Bring it back to the scientific method. The way to differentiate between competing hypotheses is by devising experiments that falsify some of them, and then running the experiments and either falsifying or failing to falsify them.
The issue here is subtle, and it's that the most popular interpretation, Copenhagen, isn't a complete theory because it doesn't tell you algorithmically when collapse occurs. For any possible algorithmic way to handle collapse, there's a corresponding experiment that could (at least in theory) differentiate between Copenhagen and Many Worlds. But the Copenhagen is inordinately slippery in that collapse is defined to occur ex post facto in whatever way is needed to make the experimental results match the theoretical results.
It's perhaps not so surprising that this shortcoming was overlooked in the beginning because Copenhagen was hypothesized before we really had a clear handle on the study of algorithms. But the fact that Copenhagen is still as popular as it is means that Yudkowski needs to spend a lot of time on philosophy of science, because that's what's holding back most people from seeing the problems with Copenhagen, and why at first glance it looks like philosophy.