They performed an experiment to see if removing the dislike button would mitigate brigade voting, or mob voting, especially against small creators. The results

They performed an experiment to see if removing the dislike button would mitigate brigade voting, or mob voting, especially against small creators. The results of their experiment, they say, is that it did address this issue. That sounds like a great result! The problem would be if they did not test any other side effect of removing the dislike button. For example, if the only thing they tested was whether it prevented mob downvoting in a small number of cases, and not whether it negatively affected the experience for a majority of users, the results of the experiment would not mean much. Based on the reaction, I am guessing this what they actually did. (If they published their methodology and results somewhere, I apologize: I looked but did not find anything like this). That's the problem with product research: it's often poorly designed, when it's not just designed to confirm a prior decision and back it up with some data.