What is the correlation between upvoting and benefit to readers of LW?
Do posts with more upvotes actually provide more (real) value to community members? Is there any evidence to support this hypothesis? How this could be measured at all?
Regarding confirmation bias: personally, as a LessWrong member, are you more likely to click on an upvote for posts you agree with or for posts you disagree with? Or to the post interesting, but not questioning your assumptions and your map of the reality? To be honest? Could you count your upvotes/downvotes during last, I don’t know, year, and check the ratio?
And do the posts you initially agree with really make you less wrong?
There are two different criteria for comments (like and agree), but only one for posts (like). Perhaps there is room for improvement here.
Below is the letter from (supposedly) admins (italic emphasis is mine):
P.S. to admins: No, I will not pay attention to what content gets upvoted, sorry. Also I will not “invest in improving my output” specifically to moderators of LW, as I already do it on a constant basis.
Quick note to say that your recent posts and comments haven’t been up to LessWrong’s standard, and I’ve rate limited you to posting once per day. To get a better sense of the site’s norms, I’d recommend paying attention to what content gets upvoted before writing again. Here are some common issues that are worth specific attention (not that any of these necessarily apply to you):
- quality of reasoning
- ignoring of well-known existing arguments or evidence
- poor formatting, punctuation, or grammar
- low clarity
- inappropriate/hostile tone
- appearing on the site over a “pet topic”
PS: If in a few months you have invested in improving your output, you can message us and share evidence of your progress and we can re-evaluate.
Amusingly, I’m not sure whether to upvote or downvote. I’m happy to have (some) discussion of LW signaling and it’s impact on groupthink or limits on contrarian exploration on uncomfortable dimensions. But this doesn’t seem to be that.
Upvotes are not agreement (and I am with you that I’d like to see the agree/disagree option on posts). They’re “I like seeing this kind of thing on LW”. And downvotes are closer to “I wish I hadn’t spent time reading this”. There’s a ton of noise, of course—people don’t have to say WHY they voted, and there’s no oversight or cabal who overrides votes based on some secret meetings. Sometimes votes just mean “I don’t get it”, or “right on!”. They’re a weak signal, but the easiest one to get.
It’s good advice from the admins, if you’re not getting the reception you want, to look for examples of things that DO get the reception you want. Points and votes aren’t the end-goal, but they somewhat correlate with engagement and approval. Unfortunately, there’s no way except trial and error (and posting smaller things, for more targeted trials) and engagement with comments (IMO a much better signal than votes) to find out what really works here.
And just to acknowledge—yes, it’s an imperfect group of people, and there are some topics and styles which just aren’t going to work here. That’s not ideal, perhaps, but it is what it is. You’re free to use LW for the things it works well for, and other sites/groups/activities for the things THEY work well for. I don’t know anyone who exclusively posts on LW.
Thanks. Agreed, different places works better for different topics and styles.
I have checked and can acknowledge that a lot of downvotes are quite uncomfortable psychologically even if you are fully prepared to them and even without explicit harassment.
That could be bad:
1) to people who would like to write about some controversial topics which could be uncomfortable but finally helpful to the community (I’m not talking about myself here, but about more sensitive persons);
2) to the community as a whole and to members of the community finally, as their views less likely to be challenged in current setup;
Interesting that this downvote problem probably will matter less for the prominent members of the community who already have a lot of karma and respect. This could lead to the situation when controversial topics become discussed only at the behest of the community leaders.
I could also point that recent changes at YouTube where you cannot see dislikes now, just their percentage, work really well. They encourage more involvement and do not hinder the incentive to write.
P.S. This comment I cannot write because of the rate limit, so it will be posted later.
P.P.S. I decided to consider this not a bug, but a feature and I will answer max 1 comment per day in the following couple of months in any of my places. It should help in different ways: to choose only most valuable comments to engage with, to not spend too much time in useless discussions etc. Personal lengthy discussions could be done through DM if anyone needs them.
Just a clarification: I didn’t mean that I think the voting was particularly broken or needs fixing, I meant the PEOPLE are imperfect and some topics, themes, and styles won’t work here, regardless of voting or feedback/filtering mechanism. It’s not a “downvote problem”, it’s a “community preference expressed through voting”, which may or may not be a problem. The voting is an indicator, not a cause.
Controversial topics on some dimensions are welcomed, but on other topics are indistinguishable (by the members of the community) from garbage or attacks. The joy and pain of a voting system is that there’s no oversight or leadership who declare whether the masses are “correct” in their judgement of what they want more or less of on the site.
If the people are broken, the system could compensate with better rules. If you tell someone to upvote or downvote on the basis of like/dislike, most of what you are going to see is bias...because people like their opinions confirmed.
I didn’t say “broken”, I said “imperfect”. The system can compensate a little bit, if most of the people want it to. But if it compensates/interferes too much with users’ preferences, the people leave (and different people may or may not join).
