
With the COVID-19 worldwide crisis, an unprecedented wave of misinformation has been observed online. Unfortunately, disinformation has swept across continents confusing the public about fake treatments, false rumours about the disease, exacerbated fears and conspiracy theories.
This “disinfodemic” surge brings new challenges for social network analysis to help understand more specifically how disinformation about the pandemic spreads, by which actors and websites it is promoted, how rumours and fake news are spreading across countries. Therefore, journalists, fact-checkers and health authorities are overwhelmed by the amount of disinformation and rumours. Due to the amount of work they’re facing, they need to be able to focus on the more impactful disinformation in order to debunk it and avoid its spread, and potential dramatic consequences.
During the EUvsVirus Hackathon, the Twitter SNA team led by EU DisinfoLab and AFP Medialab has been working to develop new useful tools to assess impact of disinformation and provide easy access to existing resources. We have built upon an incipient Twitter social network analysis feature to be released in the InVID-WeVerify verification plugin (used by 25k journalists and fact-checkers worldwide),
From the hackathon brainstorm, we are also planning to introduce new features to reveal accounts and websites spreading disinformation as well as highlighting possible disinformation cross-postings around networks.
- Provide easier access to already existing fact-checking
We’re currently facing two important shifts in the way disinformation is produced and debunked:
1 – We are facing a globalisation of the disinformation trend. Research has shown that similar narratives or conspiracy theories were spread at the same time in different countries. This trend also includes the fact that local images are de-contextualised and re-contextualised in different countries, according to local specificities (culture, beliefs, political issues, etc.)
2 – At the same time, even if growing, there is not enough coordination of debunking efforts from newsrooms, as they’re very often limited by language or territory coverage. However, many organisations have started to pull their efforts together and to provide open-source collection and/or databases of disinformation, such as IFCN, FirstDraft, Boom or Altnews.
Our first solution was to provide a service that could check if similar disinformation has already been debunked. Many journalists have been are struggling amid the Covid 19 crisis. This is partly due to the mass of disinformation being spread in the disinfodemic. To ease up their work, we set up a search engine that could let them know if a similar disinformation they’re observing has already been debunked by another newsroom.
Figure 1: Results from hashtag #FilmYourHospital an expression used by conspiracy theorists to claim the Covid19 pandemic is not real and is fabricated by elites
The search engine is a Google Custom search engine, configured to search directly in a public database of disinformation in order to find already existing debunking articles. This search engines (also available in French and in Spanish) provides results for the following resources:
Covid-19 debunk search in English:
https://cse.google.com/cse?cx=000556916517770601014:i0sxs8kmmzr
Covid-19 debunk search in French:
https://cse.google.com/cse?cx=000556916517770601014:jff3t29fxeq
Covid-19 debunk search in Spanish:
https://cse.google.com/cse?cx=000556916517770601014:3imsvcumevz
- Estimate impact and spread of disinformation
The second most important task for newsrooms is to be able to estimate the impact of the suspicious content they want to verify. . If they debunk it too soon, they might give more visibility to only marginal rumours. If they’re doing it too late, too many people might already believe in it and their debunk arrives too late.
This is why ,in order to only debunk stories that are starting to be massively shared, it is crucial to evaluate the impact of disinformation:
- how many times it is shared?
- where? on which platforms ?
- by whom?
The goal is to provide quickly to journalists, fact-checkers, civil society, means to retrieve meaningful data about the spread of COVID-19 misinformation such as:
- provide cross-network results to see if the disinformation they observe is present on multiple social platforms
- automatic information retrieval of lists and graphs of accounts / websites / hashtags promoting rumours and fake news, in order to find their origin (accounts, countries).
To address the first issue, during the hackathon we have set up a search engine providing occurences of a query (for instance a hashtag or a combination of terms such as “Coronavirus 5G”).
Figure 2: Even after its debunk, the hashtag #FilmYourHospital is still used. Through our search engine we can find several occurrences on Telegram and YouTube for instance, allowing journalists to research and investigate this thematic.
This custom search engines provides public results on 10 social networks:
- Youtube
- Telegram
- Vkontakte
- Dailymotion
- 4chan
- 8chan
- Gab
Cross Network search
https://cse.google.com/cse?cx=000556916517770601014:k08mmqlnmih
To address the second issue, we’re developing within our plugin an improved Twitter analytics module that crawls quantitative data on how much a disinformation hashtag has been used and a visualisation of clusters of Twitter accounts spreading it.
By providing both quantitative and qualitative data, we want to support newsrooms in their assessment of sourcing and evaluating the impact of disinformation, and thus to be able to have stronger arguments when debunking it.
This Twitter Social Network Analytics tool will allow newsrooms to:
- Search a hashtag or an expression associated with disinformation on Twitter
- Visualise analytics around hashtags (number of tweets, retweets, keywords most associated, most influential accounts interacting with the hashtag, etc.)
- Visualise on a map and timeline how this hashtag has evolved in time, and what were the main clusters of discussion
You can see a demonstration of the whole tools hereunder:
Want to know more?
These new functionalities will be available in the next version of our verification plugin. Download the current version here:
If you want to test some of these functionalities? Contact us
Leave a Comment
sing in to post your comment or sign-up if you dont have any account.