A computer model is fine for predictive purposes, but it is not an acceptable substitute for empirical confirmation of a scientific hypothesis.
Gödel's incompleteness theorems alone makes that pretty obvious.
Whether you call it a computer model or a mathematical model, mathematical models can and do confirm scientific hypotheses. The classic example is the 1919 eclipse measured by Eddington to verify General Relativity. Eddington could show that the observations confirmed General Relativity at (something like) the 95% confidence level, by comparing the observations to the predictions of the model.