During the 2020 election, the Department of Homeland Security’s Cybersecurity and Infrastructure Agency (CISA) partnered with the Election Integrity Partnership (EIP), a consortium of groups led by the Stanford Internet Observatory to track and counter what they considered mis- and dis-information.
EIP surveilled hundreds of millions of social media posts and collected from the cooperating government and non-governmental entities that it calls its “stakeholders” potential violations of social media platforms’ policies concerning election speech.
It coordinated its efforts primarily through a digital “ticketing” system. There, one of its as many as 120 analysts or an external partner could highlight a piece of offending social media content, or narrative consisting of many offending posts, by creating a “ticket,” and share it with other relevant participants by “tagging” them. Tagged participants could then communicate with each other, in something of a group chat, about the veracity of the flagged content, concerns about its spread, and what actions they might take to combat it.
For social media companies this meant removing the content outright, reducing its spread, or “informing” users about dubious posts by slapping corrective or contextualizing labels on them.
RealClearInvestigations (RCI) has obtained data associated with nearly 400 EIP tickets, data produced for the House Homeland Security Committee in connection with its oversight efforts.
The tickets come in the form of a series of spreadsheets. Each row represents one ticket. There are 95 fields of data associated with each ticket. The fields include descriptions, such as “Misinformation tweet regarding re-voting,” or “Voter turnout >100 percent in swing states chart, spreading on Twitter.” The fields also include URLs of the associated questionable content; “organizations,” presumably representing all partners tagged; and the comments from such stakeholders.
The Stanford group provided no key for the spreadsheets. Much of the information is redacted, and the source of the comments on each ticket is not always discernible. The data suggests there were images embedded in comments that are unviewable.
Targeting of President Trump by CISA, an Executive Agency
Ticket EIP-482 (created Oct. 27, 2020) was originated by the CISA’s Elections Infrastructure Information Sharing & Analysis Center (EI-ISAC). It concerns a tweet from then-President Trump indicating “most” states permit one to change one’s original vote after engaging in early voting, which EIP categorized as potential “Procedural Interference.”Agencies Serving as Force Multipliers
Ticket EIP-257 (Sept. 29), originated by the EI-ISAC, concerns a social media post from an unnamed user, alleging an absentee ballot had been delivered by mail to his dead father. An EIP stakeholder “flagged the post to Facebook for removal and the link is no longer active which means it has either been taken down or made private to the individual’s Facebook.” A subsequent comment notes that “We also received confirmation from Facebook (by way of CISA) that Facebook took action on this case,” again showing EIP and CISA seemingly working as force multipliers in content moderation.Targeting Content Because of How It Might Be Perceived
Ticket EIP-301 (Oct. 2), originated by the EI-ISAC, concerns a “tweet regarding voting machines.” An elected official reported that the since-deleted and unavailable tweet “is false. Voting machines work the vast majority of the time. Old machines do have issues, but to phrase it like [this] vastly overstates the scope of the problem.” CISA inquired as to whether Twitter took the tweet down. It did, suggesting that the election official’s interpretation of the tweet led to the speech being suppressed.Making Explicit EIP Is a Proxy to Avoid Legal Issues
Ticket EIP-923 (Nov. 6), originated by the State Department’s Global Engagement Center, highlights an Instagram post showing a video depicting a poll worker purportedly caught modifying ballots. The GEC submitter writes:Targeting Conservative Sources for Raising Questions About the 2020 Election
Ticket EIP-954 (Nov. 8), the origins of which are not discernible, concerns social media posts sharing an article from The Federalist, where I am a senior contributor, titled “America Won’t Trust Elections Until The Voter Fraud Is Investigated.” According to the ticket, the article “Misconstrues Disinformation as Evidence.” One tagged post comes from Federalist editor-in-chief Mollie Hemingway. A stakeholder writes to Facebook and Twitter in connection with the ticket that “this seems to be the greatest hits from the past 3 days wrapped up in one article. The article links to several of the gateway pundit links which have received action since Tuesday.” Twitter indicates it received and was reviewing the tweet, though it appears not to have taken action on it.RCI asked Hemingway for her comment on the flagging of her tweet and publication’s work. She replied:
“This unconscionable censorship of The Federalist and its reporters is sadly unsurprising. The censorship-industrial complex in this country clearly views free speech as its enemy and will do anything to shut it down, including spreading lies and using intimidation to coerce private companies to censor factual, legal speech on behalf of the regime.”
