> But what interests me most about Community Notes is how, despite not being a "crypto project", it might be the closest thing to an instantiation of "crypto values" that we have seen in the mainstream world.
The attempt to argue that any community-consensus system is "crypto" is absurd.
Read it again. He very explicitly says it's not a "crypto project", and is just saying that it feels close to something with "crypto values", which, to Vitalik, seems to mean decentralized mechanisms for building community consensus.
Except this is not about community. Communities are often exclusionary (e.g. "we, who are not those smelly hippy leftists"). This is about credible neutrality, an attempt to soar above tribal politics. Credible neutrality is an important aspect of money. Many (including Vitalik) think that cryptocurrency can't be considered money if it is not credibly neutral.
It seems like we're still at quite early stages of collective intelligence amplification. It would be cool to have a Community Notes for HN. It would be awesome to automatically find consensus on comments and posts that lack attention. I suppose the votes are currently behind a private API on HN. It would be cool to play with sliders for the thresholds and see what different content floats or sinks. It's great to see Vitalik poking around with this sorta stuff, with less than 200 lines of Python code - I'd love to see it vastly simplified. I'm glad he pointed to Polis also recently. Really interesting stuff.
Edit: Actually it occurs to me, you might be able to use ML to infer cross-partisan support from arbitrary sources like HN comments and posts. Especially if trained on Community Notes. Might need to absorb some HN audience specific preferences / biases though.
One option would be to show more of these notes but also make their rating public. For example, “you may find the following note helpful, although only 65% of diverse users did”. But all in all, it’s already a great feature.
In reference to the article's question: "could we turn Community Notes itself into something that's more like an economist algorithm?" My answer would be: no, as soon as you add an economic incentive to a system such as this, you break it, by adding an immediate and direct incentive for abusing and gaming it. FWIW that's why I think the concept of "credible neutrality" applied to money is a fantasy, like a neutral weapon (as soon as it is used it has, by definition, to have taken a side).
I don't think the idea of an "economist algorithm" is to add an economic incentive, but rather to realize that such incentives already exist (of course people want to abuse and game the community notes) and to design a mechanism so that these incentives lead to using the system as intended. (cf. https://xkcd.com/810/ )
Great write-up. I would have loved to be a fly on the wall during the white-boarding, trying to figure out which tensor math would be useful to solve this.
> This is a difficult issue, but ultimately I come down on the side that it is better to let ten misinformative tweets go free than it is to have one tweet covered by a note that judges it unfairly.
That seems strictly worse, to me. If you have a good tweet with a bad note, you have 1 piece of good info and 1 piece of bad info. If you have 10 tweets with bad info and no notes, you have 10 pieces of bad info. Community Notes does not seem to be able to address that issue in its current form as it seems to require notes not be very controversial and, as discussed in the article, many things these days are quite controversial. Or, in other terms, if you want to spread misinformation on Twitter and trivially defeat the lame mitigation that is Community Notes, you just need to make it controversial.
> That seems strictly worse, to me. If you have a good tweet with a bad note, you have 1 piece of good info and 1 piece of bad info. If you have 10 tweets with bad info and no notes, you have 10 pieces of bad info.
In the isolated case, sure, but the point is that in the long-run it's more useful to maintain a non-partisan reputation, otherwise it becomes as useless as fact-checkers are. Bad community notes corrodes public trust in them.
> if you want to spread misinformation on Twitter and trivially defeat the lame mitigation that is Community Notes, you just need to make it controversial.
The goal is that the Notes themselves are not controversial. You can have a non-controversial Community Note attached to a controversial tweet.
You are correct that Community Notes doesn't work if things are controversial, which is why I called it a "lame mitigation" against misinformation.
My point is that non-controversial is not strongly correlated with the truth. So if you want to avoid the truth in a Community Note, like China's treatment of Ugyhurs, you just need to make it controversial so that it is automatically removed by the algorithm.
True statements may not necessarily be non-controversial, but that doesn't mean there aren't less-controversial ways to state the truth.
