S · SYSTEM
Failure modes
Notice how an arrangement can go wrong.
How do I do it?
Notice how an arrangement can go wrong.
Choose a named move below to open its instructions.
- Substrate mismatch
The tool or person you picked is good at the shape of the job but bad at the part that actually matters.
- Local optimization
One part of your setup gets faster while the whole thing gets slower, costlier, or riskier.
- Coordination tax
Sometimes the messaging, scheduling, and checking around a task quietly costs more than just doing the task.
- Context collapse
When work moves to a new person or tool, the unwritten details that made it work get left behind, and private details that should have stayed put get carried along.
- Accountability diffusion
When many people and tools touch a job but no one person is on the hook for the final result, the job quietly stops having an owner.
- Correlated independence
Several opinions can agree because they share a source, a method, or a paycheck, so they feel like proof when they are really one opinion repeated.
- Automation bias
Automation bias is when you trust a tool's answer more than the evidence behind it, usually because it has been right often enough that you stopped checking.
- Exception blindness
A setup that handles the ordinary case perfectly can fail silently on the rare case, and because nothing ever alarms, you keep trusting it.
- Deskilling
Deskilling is when a tool makes your results better while your own ability to do the thing without it quietly gets worse.
- Agency erosion
A system can quietly take away someone's real say in their own life even though nothing broke and nobody objected.
- Privacy or security spill
You gave a person, app, or AI more data, access, storage time, or audience than the job actually needed, so harm can happen even when everything works correctly.
- Brittleness
Brittleness is when your whole setup rests on one login, one supplier, one person, or one rule, so the hour that single thing fails, everything stops.
- Automation debt
The rules, shortcuts, and automations you set up keep running long after the world they were built for has changed, so they quietly go wrong while still looking like they work.
- Alert fatigue
When warnings arrive so often that people stop reading them, the one warning that mattered gets ignored along with the rest.
- Dehumanization
When you speed something up by routing it through a tool, the human part of it — presence, care, apology, trust — can quietly disappear while the task still looks done.
- Coercive defaults
You get people to go along with you by making the alternative hard to reach instead of by convincing them, so their agreement is not really agreement.
- Surveillance creep
Monitoring you set up for one small safety reason keeps growing until you are watching far more than the problem ever required.
- Incentive mismatch
The person, company, or tool you depend on is rewarded for hitting its own number, which is not the same thing as getting you the result you actually came for.
- Loss of separation
When the same person or tool does the work, checks the work, and decides whether to go ahead, one mistake sails through every stage because nobody uncoupled from it is left to object.