Hacker News Reader

Comment by Wowfunhappy | Hacker News Reader

The trouble is, manual moderating systems are also a no-go at scale. So what do you do?

We don't need scale. Google needs YouTube to be the only place to watch videos out there. I don't. Smaller communities self-moderate fine enough, and bigger ones can raise money from members to hire mods / editors / etc.

But this isn't working as well as it should because the platform / advertising oligopoly is eating up everyone's revenue and using those funds to vertically and horizontally integrate everything under a handful of megacorp brands.

For as long as this consolidation (aka "scale") continues, content moderation will remain "unsolvable".

> We don't need scale. Google needs YouTube to be the only place to watch videos out there. I don't. Smaller communities self-moderate fine enough, and bigger ones can raise money from members to hire mods / editors / etc.

The thing is, you still have the scale problem even if you break up Youtube into a hundred different video streaming sites.

Say that Google would need to hire a million full-time moderators (in practice it's probably more) to ensure quality moderation for every video uploaded. If it's not sustainable for them to do that, then it won't be sustainable for a hundred companies to hire ten thousand moderators each to do the job.

This is especially bad because bad actors can just post the same content on multiple platforms automatically. Unless platforms start pooling their moderation resources (in which case you're back to square one), having lots of small-scale platforms means that attackers can just upload the same thing on as many platforms as they can and hope it slip through the net in a few cases.

The only way we can stop that, as a society, is if we decide that any platform for content diffusion needs to factor in and charge the price of filtering that content from the get go. That means no more free uploads on Youtube; if you wanna upload something you have to pay a fee for moderators to look over your video and certify it as safe.

I think that would be a viable way for the ecosystem to go, but it would make it harder for small independent creators with no budget to get started.

> The thing is, you still have the scale problem even if you break up Youtube into a hundred different video streaming sites.

No, PeerTube is a decentralized video platform with independent servers, which can work together.

How do independent servers solve the problem of scaling moderation? Every video still has to get reviewed by somebody, so the total number of reviews done doesn't change.

No single company must pay for the army of reviewers. Different servers compete for good content, i.e. have incentive for good moderation at low scale, which is possible.

Okay, that's the theory, but in practice, how good are Peertube servers at, say, filtering out djihadist beheading videos?

EDIT: A quick search shows that I was asking the wrong question; in practice, there are servers dedicated to exclusively these kinds of videos.

> Say that Google would need to hire a million full-time moderators (in practice it's probably more)

You can't use made-up numbers to prove your point. Approx 800,000 hours of video are uploaded to Youtube every day. Watching all of those in full at normal speed would require only 100,000 full time mods, but that's just about the dumbest way to go about it, so in reality it will be much less.

> it won't be sustainable for a hundred companies

> bad actors can just post the same content on multiple platforms automatically

You'll have to reconcile this fatalism with the myriad of smaller communities managing spam and moderation just fine. Millions (not "hundreds") of topical forums (incl. independent ones, and on platforms like reddit) are doing fine, often without full time or paid moderators.

> if you wanna upload something you have to pay a fee for moderators to look over your video and certify it as safe.

Since no one will opt-in to that willingly while free alternatives exist, you're essentially proposing state-mandated censorship. No thanks. I'd rather Facebook / Youtube just die if their consolidated asses can't keep up with what they created.

Shrink "at scale" until the company can handle the responsiblity. There's no inherent right to grow endlessly, and limits can be legislated if FAANG and their "best of the best" employees can't (or don't want to) figure it out.

This isn't an excuse.

e.g. Imagine the argument was "Disposing of chemicals safely is a no-go at scale"

Businesses don't get to ignore negative externalities just because it's unprofitable

"Disposing chemicals" doesn't feel like the right analogy. That's a negative externality that wouldn't exist without the manufacturer -- if the manufacturer doesn't exist, then that problem goes away. "My video is not available on this platform" is different, because if the platform doesn't exist, you still have that problem.

> The problem is, manual moderating systems are also a no-go at scale. So what do you do?

Is it? I'd have imagined there's some ratio of 1 mod per N people for N not too small that would make things work. It might be more like 1:1K instead of the 1:1M (or whatever) that companies prefer, but is that genuinely unaffordable or is it just that companies don't like to pay for moderators?

