When is it appropriate to combine expert judgments?

A recent study by Titus and Narayanan (in this issue) is a remarkably ambitious attempt to perform an end-to-end evaluation of the probability of sea level rise. At various steps in their analysis the authors incorporate the elicited judgment of experts including, in some cases, both parameter and model uncertainty. They then average the expert judgments to produce a single valued probability distribution for future sea level rise. The authors acknowledge the methodological difficulties associated with combining expert judgment, but argue implicitly that the end justifies the means.