How could a study find what an algorithm for personalization does without actually trying it with various people? Number of post views is not a relevant metric.

How could a study find what an algorithm for personalization does without actually trying it with various people? Number of post views is not a relevant metric. Musk's number of followers has risen significantly in that time frame. Of course Musk's post views rise as people who are not fans leave, his fans come, and there's not enough other content to show those who stay. Where's any proof of the claim that Twitter adjusted the algorithm? Do the authors seriously think this weak temporal corelation is enough? Do they actually think this is good science, or just bank on the current Musk hate? (I'm not a Musk fan either, my own Twitter algorithm didn't show me a single Musk post since forever. And it's smart enough to keep showing me SpaceX stuff.)