So you'd rather they manipulate the data and potentially mislead users? Sounds like a slippery slope to me \s
[1]: https://81m.org/
Archive: https://web.archive.org/web/20211023061808/https://81m.org/
Say for example I used a batch of 100 HN accounts to mass downvote your comment. After a while an hourly job runs that looks at all those votes in aggregate and determines they were coordinated (for simplicity sake they all came from the same IP) and removes them. You'd see a large shift in the net score of your comment all of the sudden. This isn't HN manipulating anything, it is them doing their job to prevent abuse on the platform.
I think that unintentionally removes down votes on obviously bad videos that people do not waste much time watching.
For instance, an hour-long video claiming to be what you're looking for, but actually consisting of a still image and a URL to a scam/spam site.
I've seen dislike attacks happen. And if it takes a non-trivial procedure to make them count, people will document that procedure. For instance, hypothetically:
"Alright everybody, here's the link. Remember, mute the tab right away but don't mute the video itself; wait 2 minutes (the timer that starts when you click the link will go green to let you know), then click the dislike button."
I've seen much more complex instructions offered as part of gaming a poll, as well as sites built to help semi-automate or simplify the process.
The key distinction that would be useful for YouTube to measure: did you encounter the video and then dislike it, or did you visit a video you were referred to for the sole purpose of disliking it? I don't think hiding dislike counts serves that purpose, though.