[rch]

Maybe e-harmony, since it'd be double blind model-driven matching, but with near real time feedback while one is refining a set of preference parameters.