Signals
What data systems read — and what it cannot tell us.
Purpose
This document defines signal principles that apply across all Governor HQ domains. Every signal processed by systems built with this framework must be understood within these constraints.
Domain-specific signals vary: Wearables read HRV and sleep stages, BCI systems read brain waves, therapy apps read mood entries. But the underlying principles are universal.
A developer reading this page should know:
- What can be read from various data sources
- What can be derived with reasonable confidence
- What must never be inferred from this data
Universal Signal Principles
1. Trend-based, not absolute
All signals are meaningful relative to the individual's baseline, not as standalone values. A single measurement means nothing without personal context.
Examples across domains:
- Wearables: An HRV of 35ms means nothing without context. An HRV that dropped 30% from baseline is meaningful.
- BCI: A beta wave frequency means nothing alone. A sustained shift from personal baseline requires attention.
- Therapy: A mood rating of 4/10 means nothing. A consistent 2-point drop from baseline is meaningful.
2. Noisy
Consumer health sensors produce estimates, not clinical measurements. Single-event anomalies are expected and must not trigger alerts alone.
3. Delayed
Many signals reflect what happened hours or days ago, not the current moment. Systems must account for this latency when interpreting data.
4. Influenced by confounders
Countless external factors affect every signal. The system cannot isolate causality — only observe patterns.
Signal Examples by Domain
🏃 Wearables & Fitness
| Signal | Source | Characteristics |
|---|---|---|
| Sleep Stages | Wearable (accelerometer + PPG) | Estimated, not clinical polysomnography |
| HRV | Wearable (PPG) | Trend-based, noisy, influenced by lifestyle |
| Resting Heart Rate | Wearable (PPG) | Delayed indicator, more stable than HRV |
| Activity Load | Derived (movement + GPS) | Training stress proxy, not medical |
See: Wearables Signal Documentation for complete details.
🧠 Brain-Computer Interfaces
| Signal | Source | Characteristics |
|---|---|---|
| Alpha/Beta/Theta Waves | EEG | Noisy, influenced by electrode placement |
| Focus Score | Derived (beta/theta ratio) | Estimate, not absolute cognitive state |
| Meditation Index | Derived (alpha dominance) | Rough proxy, individual variance high |
Note: BCI-specific signal documentation coming soon.
💭 Therapy & Mental Health
| Signal | Source | Characteristics |
|---|---|---|
| Mood Ratings | User-reported | Subjective, context-dependent |
| Journal Sentiment | NLP derived | Approximate, language-limited |
| Behavioral Patterns | Usage tracking | Correlation only, not causation |
Note: Therapy-specific signal documentation coming soon.
What Signals Cannot Tell Us
This principle applies universally across all domains.
| Impossible Inference | Why |
|---|---|
| Diagnose a condition | Consumer health tools are not medical devices. Signal accuracy is insufficient for diagnosis. |
| Determine causality | A change in any signal does not tell you why it changed. Correlation ≠ causation. |
| Prescribe treatment | No signal combination justifies medical intervention. |
| Replace a professional | Data cannot replace clinical judgment, professional assessment, or therapy. |
| Predict disease | Trends may correlate with health outcomes in research, but individual prediction is not validated. |
Population Averages Are Not Decision Inputs
Systems built with Governor HQ explicitly reject population averages as primary decision criteria.
Why:
- "Normal" ranges vary by age, demographics, genetics, and measurement method
- Comparing users to population averages creates anxiety without actionable insight
- The only meaningful comparison is personal baseline vs. current state
Rule: Population data may be used for initial context during onboarding but must never drive recommendations or deviation detection.
Developer Guidance
DO
- Compare signals to the user's personal baseline
- Require multi-signal confirmation before triggering alerts
- Treat single anomalies as noise unless part of a sustained pattern
- Display trends, not snapshots
- Acknowledge uncertainty and limitations
DON'T
- Show "normal ranges" based on population data as decision criteria
- Use language like "low" or "high" without baseline context
- Infer health status, mood, or condition from a single metric
- Present consumer health data as medically accurate
- Make causal claims from correlational data
Bottom line: All health signals are behavioral feedback tools, not diagnostic instruments. Build accordingly, regardless of domain.