Friday, November 11, 2016

Twitter's Double Standard Spam Filter

Did you know that if you post multiple tweets containing similar content (I have found that this is generally 4 or more in a row) that Twitter will begin to filter any new duplicate tweets as spam?

The sad irony is that election season is a great time to test Twitter's spam filter to see who's tweets didn't make the cut.  Once Twitter decides a tweet is spam that tweet is no longer searchable and, according to the results of my most recent CA Legislative Open Records Act request, you can tweet your state representatives telling them how you would like them to vote, but don't count on your representative being notified that you were trying to tweet them.

My previous blog post showed how the California Nurse-Midwives Association and a number of their Twitter supporters had their tweets regarding Assembly Bill 1306 censored as spam on Twitter and that the representatives they posted their tweets to did not receive any notification that they were mentioned in the tweets.  This is much like your gmail account failing to deliver email to your public official based on the discretion of Google.

Keep in mind that when you mention someone in a tweet you see a confirmation that "your tweet was sent" or "your tweet was posted."  This is nothing short of fraud for Twitter to tell you one thing and then not follow through with it.

So, what makes election season a good time to test Twitter's spam filter is that lots of elected officials have a Twitter account and if you are a group that wants to promote your favored candidate you can easily fall into a spam-trap.  The 140 character limit also helps increase the odds because what happens is that it's easy to copy and paste a repetitive message as opposed to composing a whole new tweet for each individual.

So, let's compare repetitive messages posted by the @CA_Dem and @TxValuesAction accounts.  What I discovered with my sample below is that none of the tweets posted by the @CA_Dem account are censored, but all of tweets posted by the @TxValuesAction account are.  For @TxValuesAction, I did not include the uncensored tweets posted in the series of repetitive tweets.

In another previous blog I discuss how to find a censored account, but in this instance neither account is censored, only selective tweets are.  See the blog if you want to know more about searching for censored Twitter content.

@CA_Dem Tweets:

@TxValuesAction Tweets:

Repetitive tweeting looks like it's okay depending on the account that posts it.  I see no reason why one account would have their tweets filtered as spam, but not the other.  Also, if my previous LORA request is any indication, the @TxValuesAction posted a lot of congratulatory tweets that were never delivered to their intended recipient.

Tuesday, November 8, 2016

Public Records Request Reveals Twitter Censors Political Speech

Last month I posted a blog about how Twitter Censors "Spam" Posted by the California Nurse-Midwives Association and yesterday I finally received even more disturbing proof of what Twitter actually does.

If you didn't read the blog then let me give you a brief overview.  A few months back the California Nurse-Midwives Association (@CalMidwives) tweeted their support of Assembly Bill 1306 by using the #AB1306 hashtag.  In a following blog I explained how The California Democratic Party Twitter Account Gets Away with Posting Spam when organizations like the @CalMidwives had their tweets promoting the passage of specific legislation censored.

I knew that the @CalMidwives had repetitive tweets censored, most likely as spam, yet, the @CA_Dem account engaged in the same activity, but their tweets failed to get censored.

And just to make things further convoluted for you, I knew from a previous open records request that a person tweeting from a censored Twitter account may tag people, like their local police department,  in a tweet, but the tagged account may not get a notification about it.  For example, the account of ACLU lawyer @carltonwilliams is currently censored (and has been for some time).

I found a tweet Mr. Williams sent to the Revere Police Department and I asked if the @reverepolice was notified of the tweet.  They were not.
Knowing that the @CalMidwives had a number of censored tweets tagging and/or mentioning members of both the California Assembly and Senate, I submitted a public records request to those individuals.  There were many censored tweets and numerous elected officials to choose from.  I decided to narrow down my requests to members of the California Tech Caucus.

Yesterday I received the response to my request in regards to members of the California Assembly.  You will notice that it wasn't only the CNMA that was caught up in the Twitter-spam-filter-mess, but also some of their supporters like @karenrubybrown and @JonathanBartle8.

These are the tweets I inquired about and the results I got:
Request: I would like to know if Twitter notified the @JimPatterson559 account that this tweet was sent.
Result: No notification received.

