On problem understanding

Problem understanding is really about asking the question, "What to do?" which is a matter of requirements. Why this question is so important is perhaps obvious. The issue is why is there a tendency to jump to the question, "How to do it?" which is about design. I wonder if we somehow instinctively know how hard the "what" is and simply avoid it. Even when it's written in black and white, it is still possible not to understand what to do. That's why there are courts of law and an entire legal system to interpret what is meant by such-and-such and the experts can still get it wrong. Does that mean we really need to understand the entire problem before we start working on it? I have reasons to believe that is impractical and furthermore, doesn't yield good results. Why? Because often we don't know what to do until we do it, or part of it. Only then the picture may become clearer.