First of all, we don’t want policymakers to assume the worst about the future when shaping economic policy. Nor do we want them to assume the best. Actually, we don’t want them to assume anything. Instead, we want them to come up with the most accurate forecast of the future they can, and to adopt the economic policies that make the most sense given that forecast.
Unfortunately, it's more complicated than that. Surowiecki's proposal would have us ignore unlikely-but-still-possible events -- "black swans," as we're now learning to call them -- because they don't fit into our "most accurate forecast of the future."

Rather, we want politicians to forecast the entire range of reasonable possibilities, estimate the probability of each of those forecasts coming to pass, and then adopt the economic policies that best optimize outcomes across that range of possibilities, taking into account the estimated probabilities and constantly adjusting the forecasts, the probability estimates, and the policies themselves in the light of new events and information. Plus also we'd like a pony.

Update: One might imagine that Surowiecki would be too busy to post a response in the comments section of this post. One would be wrong. In case it's not clear: the difference here is semantic and hinges on the definition of "forecast," which I had taken to mean "single predicted outcome" where J.S. was using it to to mean "predicted range of possible outcomes."