After a false positive, Twitter suspends open government blog for being too social

Twitter’s best practices for tweeting don’t appear to mix well with its rules for tweeting, as I found out last month when the social networking company briefly suspended the Twitter account for this blog. While I was able to quickly get the account back online, the episode raises somr issues regarding how Twitter’s algorithm flags media accounts and some contradictions in the company’s guidance for new users.

When I found that I couldn’t file a help request to Twitter Support to appeal the suspension of @e_pluribusunum_ through that account, I used my main account (@digiphile).

Initially, I thought the suspension was due to spam, similar to the situation David Seaman encountered in 2011.

After I directly contacted Twitter for help, the account went back online later that day:

As I found out days later, however, the suspension was for “sending multiple unsolicited @replies or mentions,” per the statement I have from Twitter Support on @e_pluribusunum_:
“This account was suspended for sending multiple unsolicited @replies or mentions. Twitter monitors the use of these features to make sure they’re not abused. Using either feature to post messages to other users in an unsolicited or egregious manner is considered an abuse of its use, which results in account suspension. You can find more information about @replies and mentions here:https://support.twitter.com/articles/14023-what-are-replies-and-mentions

I have now unsuspended your account. Please note that it may take an hour or so for your follower and following numbers to return to normal. Be sure to review the Twitter Rules, as repeat violations may result in permanent suspension: http://twitter.com/rules”

suspension-notice

The tweets in question, however, are extremely similar to the way I’ve been using Twitter for years, advise others to use Twitter, and that Twitter itself recommends to new users.

Here are the tweets sent the day before the suspension and the three that morning, which I have to assume triggered the suspension.

The seventh tweet, embedded above, had six different @names in it, but it was appropriate: I was attributing the source of the information, referring to an NPR program (The Kojo Nnamdi Show) and naming the 4 guests who were on it. The eighth tweet had three @mentions in it, as I had retweeted a media account that referred to a reporter and added the subject of the story for context.

So: Were there a lot of @mentions? Yep. Were they “unsolicited?” Yep.  That accurately describes tens of thousands of tweets that I’ve sent over the past seven years. In this case, they were far from “abuse.”

That led me to wonder how many people, journalists, government or media companies or nonprofit organizations a Twitter account is allowed to @mention before it’s suspended. Should any of the categories of users I listed now have to actively ask followers for feedback or allow others to talk about them? That doesn’t seem practical nor scalable. Are there different rules for different users, Verified or not? (I’ve asked Twitter for comment on these general questions but have received no answers after two weeks. I will update the post if I do.)

In the meantime, I’ve tried to think them through myself. The “newness” of this account likely tripped Twitter’s automated filter, leading to the suspension. That means that other new users have to think about whether they’re sending “unsolicited replies or mentions” to keep clear.

I found that deeply jarring. I used the @E_Pluribus Unum_ account exactly as I have @digiphile, for over 7 years now, resharing tweets with attributed context and quotes, tweeting about public figures and government officials, tagging mastheads, retweeting select tweets.

That’s more or less how I define being “social” and engaging on the platform. That’s how I thought Twitter defined it, too. Twitter’s own best practices for engaging followers recommends it:

Mention high-profile users
@HillaryClinton included Secretary of State Madeleine Albright username@madeleine in a Tweet welcoming the former Secretary to Twitter. In turn, Albright replied to @HillaryClinton and also mentioned the Kennedy Center (@kencen), where she had recently performed. Including so many mentions of other users makes it more likely that people will find the conversation and join in. ”

If Twitter is suspending new accounts that @mention too many high profile users or reply to them in an “unsolicited” fashion, I can’t help but have serious concerns about Twitter’s future and commitment to being a platform for free expression, government accountability, or hosting civic dialogue.

I do see potential issues with “egregious” @mentions — “@reply spam” has been an issue on Twitter for years — but isn’t that exactly what the block button has been used for, or the new abuse reporting button should be used for? People have been tweeting “#FollowFriday” recommendations for years with many unsolicited @mentions. Are they risking suspension?

Honestly, knocking new accounts offline for being “too social” suggests a tone-deaf algorithm. Ignoring my questions regarding general standards suggests something else. (The company generally refuses to comment on individual accounts.)

Given reports of retention issues and low activity by most users, an overly aggressive approach to filtering new users that are engaging in activity that Twitter itself recommends, particularly media accounts, strikes me as actively self-defeating.

Twitter and its investors should care about the people who never tweet. This experience reminded me that those same parties should care about the people who do tweet and are caught up on algorithmic censorship, followed by vague missives not to talk about other accounts too much.

As I’ve written before, Twitter is not a public utility. It’s a private company with a Terms of Service and Rules it itself sets.  If Twitter’s users don’t like them or lose trust, their option is to stop using the service or complain loudly on other platforms.

In general, Twitter’s record on censorship, Internet freedom and privacy is the best of the big tech companies, as an analysis by the Electronic Frontier Foundation last year highlighted. They’ve gone to bat for their users, from Turkey to Washington. Today, however, I just wish they’d clarify how social those users are allowed to be.

Editor’s Note: The headline of this post has been amended, with “After a false positive” added.

5 thoughts on “After a false positive, Twitter suspends open government blog for being too social

  1. Actually, i don’t think Twitter’s filters were miscalibrated on this one.

    Setting up a new account for E-PluribusUnum is structurally indistinguishable (though not semantically) from a spammer.

    The game spammers all play now is essentially a sort of filtering/Turing test. They’re trying to behave in a way such that their behavior can not be structurally discriminated from the behavior of legitimate users.

    So, if twitter can’t discriminate based on the structure of individual messages, they presumably look for other factors upon which they can distinguish real users from spammers. I presume that Twitter does this through a combination of factors, but network analysis is assuredly part of the milieux.

    You don’t raise flags for twitter when you tweet the way you do, because you have a kajillion followers and folks interact with you when you tweet about them. You also have a history of this behavior.

    So if twitter only has the history and pattern of followers to judge whether an account is a spammer or not, starting a new account (without doing all the sorts of announcement and fanfare to collect new users) and tweeting with it, exactly as you use your existing account means that you’ve stripped away all of those things that Twitter uses to tell that you’re signal and not noise.

    And given that the prior probability of whether new accounts who mysteriously behave like Alex Howard are spam is high, I’m not surprised that you tripped their spammer detection.

    Seems like it might be helpful for twitter to let you imbue new accounts with more of your “you-ness” so that they know a new account is legit. Either some notion of forking accounts or sub-accounts or something would address an issue like this.

  2. Pingback: The Blind Enforcer: The Reflex Application Of Rules | Campaign Mastery

  3. Pingback: Twitter suspends open government blog for being...

  4. Pingback: Twitterban: Pas op met ‘ongewenste @mentions’ | Twittermania

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s