Designing adaptive algorithms that cope with unanticipated changes in a self-adaptive system, the environment in which it is situated, or the goals that drive its behavior is a difficult and important challenge. The considerable attention that has been devoted to this challenge by the research community is fully warranted. However, I see such work as a pre-requisite for the ultimate challenge for self-adaptive systems: eliciting the goals that drive their behavior. I explain why this is both a fundamentally important problem and a highly challenging one, and urge the research community to start tackling it in earnest.