There’s no way to make a purely voluntary social chat group to be more accepting of dimensions of disagreement they don’t accept as valid. And I, for one, don’t want that anyway.
If you hold “membership of the group” and “voting according to agreement instead of truth” as fixed, sure. But under these assumptions a rationality promoting group is doomed to failure .
I also didn’t say “voting according to agreement instead of truth”. More “voting according to what we want to read and write about”. I did presume that we don’t have the option (nor, in my part, the desire) to radically change the membership. Though, now that I think about it, that’s isomorphic to someone finding a different group (instead of or in addition to this one) that’s more receptive, and I do support that.
I’m OK with being doomed to failure, if the only definition of success is “every idea, regardless of priors and preferences, gets an equal voice”.
If voting was only about topics , the situation would be much better. But the explicit instruction is like /dislike, not ir/relevant.
Disagree—voting is and should be about what we want on the site, inclusive of relevance, style, respect for prior work, and general tone.
And with that, I’m going to bow out. I’ll read any responses, but probably won’t reply further.
But it will inevitably include “against my biases” ,because that’s disagreement, too.
I can understand why you get downvoted on some of those posts on your profile. Nevertheless, I can understand a reason why you’d be right to criticize the downvoting of those posts. You do have real insights, but I also do think you could benefit from thinking a bit more about how to phrase them. I can’t interpret them easily, and I think that will be common here; but I also think that I could spend some effort to attempt rephrasings of your downvoted points and get massively agreement downvoted without getting karma downvoted. That said, as a person who has a tendency to get in fights online myself, I get karma voted more often than some folks, and I wouldn’t be at all surprised if this comment gets heavily agreement downvoted. I’m the type to occasionally read that one facial expression club that doesn’t like this website, even.
RE: how well does downvoting vs upvoting work—I’d say it has a lot of improvement to be made. Your point in this post is good, and I think right now there are a couple of strong upvotes (one is mine) and a strong downvote, or so.
In my view, downvote should mean “this was badly argued, please delete, rephrase, and try again”. It should be the same kind of thing as peer review rejection due to unreadable papers or missing examples.
Agreement votes should be much, much more clearly indicated as different. They should be doing the job of “I don’t agree, this seems wrong”. perhaps one should have to split their strong vote between agreement and karma. or something.
Both should probably be at the bottom of a comment, to mildly reduce perceptual bias.
Those are just some takes I have. I tend to delete a comment if it gets heavily karma downvoted—I take that to be an actual social rejection. I don’t tend to delete comments that get heavily agreement downvoted, I just move on with life knowing that I said something that was insightful enough that people thought I was the crazy one in the room, which should happen often enough that I sometimes get agreement downvoted.
Shrug. I think overall upvotes are doing a solid job making this a solid shared ai-and-human-safety research notebook website, which is what I want out of it.
Getting downvoted means collectively that content is not wanted. If that is a “bad take” (to reject that content when it is actually valuable) the forum does it at its own peril. There is responcibility in voting.
To try to evaluate whether some piece of content would have been valuable if it had been upvoted is very challenging.
I agree that you should not “karma optimise” or filter your posts, but rather than getting the content through you should be be content with the silence. Original Content that is later discovered is way more valuable than “expectation pulp” that gets the poster into the ingroup.
My voting heuristic is “do I want to encourage this kind of comment”. In most cases this boils down to “upvote if I agree” but sometimes I’ll upvote something I disagree with if I felt that it was valuable.
I didn’t check my history but my conclusion is the opposite of the one I’d get if I was biased so I think I’m okay.
A post can be valuable even if I agreed with it. For example, I agreed with most of the sequences as soon as I read them, but they’ve been extremely valuable to me.
Edit: I weak upvoted the OP because it was interesting enough for me to spend a few minutes thinking about it and responding.
Agreed on sequences example.
For me the most valuable were e.g. that Seneca’s letters, with which I initially disagreed completely, but after several days or weeks of reflection, came to the conclusion that he was right and I was wrong.
Karma is the crudest possible measure of reception, and one that not everyone pays any attention to. Your posts are also getting a poor reception in terms of the replies they receive. What, if anything, you choose to do about that is up to you. What do you want? What is your purpose?
The general theme of your downvoted posts has been “don’t be rational, don’t be utilitarian, don’t be consequentialist, because it doesn’t work.” Unless you have something better to propose, what’s the point? Besides which, your argument is a consequentialist one, and if you’re down on rationality, why are you giving an argument at all? There are ways to think better, and the problems of utilitarianism are well-known, both here and the wider philosophical community.
For example, Eliezer has recommended: “Go three-quarters of the way from deontology to utilitarianism and then stop. You are now in the right place. Stay there at least until you have become a god.” You have recommended nothing, and seem to be unaware of what has already been said here on these topics.