What about treating the collapse as the Bayesian probabilities update in light of the measurement (new evidence)?