For example, I think the fact that the Xinjiang United Front published http://web.archive.org/web/20201115143615/http://www.xjtzb.g... and kept it online for three years is less controversial to establish than whether a genocide necessarily involves mass murder, and serves just as well to contextualize the original tweet.
If public users can vote on a note’s usefulness, then you will have brigading. If an algorithm isn’t resistant to that sort of manipulation, which the example in this article leads me to believe is suspect, then I don’t see how this will be particularly useful at this point.
It is resistant to that sort of manipulation. The algorithm looks at users' voting history to see how often their voting is accurate and vice versa (i.e. if they vote something is helpful, does or tend to ultimately get marked as helpful?). The algorithm also looks at how one-sided a note is. The more a note is voted helpful by those who've disagreed in the past, the more likely it will have the helpful status.
Like most of Musk's ideas, this one is "good but doesn't work". When he's got good people around him, they turn that into "good and works!", but when he's left unchecked we get dumb bullshit like Community Notes, which any idiot could see the problem with: Musk himself.
Just, as an example, not casting any aspersions...
If your community is made up of people who either lean right, politically, or are far right, politically, then your Community Notes feature could still "work" in that there would be strong disagreements between whether Trump is a literal God King and whether Biden's policies are actually doing what fiscal conservatives want. You would get a spectrum of beliefs and you could get "corrections" that bring you CLOSER to a truth, maybe (splitting the difference between "empirical" and "some shit you made up" brings you further from truth, but let's press on), but you still won't get any real "fact"-checking. You'll just get a big fat "actually, you're an outcast in this society for this view" sticker on your thoughts.
In the name of fairness, you can replace the above thought experiment constituents with leftists and the result is the same. Notes that urge users toward or further from "defunding all police", or "abolishing the second amendment", etc.
And then, of course, the reality is in the middle-ish. I will make a firm contention, here, that Twitter IS more politically right than left and it shows in the platform's third-party observed metrics, as well as the content of the Community Notes. But I also think that the political classes of twitter are a much smaller portion than, say, the J/KPop stans, the Swifties, or the product people (individual creators selling things, like webcomic artists rather than, say, Tommy Chong's weed gummies). So there's a strong raft of leftist politics, it's just that they aren't the ones who engage in Community Notes. Children, Fans, and the chronically overencumbered do not tend - at least in the metrics I've seen (and screamed as "duh" by my brain) - to engage in policing misinformation. So you end up with a Community Notes that settles on 'rightness' via a right-leaning community consensus, rather than any actual facts. Which is... not the intended effect. Now, granted, it's less biased than other systems I've seen, but that doesn't excuse the amount of bias it ends up displaying, practically.
So the issue is that without the general public, or at least representation for all of its demographics in proportion, you don't get closer to truth, you just get closer to community consensus. It's a utility that works BEST at consolidating viewpoints into mass-appreciated ones, and has a limited tendency toward truth as a side-effect.
So why is the general public not proportionally represented on Twitter? Like I said: Musk. Whether you appreciate it or hate it, his personality is polarizing and a lot of people just won't engage with it. They left Twitter and it surged right. Again, not to it's complete subsumation as a lot of media would "have you believe" (more likely, media leaves out unnecessary details and people run with that to make it feel like it was intentionally dishonest), but it's observably more right, politically. Source: all third party reports, and spending every single day on twitter with 16 separate accounts specifically for monitoring trending content.
So, unfortunately, he built a really cool little thing that does actually work! But it worked better for old twitter, and doesn't particularly work well for getting "facts", now. All types of people have to engage and you only get all types of engagement if you aren't so hyper-focused on your own dumb bullshit.
Instead what he made was the best little shit-stirrer for his echo-chamber. Great, if that's what he was after. But hilarious if he was actually trying to get something approaching "truth".
> So why is the general public not proportionally represented on Twitter? Like I said: Musk. Whether you appreciate it or hate it, his personality is polarizing and a lot of people just won't engage with it. They left Twitter and it surged right.
