YouTube’s Algorithm Incentivizes the Wrong Behavior

>>furcyd+(OP)
It's not complicated.

They need to stop showing people the upvote and view COUNTS. Behind the scenes they can still use it to make recommendations.

Those numbers are pseudo signals of quality to people who encounter content they have never encountered before.

Even when they have doubts that are watching something unhealthy the mind goes "well if the rest of the world thinks this dumbass is important I better pay attention..."

If a dumbass hurting people on video gets 10 million views other dumbasses worldwide automatically get triggered looking at the count. "hey I can do this maybe I should run for President..."

Remove the counts and you remove the pseudo signal of quality.

>>kodz4+Dl
It is complicated. I think thats a bad solution.

I want to see the counts. I feel it is far more transparent to see the counts than for things to just be surfaced or not opaquely. Youtube is not a discussion site and it does not work as one. How popular things are is a part of the context of pop culture, and most youtube content is pop culture.

>>verall+Sm
Every single day, I watch the channel of a guy who has put out < 15 minute videos going back to nearly the founding of YouTube.

He gets an average of 10-15 views per day.

The value this guy adds to my day is literally measurable in $$$.

If I could find more people like him, that would be great, but instead these are my recommendations:

    - 5 ways to do X
    - Bill Gates breaks down blah blah blah
    - Something about Tesla
    - One video by a guy I discovered outside of YouTube who is similar to the guy I watch every day. I don't watch this one that much though.

YouTube's algorithm is not designed for discovery. It's designed for engagement. So I keep separate accounts:

    1. Account for actually useful stuff where YT's recommendations are useless
    2. Account where YT's recommendations are OK: white noise like things. Howard Stern interviews, etc

I wish you could configure the algorithm for discovery somehow.

>>cheez+gq
Of course it is a matter of metrics - it has no way of knowing what is useful. The closest way to algorithmically discover (outcomes over time) would be prone to spurious correlations and be so intrusive it would make Cambridge Analytica look like LavaBit.

>>Nasrud+WC
I'm thinking "make things more discoverable" than "find more useful things" if that makes sense. I'm willing to wade through it myself if you present me with options.

zlacker