The Australian Electoral Commission has struggled to get Twitter to remove posts that it says are inciting violence against staff and promoting disinformation about the electoral process ahead of the Indigenous voice to parliament referendum, documents reveal.
Correspondence between the agency and social media company – now known as X – obtained under freedom of information laws show frustration over “nil actions” by Twitter on AEC reports in April and May, with a referral left pending for up to 15 days and some appeals left even longer.
The documents show Twitter repeatedly ruled that tweets referred by the AEC were not against its terms of service. On 2 May, the AEC asked Twitter to reconsider its decision not to remove a tweet that raised “concerns it is inciting violence against AEC staff”.
“We believe a post claiming the AEC is corrupt and stating ‘You pack of dogs will have your day’ is intended to incite violence against our organisation and staff members,” the AEC wrote.
Another tweet the company declined to remove replied to an AEC tweet, stating “stop interfering in policy, or pay the price”, which the commission also interpreted as a threat.
On 10 May, the AEC again asked Twitter to review its decision not to take down a tweet that it said “wrongly implies that ‘fake’ voters can be enrolled on the Australian electoral roll, and encourages users to do so via a monetary incentive”.
“Falsely claiming that enrolment fraud is occurring – without evidence – only serves to bolster disinformation and undermine faith in the electoral process,” the AEC said.
An AEC spokesperson confirmed the tweets appealed against in April and May were not removed by Twitter, but said the company had in other cases added warning labels, removed one post and suspended accounts in response to the “small number” of posts it had submitted.
“People have the right to express whatever opinion they would like to,” he said. “AEC communication, including on social and in liaison with social media organisations, is only ever in relation to objective information regarding electoral processes we deliver, or threatening behaviour.”
Analysis of more than 200,000 tweets by a Queensland University of Technology academic found that while the yes side was more prolific on the platform, there was a significant boosting of conspiracy theories by some accounts pushing for a no vote, including false claims about the AEC.
Another tweet the AEC asked Twitter to review was from an account that claimed to have “voted multiple times in a federal election” and that the results of an election are “rigged”. Others encouraged multiple voting and claimed this was a “common practice” or suggested “referendum ballot papers are already ‘all filled out and ready to be counted’”.
On 15 May, the AEC wrote to Twitter to express its frustration with the repeated refusals to remove content. “Disappointingly, according to our records, nil actions have been taken against any of the below after significant time processing,” it said. The document shows some of the appealed referrals had no response for up to 27 days.
Twitter has faced criticism over the proliferation of false claims and abuse during the Indigenous voice to parliament referendum, after new owner Elon Musk cut staff from its safety team.
Twitter’s terms of service bans content around elections “that may suppress participation, mislead people about when, where or how to participate in a civic process”. It says it may remove offending posts from timelines, among other measures.
No bots, but misinformation spreads
Tim Graham, an associate professor of digital media at the Queensland University of Technology, examined 246,000 tweets about the voice to parliament between March and May 2023.
While accusations continue to fly that both sides of the debate are using bots, Graham did not find evidence of bot-like behaviour in his two-month sample. He did find, however, that accounts sharing no content were more likely to be recently created.
His analysis, which is yet to be peer-reviewed, shows a spike in account creation in November 2022, around when Musk took over Twitter and culled staff across the company – including in its Trust and Safety team as well as employees in Australia. About 285 accounts in the dataset were created in November.
Another spike in account creation occurred in April 2022. While it is not clear what is behind the spike, Musk became the largest shareholder in Twitter around this time. Many of the accounts created in these two spikes, examined by Guardian Australia, also shared anti-vaccination content and other conspiratorial claims.
“A lot of the moderation procedures and resources that were there before Musk are not there any more,” Graham said. “The effect that this has had is that events like the referendum become flashpoints for trolling, flashpoints for conspiracy theorising.”
Graham’s analysis also found that debate about the voice on Twitter was likely to be driven largely by a “small but very active” core of participants. Of the 32,453 accounts examined in the dataset tweeting about the “voice to parliament”, the top 100 accounts sent one in 10 of all voice-related tweets he examined.
Elise Thomas, senior OSINT analyst with the Institute for Strategic Dialogue, is concerned about the absence of research of social media activity around the voice beyond papers like Graham’s.
The situation on Twitter is likely to have “changed dramatically” since the data was collected earlier this year, according to Thomas. “A lot more people are tuning into the conversation since then,” she said. “It’s become much more aggressively polarised. It’s much more vitriolic.”
Twitter responded to a request for comment with an automated email: “Busy now, please check back later.”
