A system exhibits Automation Sedation when:
The pattern is self-reinforcing: suppressed signals produce calm interfaces, calm interfaces reduce operator intervention, reduced intervention allows conditions to persist, persistent conditions generate further signals that are suppressed.
- Repeated signals of unresolved failure are deduplicated, correlated, or suppressed as a function of repetition rather than resolution.
- The visual or operational salience of a condition decreases over time despite the condition persisting.
- Operators interact with an interface that presents manufactured calm rather than verified stability.
- Escalation probability decreases as a function of signal suppression, not as a function of recovery.
Modern observability and incident management systems implement several functions that, individually, serve legitimate noise-reduction purposes:
Each mechanism conflates two structurally distinct signal classes:
| Noise | Uninformative variance; does not change decisions | Suppress | | Persistence | Unresolved failure; recovery has not occurred | Escalate |
When systems treat persistence as noise — because both produce repetition — they erase the only signal that recovery has not occurred.
- Deduplication: Collapsing repeated alerts into a single incident.
- Correlation: Grouping related signals under a common root cause.
- Suppression: Reducing alert frequency for "known issues."
- Summary aggregation: Replacing granular signal streams with smoothed dashboards.
Observability collapse through incentive-aligned suppression.
The system optimizes for a metric (alert volume, dashboard calm, MTTR) that is structurally compatible with both genuine recovery and manufactured silence. The metric cannot distinguish the two. The optimization drives toward whichever is cheaper to produce. Silence is always cheaper than recovery.
This violates ZTG-0a (Observability Precondition): governance evaluation requires that system state be observable. When the observability layer actively suppresses evidence of unresolved conditions, governance has no valid input.
- Escalation delay: Conditions that would trigger human intervention under raw signal are invisible under processed signal.
- Operator deference: Repeated exposure to calm interfaces produces learned trust in unfounded confidence.
- Judgment atrophy: Human governance authority (ZTG-2) depends on humans who exercise that authority. When observability suppresses the signals, the judgment faculty degrades.
- Late, severe failure: Systems exhibiting this pattern fail late, not often. Incident frequency decreases. Incident severity increases.
- Governance invalidation: Any governance mechanism that depends on operator awareness is compromised.
Salience-Proportional Observability (SDS.RX.002): Unresolved conditions increase salience over time. Silence is permitted only following verified recovery. Failed resolution escalates the condition.
Automation Sedation => Observability Collapse => Governance Invalidation