How Twitter has an AI Problem

With the number of Tweets that get posted to Twitter every second, it would be impossible for Twitter to maintain a sufficient staff of people to review everything that gets tweeted for content which violates their rules. This means they turn to artificial intelligence, or AI, to solve this problem. But what happens when a tweet gets caught in the net because it believes it falls into a prohibited category? The obvious solution would be to first have a human review the content to see if it did in fact break any of the rules. Unfortunately, that does not seem to be the case, or at least I do not think that is the case as I found out.

Elvis has been crowned the King of Rock-n-Roll and since his time in music history there have been people impersonating him. Elvis impersonators can be found worldwide, officiating weddings, sky diving, there is a 500lb Elvis impersonator, women who impersonate Elvis, and many more. I am not an Elvis fan by any stretch but there is one Elvis impersonator I found interesting.

In Brooklyn, New York you can call a verification phone number and hear a recording on a piece of telephone switching equipment and hear someone impersonating Elvis while listing which prefixes that central office services. Given my interest in telephones, telephone history, and knowing this bit of trivia I tweeted this to the Museum of Communications on March 29, 2022: “Did you know Elvis lives in a DMS100? (NPA) NXX-XXXX” (I am redacting the phone number to not generate further calls, but if you want to read more about it visit Tedium.co and the Weird Phone Number History page.) Within a couple of minutes, I received an email from Twitter stating my account has been locked for violating the Twitter Rules, specifically rules against posting private information. I read through the rules and it is clear that it is against the rules to “…publish or post other people’s private information without their express authorization and permission…” however, it goes on to state that “If the reported information was shared somewhere else before it was shared on Twitter, e.g., someone sharing their personal phone number on their own publicly accessible website, we may not treat this information as private…” Ok, so this really isn’t Elvis and this isn’t someone’s personal phone number since anyone who calls this number is connecting to a DMS-100 telephone switch and listening to a recording. I will appeal this decision.

I received a reply to my appeal 12 minutes after submitting and their support team “determined that a violation did take place.” I don’t understand how they came to that conclusion since it wasn’t a personal phone number. I understand how their AI was triggered and flagged my tweet because there was a proper name, I said “lives”, and it had a properly formatted phone number. But let’s look at the timeline.
March 29, 2022, 5:41 a.m.
Email received my account is locked for violating Twitter rules moments after Tweet is sent
March 29, 2022, 5:49 a.m.
Submitted an appeal
March 29, 2022, 6:02 a.m.
Appeal was denied

This all happened in the matter of 23 minutes. It is possible a human was looking at the initial appeal, but what happened next, I feel is completely unacceptable and what many other users continuously face when trying to get any sort of support from Twitter.

I filed another appeal with Twitter and this time I was clearer exactly the context of the tweet, the phone number, it is publicly known, even linking to YouTube videos which show recordings of the phone number being dialed, displayed, and the recording.

– no response –

I submit a support ticket through Twitter’s support page with the same details, provide my phone number, email address, real name, and get a page saying I will receive a response in a few days but in some cases it may take longer. I give them a bit more than a week.

– no response –

I cancel my previous appeal and submit a new one, providing even more information, advising that no one looked at my previous appeal, left my contact details in this appeal, and waited.

– no response –

I saw tips on how to receive a response from Twitter by using the Gmail trick of adding a +[something] into your Gmail address. I start using my Gmail+1, +2, every other week on their support page form hoping to get a response.

– no response –

In a final desperate attempt to raise anyone’s attention at Twitter, I open a complaint with the Better Business Bureau and this time I write out a long, thorough statement including copies of my Tweet, copies of the rules they claim I violated, and submitted and waited. Finally, a response.

…In order to assist you, please provide your Twitter support case number…

This was the proposed resolution that Twitter sent back to the BBB to forward to me. I rejected it and explained I have made several attempts to raise anyone at Twitter and had not even received as much as a case number. Then a few days go by, and I get a response from Twitter via the BBB

…In order to assist you, please provide your Twitter support case number…

This is strange. It is almost as if they aren’t even reading the complaints received via the BBB either. I reject this solution as well, explain once again that I am not receiving a case number, elaborate that this is the issue I am running into. But this time nothing. I don’t hear back from the BBB for a couple of weeks. Then I look up my complaint and see that the BBB has closed it stating the “Better Business Bureau has determined that the business made a good faith effort to answer the complaint and has now closed it.” How can asking for a case number over and over, which I never received because the system is so broken, be considered a “good faith effort” to resolve the issue? I sent a follow-up email to the BBB asking them to reopen the complaint because I did not believe this was a good faith effort. Since the BBB route failed, I may escalate to the California Attorney General consumer complaint division and see what that yields.

At this point I still do not have access to my Twitter account. By all accounts I should be able to cancel my appeal, delete the tweet, and acknowledge I violated the rules, but I don’t have that option. Even if I did, should I? They don’t explain what exactly happens if you do acknowledge that you violated the rules other than further violations may result in a permanent ban. At this point I want to be contacted by an actual human to receive confirmation that yes, it did in fact violate the rules or no, we made a mistake. If it is the former, what does that mean for posting phone numbers for other entities versus people? What if a company has an un-published phone number but they tweet to someone to call that number instead of their main line for a specific issue? What if someone finds another means to contact a company through another website and Tweets it to someone? Again, these are not personal phone numbers we’re talking about and doesn’t come close to the scope of doxxing.

There are so many questions here but the biggest question of them all is does this way of monitoring tweets for violations make it unfair if you can’t reach anyone to appeal or is it just an AI problem which needs to be better programmed?

I want to be clear that this post is not about me being locked or suspended from Twitter. I understand that is needed to run the service. What this post is really about is lack of response from Twitter. There is no follow-up when following their own process. Many accounts are swept up in this AI net because certain words or phrases trigger its ban algorithm and the only option is to hope someone at Twitter replies. Who knows when that will be. So far, from my own experience, that doesn’t seem likely any time soon if at all.

For now, I will not be able to Tweet until Twitter responds, but you can still follow me @FinnleyDolfin

Did you ever get your Twitter account locked or suspended for violating the rules but really didn’t? Did you go through their appeal process? Did you ever get your account back? Let me know in the comments.

Update 05/13/2022:
It has been 45 days since Twitter locked my account. I still have not received a response from my appeal nor support requests. I still have not received a case number.

On the bright side, the owner of the blog I mentioned above tagged TwitterSupport trying to get some attention to my case, but still nothing.

For now I have been lightly using my bot account to send out messages and it posts the daily count of how many days it has been since my account has been locked.

2 responses to “How Twitter has an AI Problem”

  1. Hey there,

    This is Ernie Smith from Tedium—I feel terrible to hear that your sharing of a phone number from my article got you a Twitter rule violation. Please let me know if you need anything—I’ve tweeted about this issue on my account in hopes of drawing attention to the situation. Hopefully I can help.

    1. Hi Ernie,

      Thank you for mentioning it on Twitter. However, as my post says, I do not believe there are any humans working at Twitter anymore. Funny though is I knew of that number from the old BBS days when it was passed around on lists of interesting numbers. That must have been mid-90s. I forgotten about it and then went to look it up to share with The Connections Museum.

      For now, I will continue to cancel and re-appeal and still do the daily count of how many days it’s been since I’ve received a response.

      -Finn

Leave a Reply

Your email address will not be published. Required fields are marked *