Pattern Analysis, Not Surveillance
WhānauGuard's AI doesn't read every message word-for-word. Instead, it analyses conversation patterns — frequency, tone shifts, response times, and emotional trajectory — to detect when interactions become one-sided, hostile, or isolating.
What the AI Looks For
The system identifies cyberbullying by detecting these behavioural signals:
- Repeated negative messages from one or more senders to the same recipient
- Escalating hostility over short time windows (e.g. 10+ negative messages in 20 minutes)
- Isolation patterns — someone being excluded from group conversations
- Power imbalances — older accounts targeting younger ones
- Shift from neutral to hostile tone within a single conversation thread
Privacy by Design
The AI processes language features (tone, frequency, patterns) rather than storing or displaying the full content of every message. Only flagged interactions are surfaced to parents, with appropriate context so you can have an informed conversation with your child.
What You'll See
When a concern is detected, you receive an alert with severity level, the platform involved, and a suggested action. The AI explains what patterns it observed without exposing unnecessary private content.