If the background unhappy chatter is close to actionable (or is in the mind of the classifying algorithms), then actionable chatter (eg, actually planning an attack) can only be distinguished from background chatter near when the plans become "active" in some sense, and likely too late to meaningfully respond to. (To use a physics analogy: if you try to predict cavitation in flowing water really close to boiling, it's harder.)
In many senses, simply embracing openly that we're very angry as a society, even in the face of constant scrutiny, prevents that same constant scrutiny from determining when we're planning something. It can only really tell that we're pissed off, in general.