matt comments on Human inability to assign numerical probabilities

[matt]

But attaching those estimates is clearly useful.
Consider training: predictionbook.com