Bottom-up learning is all about trial-and-error. What’s useful about that? Simple: feedback. Direct, meaningful, applicable feedback. Top-down learning is about theory or expert-mediated knowledge — often without feedback. What’s useful about that? Well, you don’t have to try it — so it is more efficient. (Sometimes we don’t want the ‘error’ part of trial-and-error. Some errors are fatal, others are just time-consuming and really annoying… Others are no big deal! Those are the kinds of errors we want to make!)
If you think about it, there are only a few ways of answering the question “What happens when…?” If a child wants to know what happens when she sticks her finger into a pot of boiling water, there are 2 main ways of finding out: 1) try it, 2) ask someone. The first is obviously bottom-up: you try it (practical work), it hurts (feedback), you don’t do it again (learning). The learning is based on her ‘experience-of-what-happens-when-I-put-my-finger-in-a-pot-of-boiling-water’. The second is top-down: you accept someone’s answer (top-down work, like listening to a lecture), and you presumably won’t stick your finger in the pot (learning). This learning is based on a ‘theory-of-what-happens-when-one-puts-one’s-finger-in-a-pot-of-boiling-water’. These are two different kinds of learning, but both will get you the right answer on the test (i.e., different learning process, same knowledge).
Ideally, top-down knowledge is based on bottom-up knowledge; you learn from the experience of others, and as long as that experience is credible, you are good to go. In the real-world, a lot of our knowledge is based on some combination of the top-down and bottom-up learning. Generally, we triangulate expert knowledge with experience, and voilà! We don’t have to stick our finger in a pot of boiling oil because 1) mom says that’s bad (top-down, expert knowledge, probably credible), and 2) getting burned by a little splatter of very hot water hurts enough (bottom-up, direct experience and feedback). So the combination of evidence is very convincing and we don’t feel the need to go all in.
The real difference between bottom-up and top-down learning is in the feedback
Bottom-up learning gives you direct, meaningful, applied knowledge about your question (“What happens when…”). Top-down learning gives you indirect (theoretical) knowledge about the question — often with no feedback. Top-down learning actually sometimes gives you direct feedback about other things. In this example, one might collect that “Mom becomes hysterical when I ask about putting my finger in boiling water” or “Dad will yank me away from the stove when I ask about the boiling water”. Useful information; but not directly related to the question at hand. Teaches you more about your parents than about boiling water, really.
Top-down learning depends on the credibility (or expert-status) of the person answering the question. After all, if you think that person doesn’t know better than you, then you might reject top-down knowledge (e.g., “My little sis says don’t do it, but she’s no expert on boiling water! I’ll try it!”).