This is an interesting idea .. but then you are left with "how much do you trust the [human] vs how much do you trust the [algorithm]"