anxious device

📋 App Handout: Social Anxiety Context-Aware Intervention System

1. Executive Overview

Product Name: (To be determined — e.g., “ContextAnchor”, “SocialSignal”, “GroundLens”)

Purpose: A companion mobile application that receives real-time social context analysis from an existing camera-triggered device, prompts the user to self-assess their distress level, and delivers personalized, evidence-based interventions for social anxiety management.

Core Philosophy: The device captures the external context; the app captures the internal experience — together they create a complete picture for timely, contextual support.

---

2. End-to-End System Flow

┌─────────────┐     ┌─────────────┐     ┌─────────────────┐     ┌─────────────┐
│   PHYSICAL  │────▶│   SERVER    │────▶│  PUSH NOTIF     │────▶│  MOBILE APP │
│   DEVICE    │     │  (VQA + NLP)│     │  SERVICE (FCM/  │     │             │
│  (Built)    │     │             │     │   APNs)         │     │             │
└─────────────┘     └─────────────┘     └─────────────────┘     └──────┬──────┘
                                                                         │
                              ┌──────────────────────────────────────────┘
                              ▼
                    ┌─────────────────┐
                    │  USER RATES     │
                    │  DISTRESS 0-10  │
                    └────────┬────────┘
                             │
              ┌──────────────┼──────────────┐
              ▼              ▼              ▼
        ┌─────────┐   ┌──────────┐   ┌──────────┐
        │  LOW    │   │ MODERATE │   │  HIGH    │
        │  0-3    │   │   4-6    │   │   7-10   │
        └────┬────┘   └────┬─────┘   └────┬─────┘
             │             │              │
             ▼             ▼              ▼
        ┌─────────┐   ┌──────────┐   ┌──────────┐
        │Grounding│   │ Cognitive│   │ Crisis   │
        │  Tools  │   │Reframing │   │ Protocol │
        │         │   │ + Breathing│  │ + SOS    │
        └─────────┘   └──────────┘   └──────────┘

---

3. Phase-by-Phase Breakdown

Phase A: Device Trigger → Server (✅ Already Built)

Component	Status	Notes
Physical device with camera	✅ Built	Button-activated
Image capture & transmission	✅ Built	Sends to server endpoint
Server-side VQA model	✅ Built	Generates social context tags

Server Output Example (JSON payload to app):

{
  "event_id": "evt_abc123",
  "timestamp": "2026-08-22T17:34:00Z",
  "context_tags": ["crowded_room", "formal_gathering", "eye_contact_expected"],
  "environment_summary": "Indoor social event with multiple unfamiliar faces",
  "confidence_score": 0.87,
  "user_id": "usr_xyz789"
}

---

Phase B: Server → App Notification (🔧 To Be Built)

Trigger Condition: Server receives and processes image → immediately pings app’s push notification service.

Notification Payload:

{
  "to": "<device_fcm_token>",
  "notification": {
    "title": "Context Check-In",
    "body": "How are you feeling right now?",
    "sound": "gentle_chime.wav"
  },
  "data": {
    "event_id": "evt_abc123",
    "context_tags": "crowded_room,formal_gathering",
    "deep_link": "myapp://checkin?event=evt_abc123",
    "priority": "high"
  }
}

Notification Behavior:

• High priority (bypasses Do Not Disturb if user has opted in)
• Persistent notification (cannot be swiped away until acted upon)
• Custom gentle sound (non-alarming)
• Deep link opens app directly to check-in screen

---

Phase C: App Check-In Screen (🔧 To Be Built)

Screen 1: Context Acknowledgment

• Brief, non-clinical display of detected context (e.g., “Looks like you’re in a social setting”)
• No raw VQA output shown — translated into human-friendly language
• One-tap “This is accurate / This is not accurate” feedback (improves model over time)

Screen 2: Distress Rating (0–10)

• Large, accessible slider or vertical scale
• Anchors at extremes:
  • 0 = Calm, no distress
  • 5 = Moderately uncomfortable
  • 10 = Overwhelming panic / need to leave immediately
• Optional: One-tap “I need help now” button (bypasses rating, goes straight to crisis protocol)

Screen 3: Intervention Delivery (Dynamically selected based on score)

Score Range	Intervention Tier	Examples
0–3 (Low)	Grounding & Awareness	• 5-4-3-2-1 senses technique • Gentle posture reminder • “You’re doing fine” affirmation
4–6 (Moderate)	Cognitive + Somatic	• Box breathing animation • Cognitive reframing prompt (“What would you tell a friend in this situation?”) • Progressive muscle relaxation guide
7–10 (High)	Crisis + Exit Strategy	• Immediate grounding (cold water, feet on floor) • Pre-scripted exit phrases (“I need to step out for a moment”) • One-tap SOS to trusted contact • Link to crisis helpline

