I suspect their dataset was direct link dislikes. I.e. people who arrived at YouTube from another site at a specific video who dislike the video. That's the characteristic of a coordinated dislike campaign.
> the feature feeling useless, as even if you dislike, you might thing that you click will worth nothing as no one will see it
That was exactly the point. They didn't want the attack mobs seeing the fruits of their labor, making it feel useless, making fewer people mob.
I honestly don't see a problem. If someone truly dislikes a video, they put in effort instead of encouraging a mob mentality by a high number of dislikes
Maybe Youtube can somehow incorporate the dislikes into "Likely to enjoy" score similar to how its set up on Google maps. It's pretty accurate from my experience.
> the feature feeling useless, as even if you dislike, you might thing that you click will worth nothing as no one will see it
That was exactly the point. They didn't want the attack mobs seeing the fruits of their labor, making it feel useless, making fewer people mob.