Mengapa Sistem Kompleks Gagal
Kegagalan besar jarang punya satu penyebab. Ia tumbuh dari kesalahan kecil yang bertemu dalam sistem yang saling terikat erat.
When a power grid collapses or a financial market seizes, the first question is usually who made the mistake. The more useful question is why the system let an ordinary mistake become a disaster.
## Normal accidents
Sociologist Charles Perrow argued in Normal Accidents (1984) that some systems make failure close to inevitable. Two properties matter. Interactive complexity means parts can affect each other in ways designers did not anticipate. Tight coupling means there is little time or slack to stop a problem once it starts.
Put the two together and small faults can combine and spread faster than people can understand them.
> In a tightly coupled system, the safety margin is the time between the first fault and the point of no return.
## Cascades and hubs
Networks with a few highly connected hubs are robust to random damage and fragile when a hub fails. Supply chains, airline routes and the internet all share this shape. A disruption at a hub travels outward along every link.
Failure is also delayed by feedback. Systems can look healthy while slack is quietly consumed, then flip quickly when it runs out. That is why collapse often feels sudden even when its causes built up for years.
Designing for failure means accepting some inefficiency: buffers, redundancy, modular parts that can be isolated, and people with the authority to slow things down.