Modeling, Understanding, and Prediction

Sean has raised a couple of interesting points in his comments to Joe's presentation on earthquakes and natural disasters, and Tom and Meridyth's presentation on the outbreak of war. I was going to respond to his response, but, on further thought, believe that it is more of a general topic, and I invite responses.

Sean seems to be comfortable with the fact that models exist for natural disasters, and for the onset of war, but is skeptical (his word) because of the inaccuracy in using these models for prediction. I wonder if this is a general feeling among some, or all, of you? If so, what does this say about the modeling process in these cases?

For me it suggests that the modeling is yielding some deeper, or at least different understanding of the phenomena (disasters/war) as opposed to prediction. This is the complete opposite of how we started the semester - where prediction of random fractal patterns was much easier to come by than understanding why the images were produced.