To enhance learning, humans and other moving things have the capacity not just to learn from what happened but what might have happened. Gambling plays on our tendency to over-estimate what might have happened, drawing us in to play again even when the odds are still against us.
My friend Rob's little brother came up from North Carolina to visit him for the weekend. As they bounced around on Saturday, Rob suddenly "noticed" that the Massachusetts lottery was worth tens of millions and suggested that they buy a ticket. The next morning, Rob got up before his brother, opened the Sunday paper to find the winning numbers, ran out to buy a ticket with exactly those numbers, swapped it out for the one they had purchased the day before, and then waited for his brother to wake up.
As his brother ate breakfast, he read through the paper. After awhile, Rob casually suggested that his brother grab the ticket and compare it to the winning numbers printed in the paper. As you might imagine, his little brother was, er, kind of excited about the fact they'd won millions. In fact, after Rob told him what he'd done, his little brother was still insistent that he really did have the winning ticket.
Printing lottery ticket winning numbers is, of course, necessary in order to find the winners. But it also encourages people to "learn" how close they came to matching the numbers. ("4! I had 5. I was so close!") It might just encourage people to play again, in spite of the terrible odds.
As it turns out, multivariate equations armed only with real data do a better job of diagnosing patients than doctors or even doctors armed with these same equations. One reason is that doctors too quickly converge on a diagnosis and tend to ignore contravening data. Perhaps another reason is that doctors too easily invoke what might have been (or, what could be) scenarios.
This is, it seems to me, one of the problems of learning from policy. Facts can be disregarded because people who are bought into a particular ideology are able to construct what-if scenarios that demonstrate - at least to them - how this could have gone well if only. Marxists are still gaining adherents in universities and neocons are still finding supporters for invasions in the Middle East (Iran instead of Iraq this time). Not because the empirical evidence has suggested that these are wise moves but, instead, because of a kind of imaginary nostalgia, reminiscing about what might have been.
One of the most promising things about voting for change is that it suggests doing away with nostalgia and beginning, instead, with data. I, for one, have my fingers crossed that we can get past reliance on silly superstitions and nostalgia, relying instead on data.