3 ms·
I see a lot of comments here (and some in the YouTube recommendation thread currently on the front page) suggesting sometimes a Rube Goldberg-esque series of be
by cgriswald 6y ago
I see a lot of comments here (and some in the YouTube recommendation thread currently on the front page) suggesting sometimes a Rube Goldberg-esque series of behaviors to trick the algorithm into performing properly. I see this mentioned as "easy" and "trivial". I consider it "abrasive" and "tedious." The algorithms should work through normal use of the app. I shouldn't have to tell an app "Ignore this one-off piece of content I reached through an external link that is different from all the other content I've enjoyed and especially different than what I've searched." If I get to the point where I have to curate at that level of detail, what's the point?
It's less work to do what I've been doing all along: get recommendations from friends, family, articles, collabs, biographies, etc. (And it's a lot more fun.)
- Avamander 6y ago> I shouldn't have to tell an app "Ignore this one-off piece of content I reached through an external link that is different from all the other content I've enjoyed and especially different than what I've searched. Do you _really_ think they wouldn't've already implemented it already if it were possible in a reasonable way?
- farias0 6y agoActually I'd bet on the opposite: They don't think it makes sense to ignore it. They want to process this data, and they're confident in the weight it's assigned to it.