Overconfidence, in the formal language of heuristics and biases research, means an overinflated conception of the accuracy of one's model of the world, probability estimates, and a diminished ability to update beliefs sufficiently in response to new information1. Numerous studies in cognitive psychology have confirmed that overconfidence is a universal human bias, present even in professionals such as statisticians who have been trained to avoid it2. Overconfidence is associated with other cognitive tendencies such as egocentric biases, which have adaptive value in an evolutionary context3. An overconfident person is more likely to acquire status in a tribe and to attract high-value mates, passing on their genes and producing offspring with the same cognitive tendencies. This shows why overconfidence tends to persist in populations over time despite its downsides.
In the context of global risks, overconfidence can manifest in a number of ways. The most notable is to fixate on a preferred future, rather than considering a probability distribution over many possible futures4, and to ignore data contradicting that preferred future. The tendency to selectively interpret information to suit one's biases is called confirmation bias, which overlaps with overconfidence. One possible way of overcoming overconfidence is to solicit the opinions of others regarding possible futures and seriously consider these alternatives before building a more conclusive model. This is called “taking the outside view,” outside view being a term from the field of reference class forecasting, in which predictions are made using a reference class of roughly similar cases.