The case for a "deficit model" of science communication