Discounting Election Integrity Skepticism, Even When Based on US Government Documents
Ticket EIP-961 (Nov. 9), the origins of which are not discernible, includes a tweet from senior analyst for strategy at the Center for Security Policy, J. Michael Waller. In it, he quotes testimony from a State Department official in 2004 highlighting indicators of Ukrainian electoral fraud, including “‘illegal use of absentee ballots,’ ‘opposition observers ejected,’ ‘North Korean style turnout in the East’ (‘nearly 96 percent’), ‘mobile ballot box fraud,’ etc.”—the implication being that the 2020 U.S. election was similarly flawed. Waller thanked Andrii Telizhenko, a former Ukrainian diplomat, in the tweet for drawing his attention to the testimony.An EIP commenter writes to Twitter: “We believe that this content should be reviewed and labeled under Twitter’s Civil Integrity policy. Old State Department documents (originally amplified by Andrii Telizhenko, a U.S.-flagged Russian agent) are being used to infer that the 2020 election was ‘rigged.’ These inferences are inflammatory and unsubstantiated.” Twitter received the comment, though it is not clear if it took action.
RCI asked Waller what he made of the flagging of his tweet. Noting that “this tracks with what appeared to be shadow banning of my Twitter account,” Waller continued:
“It now appears that it’s ‘inflammatory and unsubstantiated’ to cite the State Department’s own standards for foreign elections as a benchmark for determining the integrity of our own.”
“If the U.S. just applied the State Department’s own foreign election integrity standards to American elections, there would be little electoral fraud in our country. Somehow, CISA thinks this is a subversive idea and must be censored.”
When RCI noted that there was no indication that CISA personnel observed or commented on his tweet, Waller replied, “I draw no distinction between EIP and CISA. One operates as a front for the other.”
Countering Credible Claims That Might Mislead
Ticket EIP-596 (Nov. 2), originated by the EI-ISAC, concerns a New York Post article alleging individuals cast ballots in New York City under the names of dead people. A commenter on the ticket writes that “regardless of whether they [the allegations] are true, they have the potential to be blown out of proportion given that only two ballots were involved. Posts that claim they are part of organized voter fraud should be labeled, and this narrative should be monitored for fact-checking and in-scope claims.” Another analyst concurs, noting that “This story is very likely true, but these two votes (which were both investigated) are not evidence of widespread voter fraud. I recommend that we notify FB/Twitter of the spread of this content with the recommendation that links to the NYPost and Fox News articles be labeled, and any comments that allege that this reporting is evidence of widespread voter fraud be actioned appropriately.” A subsequent comment indicates that Facebook “actioned” accordingly.Protecting the Narrative That Mail-In Voting Is Secure
Ticket EIP-450 (Oct. 23), the origins of which are not discernible, flags a number of social media posts around the Pennsylvania Supreme Court’s 2020 ruling “regarding accepting ballots with mismatched signatures,” whereby, according to a commenter “[t]he prevailing narrative … is that this is an indication that people can commit voter fraud through vote-by-mail.” The commenter adds, in what appears to be a directive to Facebook and Twitter, that “we recommend that the posts be labeled … clarifying that vote by mail is secure.” Facebook appears to respond in a subsequent comment that it “applied the correct labeling.”Attempting to Suppress Speculative Tweets About Mail-In Ballot Handling
Ticket EIP-996 (Nov. 11), the origins of which are not discernible, flags many social media posts concerning claims that shortly after the 2020 election, Philadelphia had destroyed all ballot envelopes. One flagged tweet came from Ron Coleman, who noted that this claimed destruction “makes it hard to prove Biden got any legitimate mail in votes at all.”“It increasingly looks like this [the destruction of ballots] didn’t happen. But it is the response I have given to people who ask me, well what if they just destroy all the ballots—cover up the crime?”
