When dealing with problems of usability, I think it’s still relatively easy to rely on experience and be able to predict if a user will be able to manage through a task. Over time, an experienced designer builds up a number of rules, best practices, or guidelines that can help him or her to make an interface more usable (that is, answering the problem of: can people use it).
When dealing with problems of usefulness (that is, answering the problem of: will people actually use it) I think it becomes harder and harder to predict the answer. Problems of usefulness are of a higher complexity and therefore are more difficult to predict and resolve with just experience or guidelines alone. Instead, in order to improve usefulness of a product, the designer has to get scientific, measure, test, experiment and simulate.
… just a sketched out hypothesis.