Here's a neat idea, and I'll jot it down even though it's not very polished. (I think I'll start doing that more here; seems a good place for it.)
Emotions offer quick and sloppy reactions. You see something that looks like a snake, you run away; only later do you realize it's just a hose. You could sit down and concentrate and take in all the details of the scene and mentally label everything and say "that is a hose", but that takes too much time. Emotions are quick at the cost of being precisely correct.
Similar things are going on in the computing world. You hear about how SQL is getting used less and new sweet techniques (BigTable here at Google, for example) are taking its place. They're not always 100% correct, for some value of correct; for example, when you search for something, it doesn't tell you the exact number of pages in the Internet that contain that word. It gives an approximation, which is fine, because the number is about 12 billion anyway.
When I'm playing Dominion, there are 16 cards, and I don't want to exhaustively research every possibility every turn; that'd take forever. So I pick a card that "feels" good. Even more when I'm playing a long match of rock-paper-scissors, I have to just go based on gut feeling. It's not 100% right, and sometimes I make mistakes (playing rock against someone who always plays paper, for example), but it is quick.
So what if we could train our emotions better? You'd get the kind of tradeoff you get between trying to use SQL for everything and using new better storage solutions. It's like the difference between Google search and, well, some slower search engine. Fractions of a second? Totally worth it. Instead of trying to put them aside and compute exactly, take advantage of them and learn to make better quick calculations.