I agree with the gist of everything you've written here. And I think you're slightly missing the point. > "so much of crime comes from social context" Agreed.

I agree with the gist of everything you've written here. And I think you're slightly missing the point. > "so much of crime comes from social context" Agreed. This is why I advocate taking that context into account. Not profiling would mean treating every single individual the exact same way, removing all social context from consideration. > at what point could you ever say: it's safe to assume that anyone with X ethnic background is not a threat This is not the point of profiling. The point is to build probability distributions, and hone them. > I don't see a way to profile without affecting the statistics. This is literally the point! You want to profile on such an exacting superset of factors (behavior, history, autonomous responses, forensics...) that you cause the incidence rate of terrorism to go down, across the board. A brittle policy like "only screen middle eastern men" will of course get gamed, but that is terrible profiling. It builds a one-dimensional "profile"--you want a profile that is harder to fool, and based on actual behavioral cues.