Thecostoffalsesignals
A system that flags too many events as significant is, in practice, as unreliable as one that misses the real ones — it just fails in a quieter way. The failure mode isn't a missed crisis; it's a slow erosion of trust. Every false alarm costs a little bit of the attention the system is supposed to earn, and the person on the other end adapts the only way people reasonably do: they start skimming, then start ignoring, then eventually stop checking at all. By the time a genuinely important signal arrives, it's competing with a backlog of noise the system itself created.
This is why precision matters as much as coverage, and arguably more. It's tempting to build a system that surfaces everything that could conceivably matter — it feels safer, more thorough, harder to criticize for missing something. But a system tuned that way isn't actually being useful; it's transferring the filtering work back onto the person, which defeats the purpose of building it in the first place. The right target isn't maximum sensitivity. It's being right often enough, and clear enough about its own confidence, that when the system speaks, it's still worth listening to.