Upvotes, knowledge, stocks, and flows
This is a short post inspired by “EXEMPLIFYING EQUITABLE GROWTH: MR. GOOGLE SERVES ME A BAKER’S HALF-DOZEN FROM THE WCEG WEBSITE, AND WHAT I LEARN THEREBY...”. Delong looks at google results for his blog archive, and finds they are very weakly predictive of what he thinks is useful about his blog. He thinks about that in terms of the “flow” of intellectual discussion — what people are talking about — versus the “stock” — what is known.
I think that’s a productive analytical frame. Websites live and die off of their salience to the flow, but their long-term value is what they contribute to the stock. Twitter is optimized for flow to the point where it’s nearly useless as stock; but high-quality work that never gets into the flow ends up so unread and hard to find that it has little value as stock.
For sites like Less Wrong, I think that thinking about stock and flow could help to improve how upvotes work. To some extent, upvotes help promote a healthy flow, which is an important and useful goal. But that is essentially a short-term value, and so upvotes like that should decay. On the other hand, upvotes also serve as long-term markers of valuable stock. That kind of upvote shouldn’t decay.
One place this kind of thinking could lead would be to having different kinds of upvotes. “Thanks” and “good question” would be more flow-y and thus decay faster; “well said” and “good link” would be more stock-y and last longer.
I think that asking people to differentiate their votes in this way is too high a complexity cost and would discourage engagement, but I strongly agree about the importance of stock vs. flow, and the long term value being in the stock.
When I’m doing a post that I hope will become valuable stock, I try to do extra editing passes, and avoid things that will tie it to a particular time, and I prioritize doing such things.
I don’t downvote things for ‘this is only flow and not stock’ but I definitely have a much higher bar before I upvote flow, wish others did the same, and sometimes consider downvoting such things because they get too many upvotes for my taste; perhaps doing this after a week or two would accomplish what we want—if one is archive diving you can make such adjustments.
Another question is, there are often-referenced posts from the past that are clearly valuable stock and have pretty low scores—many important sequence posts for example. Is there a good way to encourage people to upvote them, to allow them to be better found?
FYI, I plan to implement small-upvote and big-upvote differentiation in the near future (like, next week).
Where, basically by default you give “small upvotes” (which range from 1-3, with most users having small-vote-power of 2). And then there is a slight trivial inconvenience-hidden “big upvote” option, which ranges from 1-12, with most longtime users having a value of around 6.
In the flow/stock paradigm, small upvotes are basically for flow, big upvotes for stock.
I also still lean towards using “Facebook React Style” upvotes for differentiation (i.e. after upvoting, you have the option of adding a ‘why’ to the upvote, with a goal of finding something like 5 most common reasons people upvote things)
I don’t think this is adding much user burden, because it’s optional – the thing it’s aiming to modify is not “the part where you upvote someone” but “the part where you give a short comment saying why you upvoted something that doesn’t really add much to the conversation.”
Maybe references could be the votes. Like, if someone links an old post by someone else in their post or comment, every upvote on the new post could turn into “stock” votes on the old post, or something.
I think we should take a page from PageRank, and show highly-linked-to posts.
Wouldn’t it be interesting to know which posts have been most linked in other posts?
Yeah. Not 100% sure how to implement this but I’ve been thinking about this general principle and expect it to be important in some fashion. Getting linked (in particular by highly upvoted posts seems like it should contribute to post karma).
One important question is “what is the mechanism by which we want people to be finding old posts in the first place?”
Old stock posts that have low karma (because they came before the current inflationary period where karma got multiplied by 3). But this doesn’t really come into play at the moment AFAICT, since we don’t actually have a view for looking at old posts with high karma.
It does seem like we should probably have such a view – we removed the Top Posts view when we re-organized the frontpage because a) almost nobody was using it and b) the frontpage filtering options were getting somewhat convoluted and we were trying to keep them simple.
I do expect us to revisit this at some point, although I think it’s worth putting some thought into what we’re actually trying to accomplish here. If there’s important Stock posts that people haven’t read, it may make more sense to have them included in sequences that are promoted or something rather than a posts-view that’s basically “here read all the most important stock posts in random order.”
*cough* https://www.greaterwrong.com/archive
Ah, yeah. (Aside: someone on intercom was asking how to view all posts in order, and I sent them a link to this page because I didn’t know about the archive option, which upon reflection is ‘right there’, but also sort of tucked out of the way and I didn’t notice it)
We need silos for posts to go into. Filtered by tag, filtered by section or somehow clustered.
We need people to volunteer to curate a cluster on a topic, and we need to have a collection of curated clusters, i.e. on the wiki. Some posts can stay in the history books but we should aim to organise the backlog into some kind of accessible structure.
(I might be more help with suggesting structures if I knew just how many posts we were dealing with in the archives)
I’m not sure about this. The sequences are long. Starting a sequence sounds like signing up for a slog. But if there’s a list that’s like, “Here are the core ideas (posts) that are most referenced in other posts, in sorted order.” Then someone can bite off just a little piece at a time. And it makes it easier to 80-20 reading (or reviewing) the sequences.
I guess depends on how many dependencies the posts have. Some posts actually come with prerequisites, and it may make more sense to just read them in order.
Let the reader decide if they need the prereq!
Sure, but there comes a point as you’re backchaining through posts that I think it makes more sense to go ‘okay, it turns out there’s like 10 (or 100) of these things and it makes way more sense to read them in order.’
Agree. Except that what we do is add all the variables together and have one voting button.
You could say that spelling is very important and should be a separate vote button. But people have a rule to down vote on bad spelling and leave good spelling at zero.
Stock and flow should be good on a good post. Occasionally one is higher and the other is low. But excellent posts have both.