Friday, July 17, 2009
Why computer models are not evidence?
Computer models are sophisticated, put together by experts, vetted by peer review and getting better all the time. But even if they can predict (some) of the (physical) evidence on record, even if they are based on solid proven theories, they still don't count as evidence. They (may) predict (some) of the (experimental) evidence and are useful to predict physical behavior for (similar) situations, but their predictions are only extrapolations of the available (real) evidence that was used to support the scores of assumptions used to formulate them. The predictions (i.e., extrapolations) are only valid insofar as there is physical evidence to support them. This is not to say computer models are not useful to mankind, but to say that a computer model can generate evidence of physical behavior is, to put it mildly, "an unsupported extrapolation."