I’m sure that what we want to eat is a good guide to what we should eat, so long as you ask what our preferences would have led us to eat 100,000 years ago — before we killed off the woolly mammoths. (Curiously, I’ve never seen this obvious idea in any nutrition text.) A vast amount of trial and error is embodied in those preferences. Because we learn to like foods, our best guide to unlearned preferences may be what children want to eat.
The great essayist George Trow doesn’t quite get it, I’m afraid:
In the New History, the preferences of a child carried as much weight as the preferences of an adult, so the refining of preferences was subtracted from what it was necessary for a man to learn to do.