Well, to be sure, there's probably some reasonable empirical grounds to believe that the Taylor Rule is a reasonable model (with the emphasis on model). Many central banks across the world (BoE, BoC, RBNZ, etc) use these approaches. Like all models, these aren't perfect, but then what in life is?