by Ferre WOUTERS for FactCheckNI (25 March 2019)

At the first day of Belfast’s Imagine Festival 2019, Amnesty International held two events, both presented by Milena Marin, who leads Amnesty International’s Digital Decoders platform. It started off with a talk explaining how Amnesty is using technology to expose human rights violations, followed by a workshop where participants were acquainted with more detailed methods and tools that Amnesty uses for digital verification.

Beyond clicktivism

Amnesty was founded almost 60 years ago, but along with recent digital innovations, a lot has changed in the way they operate. While at the beginning it was very hard to find information on cases of human rights abuses, nowadays they are confronted with an overload of such information. Among the millions and billions of posts shared every day on social media are those with content of human rights violations.

Of course, we can’t just assume such user-generated content is all accurate. Amnesty works hard to investigate the accuracy of recorded images and sounds of claimed human rights abuses.

Digital investigations are done by a verification team, where students work together with research experts. They are assisted by a platform of crowdsourced digital volunteers operating as “Decoders” — people from all over the world who help these investigators by sifting through pictures and documents from their own computers. This platform makes it possible for engaged people to valuably contribute to human rights research that goes beyond signing petitions, without having to be an expert on those issues.

One example of their investigation gained much traction last year. Amnesty verified a video that circulated on social media that showed footage of Cameroonian soldiers carrying out extrajudicial executions of four women and children. The Cameroonian minister of communication dismissed it as fake news. BBC News Africa made a short documentary explaining how Amnesty gathered evidence for the credibility of the contested video.

Masterclass: Digital verification for human rights

Marin began the second part of the session by stressing the importance of social media verification. Although you can share a video with the best intentions, a little inattention can cause you to spread misinformation. It is all too easy to miss things, because you don’t see things that you’re not looking for.

Therefore, it is necessary to carefully examine every piece of shared content. Marin pointed out a crucial step is to be aware of your own biases; because if you want to believe that something is true or false, it becomes really easy to persuade yourself it really is true or false.

One example is the spread of fake war videos. Often, such content is claimed to be real drone footage of bombings, while in reality they turn out to be footage coming from video games. The Russian Ministry of Defense, for example, shared images from a mobile phone game trying to show collusion between the United States and Islamic State combat units.

Marin went on by explaining the “verification puzzle”, where each piece is needed to get the whole picture of social media content. For discovering every piece of the puzzle, different techniques are applied. A handy set of various verification tools can be found here.

Source

To check if the content is original, Amnesty uses “reverse image search”. This way, you can investigate the original source of an image. The same technique is used for videos by extracting key frames from it. The easiest tool to do so is by using Google Images, where you can search by image — either upload the image you want to check or enter its url. A more advanced tool they use to find sources of content is Intel Techniques’ search engine.

Time

To find out if images are taken at the time people claim it to be, Amnesty checks its metadata, which is data about other data, or exif data, the data of images. This data can be traced by exif viewer. Unfortunately, exif and metadata are stripped out of content on social media for privacy reasons.

Therefore, verifiers have to be more creative to check the date of social media content. One useful way to do so is by looking at the weather. What kind of weather was it the date the content is claimed to be from? If it is raining on a video, but the day it is said to be recorded was a sunny day, it is likely that the recordings were taken at a different time. Using sun calculator, you can verify the date and hour by tracking the position of the sun through analysing the shadows on the image.

Location

Are images really showing the place claimed on social media? Looking at all sorts of visual clues there are on an image, you can find the exact location by spending hours searching through Google Earth and Street View. Verifiers carefully pay attention to landscapes, mountains, trees, and buildings, but also cultural clues like the clothes people are wearing.

Finding locations is something you can get better at with practice. Want to try this yourself? Play GeoGuessr, and fact check where you think you are. (Be careful though, as Marin warned, the game gets very addictive.)

Proactive verification

I asked Marin if Amnesty has methods to catch content before it gets widespread online. She answered that this is indeed what they are aiming for. Their verification team proactively are doing research to prepare themselves for moments when violence is expected, e.g. when elections in Africa take place.

They also use tools like Data Miner, to track social media messages containing content on human rights abuses. Tweetdeck makes it possible to proactively search for new tweets on a certain topic.

Amnesty International is not a technology company. Marin said ideally other organisations should develop the tools for digital verification. Since this is not always the case, sometimes they have to take the initiative themselves. Citizen Evidence Lab is an online space created by Amnesty, where practices, techniques, and tools are shared for authenticating user-generated content concerning human rights.

Conclusion

The dedicated work Amnesty is doing in verifying social media content basically consists of fact checking evidence of human rights violations. Their fact checking is specific both in form, by investigating videos and images, as well as in content, namely claimed human rights abuses. It reminds us how diverse the world of fact checking is, and how serious and important it can be.