Digital Placebos: Alarms That Teach

Some alarms should teach, not just beep. This book argues for a quiet, verifiable alternative to crisis theater: short, scheduled exposures that build competence in the flow of work. “Teaching alarms” prefer proof over performance and repeatability over charisma. They make explicit claims in units the business cares about, and they carry their own receipts, so trust does not depend on speeches — or dashboards that hide harm in averages.

You’ll learn to wire a complete circuit you can verify end-to-end: telemetry → alert → action → receipt → tuning. The alert states the claim; the runbook proposes the minimum steps that could falsify it; the receipt tests the same unit the claim lives in; the promotion gate decides ON/OFF based on evidence, not hope. A κ-gate protects attention by pausing training when the curvature between demand and capacity spikes. Fatigue has a hard budget by team and shift. Policies are unit-bound, auditable, reversible — and boring in the right way.

Ethics is not a disclaimer here; it’s the operating system. Teaching alarms run transparently or under prior consent, collect only what safety and learning require, and can be turned off without breaking the service. No punitive leaderboards. No silent surveillance. Exposure is distributed fairly. We measure systems, not people. You’ll see formulas because they cut ambiguity, cases with control groups. After all, effect matters more than story, and dashboards are used as anti-narrative, not decoration.

You will learn how to:
• Design teaching alarms with explicit units, bands, and gates.
• Write one-page receipts that make claims pass/fail today.
• Budget fatigue and apply κ-gates to protect attention.
• Govern data use with consent, minimal collection, and reversibility.
• Replace averages with tails and segments that reveal harm.
• Prove competence gains with traces, manifests, and controlled tests.