“That don’t work here.”
Coleman added: “Amazing how that works. I corrected the record without any censorship because of my own regard for the truth and my reputation.”
“In contrast,” Coleman said, EIP “operate[s] in the shadows” and is “unaccountable.”
“Why didn’t they confront me instead of ‘telling on me’?” he asked rhetorically.
Political Stakeholders Target Opponents
Ticket EIP-325 (Oct. 7), originated by the Democratic National Committee (DNC), concerns a post from the conservative Texas Public Policy Foundation (TPPF) on mail-in voting linking to a report from the think-tank. An EIP analyst writes: “At this time, there is limited traction on the TPPF’s report on Facebook and Twitter. Further, the FB [Facebook] post’s text is not necessarily false, but the link to the webpage does contain false/misleading information. We are moving this report to ‘ongoing monitoring’ for now due to the limited traction on it, but recommend keeping an eye on TPPF for other misleading or false reports.”When notified of this flagging effort, Chuck Devore, an executive at TPPF who authored the think-tank report, told RCI he was “very disappointed that an academic quality white paper on the efficacy of voting by mail would be flagged.”
The former California state assemblyman continued:
Flagging Political Figures’ Satirical Tweets
Ticket EIP-460 (Oct. 25), originated by the EI-ISAC, concerns a tweet from former Arkansas governor Mike Huckabee in which he writes that after voting in person, he “then voted the ballots of my deceased parents and grandparents. They vote just like me! #Trump2020.” EIP comments that:“Given Huckabees (sic) following and the viral nature of this content, this tweet has the potential to cause harm and confusion about mail-in voting, in addition to propagating/encouraging false claims about how to vote illegally. We recommend that Twitter labels the post with (a) proper voting information, (b) ‘This Tweet violated the Twitter rules about misleading information that causes confusion about the laws and regulations of a civic process. However, Twitter has determined that it may be in the public’s interest for the Tweet to remain accessible,’ or (c) removes the post entirely. We believe that labeling it as ‘could be in the public’s interest’ could be an opportunity to provide correct voting information, while complete removal could amplify the claims further.”
Flagging Political Figures’ Tweets on Election Outcomes
Ticket EIP-936 (Nov. 6), the origins of which are not discernible, concerns a tweet from U.S. Senator Thom Tillis (R-N.C.) from just after midnight of election night 2020, in which he declared victory. EIP labeled this an instance of potential “procedural interference.” A commenter forwards the ticket to Facebook and Twitter noting the senator is saying he “already won the election when official sources have not yet called the election,” a potential “early claim of victory” infraction. The platforms had adopted content moderation policies concerning premature claims of victory. According to EIP’s 2020 report, platforms were most likely to take action on such claims among all categories of offending content. It is unclear whether Twitter actioned Sen. Tillis’ tweet and associated posts.Targeting Progressive Narratives
While the tickets overwhelmingly targeted speakers on the right, RCI identified some notable narratives targeted by EIP propagated from the left.“The problem with going to court for Trump is that judges require actual evidence — not just assertions — to support a claim. But the only available evidence of wrongdoing appears to be the USPS’ massive failure to deliver properly mailed ballots to Democratic-leaning polls.”
A commenter writes that the claim is false: “Provisional ballots are counted, but they may not be counted on Nov. 3rd .... We recommend that this tweet be at least flagged as misleading, if not removed.”
It is not clear whether Twitter took such action.
Hartmann did not reply to RCI’s request for comment on the targeting of his tweet.