You have to load the page to see the dislike count. They already got ad revenue by the time you see it.
I am sure they have better data than anybody as to how often a video is loaded for the first time and disliked within a few seconds of starting.
While once in a while this could be caused by a bad video start it probably mostly isn't caused by the content of the video.
By looking at behavior of users and pattern matching it probably wasn't hard to see that certain users were "making the number big".
Additionally as the article details they likely targeted those users with a new "don't show the number" feature and saw an improvement in the unwanted behavior (unnecessary dislikes).
The reality is that while dislike is fantastic from a quality filter standpoint human behavior likes making numbers big and sometimes they do weird things you have to correct for.
I am sure they have better data than anybody as to how often a video is loaded for the first time and disliked within a few seconds of starting.
While once in a while this could be caused by a bad video start it probably mostly isn't caused by the content of the video.
By looking at behavior of users and pattern matching it probably wasn't hard to see that certain users were "making the number big".
Additionally as the article details they likely targeted those users with a new "don't show the number" feature and saw an improvement in the unwanted behavior (unnecessary dislikes).
The reality is that while dislike is fantastic from a quality filter standpoint human behavior likes making numbers big and sometimes they do weird things you have to correct for.