Why “reduce to”, not the other way around?

This action confuses me for a while. As you know, once we come up with a problem and we want to know how hard it is, we will do the "reduction". Intuitively, if we prove problem B is as hard as problem A, we can know the complexity boundary of B. But if problem B is harder than A, why we call the reduction process -- reduce problem A to problem B, not reduce problem B to A?