I was reading this quote from Kahneman and Tversky, and found this part very revealing:
Low probabilities, however, are overweighted, and very low probabilities are either overweighted quite grossly or neglected altogether, making the decision weights highly unstable in that region.[page 8, Choices, Values and Frames]
Given this definition, I don't see why this theory is considered useful. If a disaster happens that seems to cause a lot of angst, one can say it was underweighted. If someone is paying a lot for insurance that seems to generate a positive return, one can say people are overestimating this risk. Ex post, the theory explains everything, and nothing. Looking a small probability event before the fact, it predicts everything, and nothing.