An important thing here is that we’re _not_ wanting “number of people who’ve interact with a post” to be the dominant thing people find themselves optimizing for. That incentivizes content that is some combination of clickbaity, tribal, viral, or minimally-and-unobjectionably-good. i.e. the rest of the internet.
We want to incentivize content that thoughtful people consider and think is real good, and the central thesis here is that the natural impulse to ask “how many people liked a thing” isn’t actually the best guiding star for incentivizing good content (either from a site design standpoint, or from a personal-motivation of individuals standpoint)
Also, it’s kind of weird to me that I have 5 vote power given I’ve only really interacted with this site for… a few months? And you guys have, like, 6? 7? Are you sure your scaling is right here? :/
Well, that’s part of the reason why we’re increasing the scaling somewhat – if Strong Votes capped out around 5, there’s even less gradation between experienced users and new/medium users.
We do want want thoughtful, good writers to have a fairly easy time getting to a voting power that feels midrange, but then room/incentivization to continue to grow. In the 1-5 strong upvote scale, that’d mean getting “3” fairly quickly, and then little gradation between people with 2000 karma and 200,000.
But it’s really confusing for my models of the post.
Cause there is a real difference between (lots of 2-users voted on this vs. a few 5-users voted on this). Those feel very different to me, and I’d adjust my views accordingly as to whether the post was, in fact, successful.
I get that you’re trying to make “lots of 2-users” and “a few 5-users” basically amount to the same value, which is why you’re scaling it this way.
But if a post ACTUALLY only has 2-users upvoting it and almost no 5-users, and other posts have 5-users voting on it but very few 2-users, that seems … worth noting.
Although, you could prob achieve the same by publishing an analysis of upvote/downvote patterns.
You could, for instance, release a list of posts, ranked by various such metrics. (Ratio of low:high user votes. Ratio of high:low user votes. Etc. Etc.)
Although, you could prob achieve the same by publishing an analysis of upvote/downvote patterns.
You could, for instance, release a list of posts, ranked by various such metrics. (Ratio of low:high user votes. Ratio of high:low user votes. Etc. Etc.)
This seems neat and probably worth doing, although this seems like even more interpretive effort than “mouseover to see how many votes of each type you got.”
Not sure how it’s a one-time cost? I was assuming the list of posts only comes out every so often, so every time you want to know the results for a new post you have to wait for such a list and then check it, and then you’d have to check it again for each new post you’ve written.
(If you were imagining the list-of-posts getting continuously updated, that doesn’t seem much different than simply providing the “number of votes (upvotes?) metadata on hoverover, in addition to or instead of the number of strong upvotes”. And while we could do either of those things, the main thing I was getting at is “we’re hesitant to make the goodhartable number the most easily accessible one.”)
I was assuming the list comes out once → I learn enough to understand what types of posts get what voting patterns (or, I learn that the data doesn’t actually tell me very much, which might be more likely), but after that I don’t need any more lists of posts.
I don’t care if it has my own posts on it, really. I care more about ‘the general pattern’ or something, and I imagine I can either get that from one such list, or I’ll figure out I just won’t get it (because the data doesn’t have discernible patterns / it’s too noisy).
An important thing here is that we’re _not_ wanting “number of people who’ve interact with a post” to be the dominant thing people find themselves optimizing for. That incentivizes content that is some combination of clickbaity, tribal, viral, or minimally-and-unobjectionably-good. i.e. the rest of the internet.
We want to incentivize content that thoughtful people consider and think is real good, and the central thesis here is that the natural impulse to ask “how many people liked a thing” isn’t actually the best guiding star for incentivizing good content (either from a site design standpoint, or from a personal-motivation of individuals standpoint)
Well, that’s part of the reason why we’re increasing the scaling somewhat – if Strong Votes capped out around 5, there’s even less gradation between experienced users and new/medium users.
We do want want thoughtful, good writers to have a fairly easy time getting to a voting power that feels midrange, but then room/incentivization to continue to grow. In the 1-5 strong upvote scale, that’d mean getting “3” fairly quickly, and then little gradation between people with 2000 karma and 200,000.
That makes sense.
But it’s really confusing for my models of the post.
Cause there is a real difference between (lots of 2-users voted on this vs. a few 5-users voted on this). Those feel very different to me, and I’d adjust my views accordingly as to whether the post was, in fact, successful.
I get that you’re trying to make “lots of 2-users” and “a few 5-users” basically amount to the same value, which is why you’re scaling it this way.
But if a post ACTUALLY only has 2-users upvoting it and almost no 5-users, and other posts have 5-users voting on it but very few 2-users, that seems … worth noting.
Although, you could prob achieve the same by publishing an analysis of upvote/downvote patterns.
You could, for instance, release a list of posts, ranked by various such metrics. (Ratio of low:high user votes. Ratio of high:low user votes. Etc. Etc.)
That would be interesting!
This seems neat and probably worth doing, although this seems like even more interpretive effort than “mouseover to see how many votes of each type you got.”
I prefer the one-time cost vs the many-time cost.
Not sure how it’s a one-time cost? I was assuming the list of posts only comes out every so often, so every time you want to know the results for a new post you have to wait for such a list and then check it, and then you’d have to check it again for each new post you’ve written.
(If you were imagining the list-of-posts getting continuously updated, that doesn’t seem much different than simply providing the “number of votes (upvotes?) metadata on hoverover, in addition to or instead of the number of strong upvotes”. And while we could do either of those things, the main thing I was getting at is “we’re hesitant to make the goodhartable number the most easily accessible one.”)
I was assuming the list comes out once → I learn enough to understand what types of posts get what voting patterns (or, I learn that the data doesn’t actually tell me very much, which might be more likely), but after that I don’t need any more lists of posts.
I don’t care if it has my own posts on it, really. I care more about ‘the general pattern’ or something, and I imagine I can either get that from one such list, or I’ll figure out I just won’t get it (because the data doesn’t have discernible patterns / it’s too noisy).
Oooohh gotcha that makes total sense.