Instant    │
│  💆 Wellness Nudge   Duration + profession → Suggest    │
│  📊 Stress Flag      Velocity spike → Context switch    │
└────────────────────────┬────────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────────┐
│              LAYER 5 — VISUALIZATION                    │
│                                                         │
│  • Switching Velocity   Context-switch spike graphs     │
│  • Gateway Analysis     Trigger app heatmaps            │
│  • Mood Correlation     Loop → wellbeing tracking       │
│                                                         │
│  Stack: MPAndroidChart / Compose-Charts                 │
└─────────────────────────────────────────────────────────┘

---

🧠 Key Behaviors

Behavioral Loop Detection

Tracks sequences, not just durations.

• Gateway Apps — detects apps (e.g. Slack) that consistently lead to unintended sessions (e.g. 30min Instagram scroll)
• Loop Detection — flags circular chains: App A → App B → App A using on-device N-Gram mining

Predictive Sleep Threat Score

Dynamic 0–100 score, recomputed every 5 minutes during late-night hours.

Input	Weight
Minutes past target sleep time	High
App-switching velocity	Medium
Active loop detected	Trigger

Fires a High Priority alert before digital exhaustion sets in.

Real-Time Intervention Engine

Built on AccessibilityService — monitors foreground transitions with zero perceptible lag.

---

📐 Component Responsibilities

Component	Responsibility
SenseEngine	N-Gram mining, sequence matching, Sleep Threat scoring
ObservationService	Real-time app-switch detection at system level
InterventionManager	Multi-channel alerts: Sleep, Loop, Stress, Nudges
Room Database	On-device storage: AppUsageEvents, LoopFingerprints, UserProfile
Visualization Layer	Interactive charts via MPAndroidChart / Compose-Charts

---

⚙️ Tech Stack

Language      Kotlin
Concurrency   Coroutines + StateFlow
Background    WorkManager
Database      Room SQLite
Monitoring    UsageStatsManager + AccessibilityService
Charts        MPAndroidChart / Compose-Charts
Storage       On-device only — zero cloud

---

📥 Setup

Permissions required

• Usage Access
• Accessibility Service
• Notifications

Onboarding Enter profession (Student / Professional) and target sleep time → calibrates scoring engine.

---

🛡️ Privacy

Principle	Detail
Zero-Cloud	All N-Gram mining, hashing, and scoring runs on-device
Minimal Data	Package names + timestamps only. No message content, passwords, or PII
Medical Disclaimer	Wellness support tool only — not a diagnostic tool for migraines or clinical illness

---

👥 Team

👥 Team

Member	GitHub
Nimish Sharma	(https://github.com/Nimish-Sharma-dev)
Parasmani Kushwaha	(https://github.com/parasmani-dev)
Manaswi Raj Dubey	](https://github.com/manaswiraj)
Divyansh Singh	(https://github.com/Divyansh-Singh-1)