**Social-Contact / Network Sub-Score (Pedalgo Methodology)**

The network sub-score is one of several modular inputs used solely to reorder a human review queue. It never produces an automated action, public label, or determination of guilt. All values below are illustrative and synthetic until platform-specific, reviewer-confirmed data exist.

### Core Network Components

- **Minor-fanout**  
  Measures the rate and breadth of outgoing contact attempts toward accounts whose self-reported or platform-inferred age falls in the minor range. High fanout (many distinct minor-linked accounts in a short window) increases the sub-score. The metric is age-agnostic in weighting and uses a single universal prevalence base rate.

- **Initiation asymmetry**  
  Captures imbalance in who starts conversations. When an account consistently initiates contact with minor-linked accounts while receiving few or no inbound messages from them, the sub-score rises. Reciprocal or adult-initiated patterns receive neutral or lower weighting.

- **Burst activity with new accounts**  
  Detects sudden spikes of interaction involving accounts created within a narrow recent time window (e.g., days). Coordinated bursts across multiple newly created accounts elevate the sub-score because they deviate from typical gradual network growth.

- **Off-platform migration attempts**  
  Flags patterns in which an account repeatedly suggests, shares, or requests transition to external channels (encrypted messaging, private servers, etc.) shortly after initial contact. Each verified attempt contributes positively but is capped to avoid over-weighting single events.

- **Proximity to previously actioned actors**  
  Records graph distance to accounts that platforms have previously removed or restricted for child-safety violations. This component receives very low weight because:
  - Shared social circles can occur innocently (school, hobby, location-based communities).
  - Historical action data may contain errors or over-removals.
  - Over-reliance risks indirect stigma without direct behavioral evidence.
  It functions mainly as a weak tie-breaker rather than a primary driver.

### Integration with DAC into Final 0–100 Score

The network sub-score is linearly combined with the DAC (Direct Account Content) score using fixed, publicly documented coefficients. The resulting raw value is then passed through a logistic calibration function that maps it to the 0–100 range. Calibration is performed so that the distribution of scores across a large reference population approximates expected prevalence under the universal base rate; the mapping is monotonic and preserves rank order only.

Final score = calibrated( w₁·DAC + w₂·Network_subscore )  
where w₁ and w₂ are public weights chosen during methodology development. The 0–100 output is used exclusively to sort the review queue; human reviewers apply full context and legal thresholds before any escalation (NCMEC CyberTipline, Swedish Police, IWF, or DSA trusted flagger channels).

### Synthetic Network Patterns That Can Elevate Review Priority

**SYNTHETIC — FOR ILLUSTRATION ONLY — NOT REAL CASES**

- Pattern A: An account created 14 days ago sends first-contact messages to 11 different accounts whose profiles indicate ages 13–15 within a 48-hour window, with zero inbound replies. Network sub-score rises sharply even if textual content remains minimal.
- Pattern B: The same account initiates 8 conversations, proposes moving to an external encrypted app after the third message in each thread, and shows two-hop proximity to one previously removed account. Combined with moderate DAC, the calibrated score places the account in the upper review queue.
- Pattern C: Three accounts created within 36 hours of each other exhibit identical burst timing and identical off-platform migration phrasing toward overlapping sets of minor-linked profiles. Each receives an elevated network sub-score despite sparse individual content signals.

These patterns reorder the queue for human attention. No automated enforcement occurs. All downstream decisions remain with trained reviewers and lawful reporting pathways.