Imho it cannot work because we dont globally agree whats okay to do and say.

I think mass media has some fundamental flaws that we havent accepted yet.

Edit: take out advertising in intellectual property and you would get rid of a bunch of bad actors and cobra effects.

I think your points are valid. But I'm not sure if I see them as moderation issues. As I see it, there's the question of what should be allowed, and there's the question of how to enforce it adequately. As I see it, the former is a policy issue (and likely tougher), whereas the latter is a moderation issue.

I 100% agree that moderation could be lightyears better.

However i see this as being a bit like the BBC problem: beleiving there is such a thing as being objective, there just isnt.

You can have a group of likeminded people who mostly agree and need a bit of moderation for when someones external bad day leaks into their interactions on the plaform. However if there is an underlying fundamental difference of opinion or goals then the moderators become just another weapon.

We need leaders not moderators. Moderators in difference of opinion just end up punishing the weak one way or another. Leaders show by example what we can aspire to, which in turn guides moderators in how to apply the rules.

If you have global platforms who are the leaders? What direction should they lead that works for china and the US?

Edit: to be clear if you cannot afford enough moderators to stop child abuse or murder videos being posted then shut your platform down as its not viable. I am more tlaking about how moderation gets weaponised etc.

Facebook would need to hire around 2.489 million moderators to meet a 1:1k ratio. Even at 1:10k, you’re looking at increasing the size of the company six-fold.

I gave 1:1K as just a random number, I don't know how much it would actually need. I'd expect it to be much lower. Maybe like 1:50k or less.

What exactly is the problem you're pointing to? Are you saying Facebook couldn't afford to pay that many mods? Note that they don't have to be employees or in Facebook offices...

Because to truly moderate YouYube, you have to actually watch the videos, and YouTube says they get 500 hours of video uploaded every minute[1], it's pretty easy to estimate the staffing it would take each video once. They're basically getting 30,000 minutes of video every minute, so if they have 30,000 people watching it, 24/7, it basically works. You need 4.2 fourty hour weeks to cover a whole week, so 126,000 people will do it. They've also report 2 billion monthly users, so around 1:16k moderator:user would get full coverage of videos; plus another team for comments. And of course a team of 1/8th of a million won't manage itself, and you need an escalation path, and non-robot moderator are actually not robots, so they won't (and shouldn't) spend their entire worktime watching.

126,000 workers actually seems reasonable to find, the 2 million you'd need at a 1:1k ratio would be hard, I think. The pay is going to be bad, and the garbage you have to watch won't be worth it for most, but you still need a reasonably analytical person to decide if things definitely follow the rules, definitely don't follow the rules, or need to be escalated. And they need to be attentive, because videos can be totally normal and fine and then get abusive later.

[1] https://blog.youtube/news-and-events/youtube-at-15-my-person...

The notion that adequate moderation requires watching every second of video at 1:1 speed is such a strawman on so many levels I don't even know how to respond.

Exactly what I was thinking. There are obviously significantly more effective ways to do things. You could probably just keep the automated systems in place, but have human moderators verify every video that the automated moderators flag (as you stated, obviously not by watching the entire video at 1:1 speed). That alone would be strictly better than how they are currently doing things, and probably not too difficult to pull off.

Note that humans-in-the-loop is the status quo already at YouTube.

Yet the humans are making mistakes like this thread is about.

I think the intent is to have better humans in the loop and to have a human appeal process.

Currently it seems like the human approved removal has no recourse and you’re just hosed.

I was under the impression that bans were automated, and appeals were handled by humans. I think it would be better if ban “suggestions” were automated and actual bans always came from a human.

Not sure the exact process at YouTube though.

you would really only need humans that watch things that were reported by humans, not every single video.

The human moderators only have to watch what is reported, it is totally unnecessary to watch everything.

126,000 people, assuming a $20,000 annual salary, would be 2.5 billion, or 12.5% of youtubes revenue in 2020. That seems more than reasonable.

And when you consider only flagged videos should *need* to be watched by a human.

$20k annually is below poverty line. I'm sure Google would love to be able to get people to solve their moderation problem for so little money they can't actually afford to both eat and pay rent, but if we're making hypotheticals here, we should at last hypothesize 126,000 people who get to live like real human beings.

