http://seedmagazine.com/content/article/on_overconfidence/

An extract:

“We can predict, therefore, that overconfidence is most likely to help in “evolutionarily salient” contexts that are similar to the adaptive challenges of our past, such as individual combat like boxing. It is least likely to help in “evolutionarily novel” contexts—such as sitting at a desk in a bank guessing the behavior of 100 million strangers, or commanding thousands of invisible troops from an underground bunker.

The good news is that evolutionary reasoning suggests ways to avoid the most dangerous situations in which overconfidence is likely to arise. First, we can make sure that overconfidence is nurtured in settings where a go-getting attitude helps performance (such as sports), and suppressed in settings where accurate assessment is more important than will (such as global climate change). Second, we can target the underlying contexts that make or break overconfidence.”