How an international troll network tried to meddle in the Dutch election

Surprisingly large network

What prompted the research was a series of eye-catching reposts about Dutch politicians. Many of the accounts in question shared outspoken political content, mostly intended to drive polarisation or to raise the profile of certain candidates. Some expressed support for individual politicians, while others were calculated to fan social division. Broadcaster RTL looked into the posts, identifying hundreds of suspect accounts in a single day. Their research team analysed a total of about 550 accounts on X (previously known as Twitter) and all their posts. It was a huge undertaking, because absolutely everything had to be examined to provide a basis for reliable conclusions. However, the team was also aware that what they were looking at was just the tip of an iceberg. Thousands more accounts remained unseen.

No amateur campaign

The investigation was aided by new functionality on X. The platform has recently started to disclose where account-holders are based. “As a result, we know for sure that the posts came from Nigeria, Ghana and Ivory Coast,” explains Robert van der Noordaa at Trollrensics. Using specially adapted software, the team has now found tens of thousands of new accounts, most based in those same 3 African countries. Both purchased accounts and new profiles were involved, some reposting hundreds of times a day.

However, it remains difficult to be sure who’s behind the network. The accounts generated thousands of likes and reposts. The geographical origin was clear: much of the activity was traced back to Nigeria, Ghana and Ivory Coast. “That tells us about the infrastructure, but not who’s pulling the strings,” says Pieter van Boheemen. It’s unlikely to be a few unconnected individuals, though. The network is too large, too well coordinated and too expensive to maintain. While hard evidence of control by a state actor or the like is lacking, everything points to an organised operation, not a spontaneous or amateur network.

TikTok as the hotbed for interference

The research wasn’t confined to X. Van Boheemen took a detailed look at TikTok, studying the livestreams of the political parties and their leaders. And he describes what he observed as worrying. During the livestreaming of a speech by centre-left leader Rob Jetten, death threats and antisemitic comments were openly posted without prompting intervention. “Lots of people will have seen the posts, certainly including young people. For a growing group of current and future voters, TikTok is an important opinion-shaping forum.”

It’s also an easy platform to manipulate. Accounts can be recreated in a few minutes, and the algorithm gives priority to new profiles. As a result, videos with political messages can circulate extremely quickly, even if the poster doesn’t have a lot of followers. In the days leading up to the elections, numerous new accounts sprang up, encouraging people to vote for particular parties. “You don’t need political advertising if the algorithm boosts the reach of your content.”

So, were the elections fair?

The researchers emphasise that there was no evidence that the elections were unfair. However, they were threatened by the activities described. It’s difficult to measure the effect of online influence. But the scale gives rise to the suspicion that it’s part of a deliberate campaign. “No one would invest all the time and effort that an operation like this requires if they didn’t think it would have any effect,” says Van der Noordaa. He believes that social media has an increasing influence on how people vote.

A further complication is that disinformation always has an effect, even if people know about it. “If people aren’t aware what’s being done, disinformation works. If they are aware, they become dubious about the whole process,” Van Boheemen explains. As a result, it’s really hard to assess how much harm is done to society.

Lax moderation

TikTok, X, Facebook and Instagram say they have sophisticated systems for detecting abuse, but the checks appear to be minimal in practice. “Even searching for simple keywords doesn’t appear to work,” adds Van Boheemen. The question therefore arises, why is it always left to civil society groups and journalists to highlight problems, when surely it should be up to platforms to demonstrate in advance that their moderation is in order? There is no objective assessment.

Van der Noordaa points to the ease with which locations were made visible. “They could easily have done that much earlier.” He is unimpressed by the social media companies’ defence that they delete large numbers of fake accounts.

The role of SIDN Fund

SIDN Fund made the study possible by paying for software, data storage and personnel. The project involved a 7-strong team that spent months analysing data. Without financial support, that simply wouldn’t have been possible. “We’re incredibly grateful to the Fund. It’s very intensive work that can’t be done alongside other things.”

Because of the Fund’s support, the scale of the foreign influence was revealed for the first time. And the work continues. An increasingly detailed picture of the network is emerging. Many accounts look Dutch at first sight, but are actually run from Nigeria, Ivory Coast or Ghana. “The problem hasn’t gone away,” says Van der Noordaa. “So it’s important to maintain our surveillance.”