Six takeaways on digital disinformation at EU elections

The Oxford Internet Institute noted that only a few social media items related to the elections were 'junk news'.

The University Duisberg-Essen showed that the best-performing Twitter accounts in the UK and Germany were not associated to extremism or disinformation, but rather those of well-known pro-European politicians.

At the other end of the spectrum, the campaign group Avaaz indicated that extremist and disinformation content was viewed hundreds of millions of time across the EU.

Commentators claimed that "in many places disinformation had drowned out the messages of mainstream parties".

How to understand this wide spectrum of opinions by experts who purport to work with sophisticated techniques of social media analysis?

1. Which platform, exactly?

'Social media' is not social media.

Many studies only look at one platform and some talk about their results as if it concerns all social media. Journalists mostly failed to make a difference between "social media" and a specific channel. Thus, Germany's primetime TV news reported that 47 percent of political social media discussions were related to the extreme-right AfD party, when in fact this was the case only for Twitter, used by 4 percent of Germans - and often these posts were negative.

2. Facebook vs Twitter

Most studies only look at Twitter, because it is more open about research on its data.

They found problems, in particular the artificial boosting of content through fake accounts, but none reported massive challenges.

Only a few examined Facebook more deeply, among them Avaaz. Its findings where indeed troubling, especially the quick rise of pages that spread extremist content under the guise of being media outlets.

According to Avaaz these accounts often reached more people than official pages of extremist right-wing parties across Europe. Facebook, which regularly reports on take-downs of pages, remained strangely silent on these phenomena and did not confirm Avaaz' claims that it took them down. Measures it had announced to stifle the growth of such pages apparently did not work.

3. No hard-right 'breakthrough'

A complete upset of the European Parliament election outcomes did not occur.

The authoritarian extreme-right did not suddenly take over the European parliament. Such fears were hysterical from the start, fuelled by the widespread perception that Trump's surprise election and the Brexit vote must have been the result of online disinformation.

Both polls has however been close calls. A few percentage points in three US states or across the UK would make the difference. Whether it was disinformation that made that difference has never been proven. Most studies on social media influence on electoral choices have not found a significant impact.

The European elections were at no point a close call, where the extreme right might have been close to winning an absolute majority to 'take over the EU'. The threats to democratic discourse online are more subtle and more long-term. They are more about an insidious change in public perceptions over time, drip-by-drip. If such threats changed the election results by one percent or two it may not upset an election, but it would still be a serious worry.

4. Better prepared

We are further than we were in 2016 in understanding and responding to threats to online discourse.

Investigative journalism and policy responses have born some fruits. The story of Trump's election was associated with targeted advertising, partly based on the data of Cambridge Analytica. In public perceptions this was the dark science that explained his seemingly inexplicable victory.

In the European elections, Twitter, Google and Facebook published political ads online, so they are not dark anymore. Most ad money was spent by the traditional European parties, not by extremists. Thus, one of the most important explainers of the Trump victory, played no role here.

5. Messy archives

Despite this progress, a lot more can be done.

For a starter, Facebook, Google and Twitter, who praise themselves for good 'user experience' could do a lot more to make the archives of political ads a good user experience.

These archives were thrown together at the last minute, difficult to search and not intuitive. This aspect could be easily regulated.

Also, it remains troubling that Facebook claims to have 30,000 staff dealing with security and integrity on the platform and yet small organisations manage in short time to identify highly-problematic pages.

6. Standardise reporting criteria?

The field of social media analysis needs to become more professional and transparent to avoid becoming a feature of disinformation itself and to weed out superficial, headline-hunting reports.

At a minimum any report on social media behaviour should state which platforms it observed, in which timeframe, how many posts, which tools it used to do so and explain its terms (i.e. what is "extremism"?).

Many reports do not conform to such basic standards of transparency. Without such transparency behind published evidence, there is no basis for a facts-based policy debate.