Request: I would like to know if Twitter notified the @ScottWilkCA account that this tweet was sent.
Result: No notification received.

Request: I would like to know if Twitter notified the @asmMelendez and @AsmL2Chang accounts that this tweet was sent.
Result: The Melendez account is personal and Chang did not receive any notification.

Request: I would like to know if Twitter notified the @jacquiirwin and @MattDababneh accounts that this tweet was sent.
Result: The Irwin account is personal and Dababneh did not receive any notification.

Request: I would like to know if Twitter notified the @NoraCamposCA and @Evan_Low account that this tweet was sent.
Result: The Campos account did not receive any notification and the Low account is personal.

Sadly, if you thought that Twitter was a great way to promote your ballot measure, political campaign, or agenda you are mistaken.  These people thought they were tweeting their elected representatives and that there was a possibility those representatives would read their message.  The only problem is that their tweets ended up censored and the accounts being tagged failed to received any notification from Twitter that they were being mentioned.

So go ahead, feel free to tweet the members of your state legislature.  Just don't tweet them something that gets filtered out as spam.

Friday, November 4, 2016

The California Democratic Party Twitter Account Gets Away with Posting Spam

The Equality California Twitter account, @eqca, has posted tweets that shows how Twitter is censoring their tweets, but not similar tweets posted by the California Democratic Party Twitter account, @CA_Dem.  However, Twitter didn't always consider the @eqca tweets to be spam.

On October 24, 2016 the @eqca posted a series of tweets that ended up being censored because (my best guess) they were filtered as spam.  I've posted a previous blog about how Twitter Censors "Spam" Posted by the California Nurse-Midwives Association and the tweets posted by the @eqca look like they have fallen into the same trap.  The problem is that the tweets are posted within a short time period and contain the same or similar text.

Here are some of the tweets the @eqca posted that are censored:

I won't post all 135 of the tweets here.  They all look much the same.  Of the 135 tweets, 127 of them ended up being censored (by this I mean that the tweets are unsearchable on Twitter).

For example, I know the tweet mentioning @AsmAutumnBurke is censored because when I search the text I get no results:

What I also know is that there is a very HIGH chance that Twitter did not notify the tagged account that they were mentioned in the censored tweet.  The response I received to a recent public records request regarding a known censored account indicated something peculiar.  I knew that the account of @CarltonWilliams was censored.  So, I found a tweet where he had tagged the @reverepolice.  Knowing I could expect a response submitted to a police department, I submitted a public records request asking the Revere Police if there account received a notification that they were mentioned in the tweet posted by Mr. Williams.

This is the public records request I submitted and the response I got:

So, an ACLU lawyer is censored on Twitter and if he tweets a public agency that agency will have no clue he was talking about them on Twitter.  Um, Wow!  I didn't know that was how Twitter and free speech were supposed to work.  That's like Google not liking my "police" searches and deciding I shouldn't be allowed to email my local police.

I have submitted a public records request to Assembly Member Burke and a few others to see if they received notification of the censored tweet posted by the @eqca.

But here's the funny thing, doing an advanced Twitter search for tweets sent by the @eqca account using the text "We're proud to endorse" posted during the year 2014 yielded 15 related tweets:

I have not had the time to scroll all the way back to 2014 on the @eqca timeline to see if they actually posted more tweets than are showing up in the results, but keep in mind that by 2016 they only got away with posting 8 searchable endorsement tweets.

Here's a few more unlucky accounts who fell into the same Twitter-spam-trap:

But here's the sad part for Equality California and their @eqca Twitter account.  As recently as November 1st the @CA_Dem Twitter account engaged in the same type of tweeting behavior, but somehow their tweets managed to not get filtered as spam:

For example,
This tweet is searchable and 9 others just like it were posted the same day:

So, somehow a series of similar tweets posted by the @CA_Dem are not spam, while a series of similar tweets posted by @eqca@EQILPAC@EqualityNewYork@ELMActionFund, and the @CalMidwives are.  We should all be asking ourselves how we are to know what gets censored as spam and what does not and Twitter should be transparent about it.