Live data from Hacker News

You either die an MVP or live long enough to build content moderation

mux.com

301–310 of 429 posts

Re: You either die an MVP or live long enough to build content moderation

#301

It always blew people's minds when I told them that 50% of the engineering time at reddit was spent on moderating. What's interesting though is that we didn't even have any moderation for the first year or so, because the community would just downvote spam. It wasn't until we got vaguely popular that suddenly we were completely overwhelmed with spam and had to do something about it.

What blows my mind is that only 50% is spent moderating.

At some point the engineering work can be something approximating “done” (or should get asymptotically close to it), and the health of your platform ultimately rests a lot more on the quality of the community than on any particular technical project.

What is engineering doing that it takes up as much person-hours as moderation? Running A/B tests to tweak the “try the app” dialog?

(Yes, I’m bitter… I deleted my Reddit account years ago and I’m still lamenting the loss of what the site used to be.)

Re: You either die an MVP or live long enough to build content moderation

#302

Earlier quoted context omitted.

If you can identify text written with mixed glyphs just ban it outright. Normal users don't use text like this, the pure binary presence of such "homomorphic" text at all is probably a better signal for spam than whatever your neural net when running it after normalization.

> Normal users don't use text like this I think that depends on the users. People copying and pasting bits of text that was in English or another common language— think documentation, code, news articles, tweets, etc.— with a different character set could be problematic. Also, 𝒮ℴ𝓂ℯ 𝒜𝓅𝓅𝓈 marketed as "𝔽𝕠𝕟𝕥𝕤 𝕗𝕠𝕣 𝕤𝕠𝕔𝕒𝕝 𝕞𝕖𝕕𝕚𝕒" would be ℭ𝔞𝔲𝔤𝔥𝔱 𝔲𝔭 𝔦𝔫 𝔱𝔥𝔦𝔰. (math symbols) A user base with…

> pasting bits of text that was in English or another common language

If they use many (maybe three? four? or more) character sets in the same post, or different character sets in any single word, then that'd be highly suspicious?

Whilst still letting people copy paste from another language

Special case needed for the shoulder shrug with an Hirigana letter tsu I mean katakana tsu

Re: You either die an MVP or live long enough to build content moderation

#303
post #55

If you host blobs for free, somebody is going to use you as their host. Even if you just hosted audio, I'm sure somebody will quickly come along with a steganography tool to hide their content on your site (and use your bandwidth). Similarly, if you make compute power available, people will use you to mine cryptocurrency. Even if all you host is text, somebody will come along to be abusive. When you put a computer on…

> If you host blobs for free This is the key distinction. If you charge money, from the beginning, most of your content moderation woes go away. At Transistor.fm we host podcasts and charge money for it (starting at $19/month). We've had very little problems with questionable content. We're a counterpoint to the narrative here: small (4 full-time people), profitable, and calm. > Even if you just hosted audio Most DMC…

> We're a counterpoint to the narrative here: small (4 full-time people), profitable, and calm.

I hope your model wins in the end but let’s face it - the internet makes available a global market for nearly no marginal cost per user. Networks effects and all that on top of that premise.

You can carve out a niche but the serious money will be spent on scale.

Re: You either die an MVP or live long enough to build content moderation

#304
post #17

Around 1 year ago we got hit badly on our [blogging platform][0] by people/groups submitting fake customer support description of other big companies, either being Microsoft, Facebook, Comcast etc. Rolled out a machine learning model and trained it on the database. 99% of them vanished. Next day, the machine didn't work and success rate was around 5%. Found out, they have learned the trick and now using symbols from…

> Found out, they have learned the trick and now using symbols from different languages to make it look like English. I wonder if you can train a ML model using text as images . For example, taken as strings, "porn" and "p0rn" are not very similar, but visually they are.

That could detect bad text in actual images too

Re: You either die an MVP or live long enough to build content moderation

#305

It always blew people's minds when I told them that 50% of the engineering time at reddit was spent on moderating. What's interesting though is that we didn't even have any moderation for the first year or so, because the community would just downvote spam. It wasn't until we got vaguely popular that suddenly we were completely overwhelmed with spam and had to do something about it.

What blows my mind is that only 50% is spent moderating. At some point the engineering work can be something approximating “done” (or should get asymptotically close to it), and the health of your platform ultimately rests a lot more on the quality of the community than on any particular technical project. What is engineering doing that it takes up as much person-hours as moderation? Running A/B tests to tweak the “t…

Super-niche subreddits still have something to give, but inevitably success > popularity > memes unless there is strong moderation.

Re: You either die an MVP or live long enough to build content moderation

#307
post #94
post #58

Earlier quoted context omitted.

For adversarial problems like this, a shadowban approach can sometimes be necessary. Perhaps people can still see their blogs but GoogleBot gets blocked from indexing them, or they only appear to someone with the spammer's cookies. That way it takes them longer to catch on and evade the model. Of course, that means you'll need to at least spot check your bans because you can't rely on legit users escalating to you.

Yep, shadow ban was in place as well. The thing is, the people that weren't the ones posting the content. It appeared their computer was affected by some type of malicious file to be part of a bigger network. (botnet?) I could see from the thousands of different IPs in different countries around the globe, that it could be affected personal computers. Very few of them were computers from hosting companies, the rest w…

This seems like a lot of effort? What do you think was the motivation for something like this on your site?

Re: You either die an MVP or live long enough to build content moderation

#308
post #55

If you host blobs for free, somebody is going to use you as their host. Even if you just hosted audio, I'm sure somebody will quickly come along with a steganography tool to hide their content on your site (and use your bandwidth). Similarly, if you make compute power available, people will use you to mine cryptocurrency. Even if all you host is text, somebody will come along to be abusive. When you put a computer on…

Yeah, there's an entire category of "idea guys" who don't get this. They repeatedly try to crack the code on a truly moderation-free or purely crowd-moderated platform, and it never, ever, ever works. It almost always boils down to a poor understanding of how humans work (usually some sort of "homo economicus") or how computers work (usually some sort of "AI magic wand").

Why can't you verify real identities? Can't you use phone numbers, photos of IDs, charging a credit card, verification of physical addresses, or invites to increase the difficulty if creating fake accounts?

Yes, there is an issue of increasing the difficulty if signing up for real users but once accounts are tied to real identities doesn't that allow crowd moderation?

- person whose fallen into your apparent fallacy

Re: You either die an MVP or live long enough to build content moderation

#309
post #65
post #39

Earlier quoted context omitted.

> I'll never forget having to be a moderator for a somewhat popular forum back in the day Similar experience, though I'll say that the worst was dealing with other teenagers that threatened suicide when you banned them. That always took a lot of effort to de-escalate and was a complete drain on personal mental health. I could deal with porn, shock images, and script kiddie defacements, but having people threaten to k…

Several years ago a popular gaming forum with a significant teenage audience I used to read had declared a simple policy toward threats of suicide. If you were threatening to kill yourself, do it , and stop messaging the mods, they are not here to talk you down from a ledge. It seemed pretty effective.

Oof. That is a genuinely terrible and cruel approach.

Re: You either die an MVP or live long enough to build content moderation

#310
post #148

At this point I'm surprised companies aren't blanket-requiring phone numbers so they have something a little more concrete to ban, much less going whole hog and demanding government-issued ID like drivers' licenses or something.

There are multiple channels to evade phone based registration limits. Everything from free VOIP apps to SMS harvesting to SIM farms.
Post reply on HN