---

4. Technical Requirements

4.1 Backend (Server-Side Additions)

Requirement	Specification
Push Notification Service	Firebase Cloud Messaging (FCM) for Android + Apple Push Notification Service (APNs) for iOS
Real-time Delivery	WebSocket fallback for instant delivery if push fails
Event Logging	Every trigger → rating → intervention selection logged with event_id
API Endpoint	POST /api/v1/notify — called by VQA server after processing

New Endpoint Specification:

POST /api/v1/notify
Headers: Authorization: Bearer <server_token>
Body: {
  "user_id": "string",
  "event_id": "string",
  "context_data": { ... },
  "timestamp": "ISO8601"
}
Response: { "status": "delivered", "notification_id": "..." }

4.2 Mobile App

Requirement	Specification
Platforms	iOS 15+ and Android 10+
Deep Linking	myapp://checkin?event={id}
Offline Support	Interventions cached locally; rating stored locally if no connection, synced later
Background Handling	Notification must wake app even if killed; audio intervention should play in background
Accessibility	VoiceOver/TalkBack compatible; large touch targets; high contrast mode

4.3 Data Model

{
  "session": {
    "event_id": "evt_abc123",
    "user_id": "usr_xyz789",
    "triggered_at": "2026-08-22T17:34:00Z",
    "opened_at": "2026-08-22T17:34:12Z",
    "context_tags": ["crowded_room", "formal_gathering"],
    "user_context_feedback": "accurate",
    "distress_rating": 6,
    "intervention_type": "cognitive_somatic",
    "intervention_completed": true,
    "post_intervention_rating": 3,
    "session_duration_seconds": 145
  }
}

---

5. Intervention Content Library (Sample)

Tier 1: Grounding (0–3)

• 5-4-3-2-1 Senses: Interactive countdown with haptic feedback
• Posture Check: Animated figure demonstrating open posture
• Ambient Sound: 60-second nature audio overlay

Tier 2: Cognitive + Somatic (4–6)

• Box Breathing: Visual pacer with 4-4-4-4 rhythm
• Thought Reframe: Mad-libs style cognitive restructuring (“I feel ____ because ____. But actually, ____.”)
• Progressive Relax: Tense-and-release guide for shoulders, jaw, hands

Tier 3: Crisis + Exit (7–10)

• Physical Grounding: “Find 1 cold thing. Press it to your wrist. Breathe 3 times.”
• Exit Scripts: Pre-written, customizable text messages to send to host/friend
• SOS Contact: One-tap call/text to designated emergency contact
• Crisis Resources: Local helpline numbers (geolocation-based)

---

6. Privacy, Security & Ethics

🔒 Data Handling

Aspect	Policy
Image Storage	Raw images are NOT stored on the phone. Only context tags transit to app.
Encryption	All data in transit (TLS 1.3) and at rest (AES-256)
Retention	Session logs retained for 90 days; anonymized for research after
User Control	One-tap “Delete all my history” in settings
Consent	Explicit opt-in required before first use; revocable anytime

⚠️ Ethical Safeguards

• No Diagnostic Claims: App is a support tool, not a clinical diagnosis instrument.
• Mandatory Disclaimer: “This app is not a substitute for professional mental health care.”
• Crisis Escalation: If user rates 10 + taps “I need help now” → immediate offer to call emergency contact or crisis line.
• No Surveillance Feel: Context is described, never shown as “we saw you at X with Y people.”

---

7. User Journey Map

Step	User Action	System Response	Emotional Goal
1	Presses device button	Camera captures, server processes	Discreet, low-friction
2	Phone buzzes	Notification appears with gentle tone	Non-alarming, inviting
3	Taps notification	App opens to context screen	Validating (“We noticed”)
4	Confirms context	Slider appears for distress rating	Empowering (user in control)
5	Selects 0–10	Matched intervention loads	Supportive, not judgmental
6	Completes intervention	Optional: “How do you feel now?”	Reinforcing progress
7	Closes app	Session logged, insights accumulated	Seamless, no guilt

---

8. Success Metrics (KPIs)

Metric	Target	Measurement
Notification Open Rate	>85%	% of pushes that lead to app open within 5 min
Rating Completion Rate	>90%	% of opened sessions where user submits 0–10
Intervention Completion	>70%	% of users who finish the intervention flow
Post-Intervention Reduction	Avg ↓2 points	Difference between pre and post distress scores
7-Day Retention	>60%	Users still engaging after one week
Crisis Escalation Usage	<5%	% of sessions ending in SOS/crisis contact (indicates appropriate tiering)