As best I can tell only a tiny fraction of people actually left twitter without coming back a few months later.
Do you have any numbers on what proportion of leftists left twitter and where they went?
A research team led by Charlotte Chang, assistant professor of biology and environmental analysis at Pomona College, found that in the six months following Musk’s acquisition of Twitter—now known as X—nearly half of Twitter users identified as environmentally oriented had ceased being active on the platform. These active users, which the researchers called “Environmental Twitter,” were defined as posting on the topic at least once in a 15-day period.
Oh, neat! Thanks for posting this! I hadn't seen it before.
I wish there were more actual data, personally, but it's still nice to see unrelated parties seeing similar trends. Especially when they are able to cut out lots of noise, like you can do when focusing on a niche.
Oh, no, nothing empirical! Musk, the "free speech absolutist" is actually very closed-mouth about actual numbers. Loves to tout metrics that no one can validate, but will not, under any circumstance that I've seen, let anyone do any real research on the metrics. It's all just hearsay based on wire monitoring and patchwork investigations.
The best estimate I've heard is that around 20-30k people really left, based on reported numbers over half a million. That seems about right, to me. 500k said they were going to leave, only 20k actually did.
Of course, why that pushed the site more rightward is likewise speculation. If those numbers are accurate, I'd guess that the people who did leave were the most politically or socially aggrieved, so they would be the exact people to have outsized impact on Community Notes. And given the actual impacts that I've observed, anecdotally, that tracks pretty well with how much further right Twitter got. Not "much" but noticeable to anyone not making excuses for it.
I'm one of the 20k, but I never said I would leave. I just removed my account, simple as that. And I'm probably not the only one who left without fanfare. Why would I announce to leave?
There are public resources available; from what I understand Bloomberg did a report on this a little while back? My evidence is anecdotal and only referenced against third-party reports of the same kind of content. There is one that I think is kind of funny, which I'll link to here:
Grapes on the vine bit? Sure. But Jesus turned water into Wine? That's closer to truth? No, it's closer to a well-known consensus. Whether you believe in god, or not, you know the bible story and find it inoffensive (playing the averages with that "you"). So the Community Note stands. But does it correct the invalidities of the tweet? No. Does it do anything meaningful at all, other than brand the tweet as "unaligned with our community's thoughts"? No it doesn't. So why wasn't a better Community Note chosen? Whatever reason you come up with: that reason is part of the failure. Whether it was "for the lulz" or just because it was "palatable", it doesn't matter.
For a more serious example, you can look at the kind of Community Note attached to Jordan Neely's death. How did that note add context? In what was was it more truthful? Does every video of a person need the context that it doesn't include their actual death, or is that only for people who recently died? If so, when should the note be removed? And, honestly, why was it helpful in the first place? Who needed that information? That a bunch of people can agree that a note is "not incorrect" doesn't mean that a note should be applied.
This is the problem with criticizing Community Notes - or anything that actually deals in truth. If you say anything about it, you are attacking something that people have put trust in. Which is to say, you are attacking their trust and self-confidence in said trust. I understand that is frustrating for you, but I'm not passionate about this at all. I would love for Community Notes to be a better idea than it is. I'm sure someone will come up with a way for it to be. But until then, I'm going to call out the complexities of the complications as best I can, without regard to how it will frustrate anyone's trust.
Put simply, if you have a problem with one of my "lots of assertions", we may be able to discuss it directly. But dismissing them all is not a good way to start.
> Even if less than one percent of misinformative tweets get a note providing context or correcting them, Community Notes is still providing an exceedingly valuable service as an educational tool. The goal is not to correct everything; rather, the goal is to remind people that multiple perspectives exist, that certain kinds of posts that look convincing and engaging in isolation are actually quite incorrect, and you, yes you, can often go do a basic internet search to verify that it's incorrect.
The attempt to argue that any community-consensus system is "crypto" is absurd.