I'm sorry, i actually had £20k, and just changed it to $ without converting; £20k is about what a 'Accounts Administrator' earns (https://uk.jobted.com/salary). So that's about $28k which is about 3.5 billion, or about 18% of their 2020 profits. Still very reasonable, most businesses have to put more than half their turnover into wages..

The other thing I'd say is that this isn't low-skill work. If you get thousands of people, train them for a couple days, pay poverty wages and tell them to make each decision in less than 30 seconds, you're not going to get good results.

You accept that there is an upper limit on community size, beyond which unsolvable problems await.

> manual moderating systems are also a no-go at scale

Yeah, I concede this is a very hard problem (especially when advertisers are involved). And even worse, maybe it's not even worth it to solve. After all, if a minority x% of your creators are expendable to false DMCA takedowns because your market share is so huge, who cares?

You revert back to becoming a 'dumb pipe' and make moderating someone else's problem.

Your ISP doesn't moderate for you either.

Okay, now we are back where we started. How does that 'someone else' moderate content at scale? Eventually someone has to take the burden.

I think it basically looks like Reddit. You carve up the site into fiefdoms and then have users self-moderate. Moderation that happens at the site layer largely targets fiefdoms rather than users.

New accounts are easy to create, but new communities are harder because you have to wait for members. Plus most of your moderators are free. I wouldn't be surprised if Reddit only pays for people to moderate the default subs since advertisers and new users are more likely to see them.

If we're all bunched into small communities, there is no 'at scale'.

I regularly hang out in several different communities of thousands of people, including this one. All of them, except the really small ones, have one or a few people who moderate the forums, and ways to report things that are posted. If someone posts a link to disturbing video to HN, it's going to disappear right quick; even if the source lives on. If I try out a community where there's a lot of disturbing videos posted, I'm going to stop going there; but other people might go there specifically because of those video links. If the videos are legal, that's their business and none of mine; I don't want to see it, and by avoiding their community, I won't.

They do if you use their email. All email systems are unusable without spam and origin filtering.

Spam filtering is "keep you from seeing things you don't want to see". Moderation is "keep you from seeing things I don't want to show you".

That's not true. It's more that spam is automatic, and as such is easy to automate away. If someone replies "BUY MY PRODUCT http://product.link" on every post, that's spam. If someone is consistently shitty to other community members and generally acts like an asshole, that's someone who needs to be moderated away.

I partly agree. There's a skill and a moral quality internal to moderation, which means that good moderators will exclude the content they judge the community doesn't want without letting this judgement be swayed by their personal agenda. But bad moderation, where the moderator lacks the skill or the quality to do this, is very much a thing.

They also are unusable with spam and origin filtering. Not getting a message you were waiting for? Turns no message from that service is getting through, no notification, no nothing.

This is actually true, I can’t get any email for one specific Substack I paid for (and others work!). And since they do login by email I can’t log in at all.

>So what do you do?

You could employ a little common sense when designing your product. Is it really wise for platforms like Facebook to allow anyone to start livestreaming unrestricted? I would argue no, at the very least they should be taking measures to minimise the blast radius should someone decide to walk outside and start shooting people.

Slashdot's system works just fine. With the exception of overtly-illegal content, moderation should consist solely of communal inputs to a function that operates solely at the discretion of the individual user.

> moderation should consist solely of communal inputs to a function that operates solely at the discretion of the individual user.

Exactly, Slashdot works because of its transparency; the user can see, what was moderated down, and thus whether the moderation was done to something he doesn't want to see, or whether it is something that someone else doesn't want him to see.

Keep forums smaller than 10x Dunbar’s Number.

Pay for it and stop being greedy.

Stop using bullshit centralized platforms. There is no gatekeeper on the matter of posting videos on the Internet. Established players (like Rossman) continuing to publish through the big players lends power to them. So stop doing that.

For what it's worth, Rossman also publishes on alternate platforms such as LBRY.

You've basically requested to cede all control of the message and to fade into irrelevance.

What about user moderation with [+] and [-] buttons ?

Can easily get you echo chambers if the numbers are visible and people start using them as agree/disagree buttons.

User moderation via "flag as inappropriate" might work though.

Reply on news.ycombinator.com