I don’t expect the system to be bulletproof and remove advanced spam rings, but when obvious spam is allowed to stay on the platform I’m wondering what the idiots at Twitter are really doing.
I just checked, it's still there https://mobile.twitter.com/ksfAKBARI check followers/following for a massive ring of obvious spam.
Here’s a screenshot of the account just in case this actually works: https://i.imgur.com/xNviClq.png
However this particular account is a spam bot as it occasionally tweets financial frauds such as "yes loans any good (link: yes-loans-any-good.txtloanltd.com) yes-loans-any-good.txtloanltd.com #no fax online loans".
What if they're not idiots? What if that system is working exactly as intended?
Why? If you don't know what the goal is, how can you judge the results? Just because you don't care for the result doesn't make the person who designed the system an idiot. It makes you an idiot for continuing to use a system you hate.
I loved Twitter and I wish they get their stuff together and make it enjoyable again, but I’m not expecting anything anymore. They’re most likely going to go down the same way Yahoo did.
Unlikely. Yahoo actually had to make a profit. In 12 years, Twitter has only reported 1 profitable quarter. And yet, somebody (or somebodies) keep forking over billions of dollars to keep twitter up and running. Whatever their goal is, making money isn't it.
My guess is their goal is the same as the major media outlets: to control the flow of information. It's a lot easier to do that without scrutiny if the tools appear to kick people off at random. The harder it is to discern a pattern, the less scrutiny they'll face when taking down accounts that say things they don't like. It also provides cover if there's a big stink and they have to reinstate an account.
I mean, it works for spam. Obvious bad english, poor spelling and punctuation, etc. weed out people who know this is an obvious scam, leaving only the type of people who will overlook every red flag imaginable because their gullibility and greed overwhelms them.
It is bots and the algorithmic timeline (which Twitter spam with lots of people i don't follow) that ruined it for me. The bot issue seems simpler to solve than spam, (a Turing test?) and I can only think that Twitter feel the bots will one day make them profitable.
Taking an user’s reputation, account lifetime, reputation of their followers, etc (so for example, a report from a new account has less weight than an established account with a good history) will sort that out.
And when everything else fails, a proper appeals system where you talk to humans will solve this problem once and for all. False positives and abuse are bound to happen eventually, but they’re tolerable as long as you have a proper appeals process to deal with them.
That's trivial to abuse, given a group of accounts filing false reports.
That's not a hypothetical; as an example, Facebook's reporting functionality has been abused to attack people's accounts as part of various hate campaigns.
Hiring more engineers isn’t going to solve it. Management and product is where the real problem is IMO.
It got banned within 5 minutes of being created, I didn't use Tor, proxy or anything shady, a normal Chrome browser and a normal gmail account. All I did was Tweet once (their own localized hell world tweet where I didn't even fill in the text).
Reason was something that amounted to breaking their ToS via 'automated use' or some shit. Of course if I were to provide my mobile number it'd get instantly unlocked, 0 human oversight on that too.
Opening a support ticket sent me their automated crap email saying the same thing and asking for phone number again to instantly get my 10 minutes old account unlocked.
I never heard from a human.
They reap what they sow.
Edit: and by 'without a phone number' I mean I didn't fill it in when creating the account (I don't even remember if it asked there but I somehow got an account open without one), just out of principle of not wanting to give any more information than I need to any service I use. And of course I didn't divulge it after that borderline ransomware situation and my opinion of Twitter has dropped (generously speaking..). It was also my only ever attempt at having a Twitter account, not a secondary account or something like that. I also have a static Polish IP for years now, no other service ever 'caught' me for being a bot, it was in an up to date Chrome on Windows 10, etc.
Focusing on profit.
Not to mention, having spam on the platform isn’t really a good strategy for profit anyway.
If you have worked in any company that deals with internet-facing traffic serving the size of audience that Twitter does, it should be obvious that 1k employees is nothing if you want to have all the necessary teams you need to keep the business running, for example, software engineering, software support, hardware engineering, hardware support, network engineering, software engineering, HR, finance, legal, PR, management, IT department, etc.
A company of this scale is not just code and raw technical skills. There are many nuances that you need to appreciate before you understand why it involves 1k-5k employees.
Typically it doesn't happen because it's more profitable to use data scientists to sell more ads.
I've reported someone leaking Polish court documents of an ongoing case on Facebook - nothing happened.
I've reported a YouTube video praising Anders Breivik as an European hero defending whites from Marxists, Leftists and Muslims - it's still up.
I've created a Twitter account and it got instantly locked for being 'suspicious' or 'automated' (but of course giving them my phone number would fix it instantly) - https://news.ycombinator.com/item?id=17014511
What are the nuances that I'm missing?
I challenge you try and think through why your simple spam filter and human appeals process will not scale to Twitter's size. What kind of throughput can it handle? What happens when there is more content being appealed than the review team can handle? How do you prevent attackers from maliciously flagging other users' content? How do you handle advanced threats where attackers compromise existing accounts then begin to spam, thus bypassing any "easy" checks like looking at account age? How do you classify "hateful" speech? How does it scale to foreign languages, slang, etc?
A constructive comment would have been to point us at good spam filters that are very good at not allowing obvious spam so that we do not make the same mistake that Twitter did and be labelled as "idiots" by the likes of you on the Internet.
Seriously, this problem is not as easy to solve for machines as you think it is.
And email seems like a harder place to police, being decentralized. These guys control the gateway to the service!
The first rule is you don't delete user data without consent. The second rule, is you dont delete user data without consent!
And yes, I work in the industry. Bad shit happens, and data can be lost. But trusting some automated 'spam detection algo' and making actionable, irrecoverable decisions on fuzzy math is completely idiotic, boneheaded, and fucking stupid! If you're in that product team, you deserve every shred of scorn you're getting.
I think some of the reason this comes about is that it's deprioritized until it's a major problem, and you're already at a large scale, and then the decision is "require human review, even of what appears to be a fraction of the deletions, and get swamped with thousands or tends of thousands of reviews immediately", or "just delete them".
That's not an excuse for the decision, since obviously they could have taken care to address this problem at least minimally much earlier, and there wouldn't be a deluge of work, but I think we all understand whey they weren't keen to do that (even if it's a crap decision for their users).