Evaluating methods for estimating existential risks.

Researchers and commissions contend that the risk of human extinction is high, but none of these estimates have been based upon a rigorous methodology suitable for estimating existential risks. This article evaluates several methods that could be used to estimate the probability of human extinction. Traditional methods evaluated include: simple elicitation; whole evidence Bayesian; evidential reasoning using imprecise probabilities; and Bayesian networks. Three innovative methods are also considered: influence modeling based on environmental scans; simple elicitation using extinction scenarios as anchors; and computationally intensive possible-worlds modeling. Evaluation criteria include: level of effort required by the probability assessors; level of effort needed to implement the method; ability of each method to model the human extinction event; ability to incorporate scientific estimates of contributory events; transparency of the inputs and outputs; acceptability to the academic community (e.g., with respect to intellectual soundness, familiarity, verisimilitude); credibility and utility of the outputs of the method to the policy community; difficulty of communicating the method's processes and outputs to nonexperts; and accuracy in other contexts. The article concludes by recommending that researchers assess the risks of human extinction by combining these methods.