And yet, counterintuitively, it's in precisely these areas (relativity, quantum physics) that our theories are best (most predictive with the strongest mathematical foundations) and the areas where our human intuition ought to supply the most assistance (psychology, sociology) where we seem to be lost at sea.
One explanation that I've come up with to explain this apparent paradox is that while humans may not have any special insight or intuition into microscopic or macroscopic phenomenon, at least we don't have any biases either. Thus, we are able to make progress in these areas by simply pouring in huge amounts of time and brainpower.
On the other hand, for subjects that are more "human," we come pre-equipped with a large number of instinctive insights; or to put it another way, we're burdened with a huge number of innate biases. These are mostly shortcut heuristics that only vaguely approximate the truth, and are deeply and unavoidable biased in ways that we ourselves are blind to. Thus, no matter how much time and brainpower we pour into these subjects, progress remains slow and theories remain poor.
It's a well-known theorem in machine learning that an ensemble of weak (only slightly better than random chance) but unbiased learners will eventually converge to a strong learner, but a collection of biased learners will never become unbiased unless their biases are all uncorrelated. In humans, of course, the biases are shared across all members of the species so do not cancel out in this way.
One explanation that I've come up with to explain this apparent paradox is that while humans may not have any special insight or intuition into microscopic or macroscopic phenomenon, at least we don't have any biases either. Thus, we are able to make progress in these areas by simply pouring in huge amounts of time and brainpower.
On the other hand, for subjects that are more "human," we come pre-equipped with a large number of instinctive insights; or to put it another way, we're burdened with a huge number of innate biases. These are mostly shortcut heuristics that only vaguely approximate the truth, and are deeply and unavoidable biased in ways that we ourselves are blind to. Thus, no matter how much time and brainpower we pour into these subjects, progress remains slow and theories remain poor.
It's a well-known theorem in machine learning that an ensemble of weak (only slightly better than random chance) but unbiased learners will eventually converge to a strong learner, but a collection of biased learners will never become unbiased unless their biases are all uncorrelated. In humans, of course, the biases are shared across all members of the species so do not cancel out in this way.