Fact checking and verification at WDR – News – WDR – News

Editorial staff in professional media receive images and videos every day that convey news and claims about current events. It is not always clear whether these are real recordings or images in the right context. In addition, complex topics often have to be dealt with under great time pressure – even in WDR.

The sender therefore has a verification and fact-checking project team that can respond to requests flexibly and quickly. With the war in Ukraine, requests have increased dramatically, reports Dustin Haack, a scientific documentarian in the WDR verification team. Suddenly, a flood of videos appeared that showed supposed acts of war. Not all of them were real. Some of them even turned out to be clips from computer games.

We spoke to Dustin Haack about his work, misleading images and rules for exposure.

WDR: How does the material come to you and how do you then handle it?

Dustin Haack: If someone from WDR finds a video or receives one and thinks it is relevant for broadcast, they can send it to us before publication. We believe we can give an initial assessment within 45 minutes.

We then split up into teams and check different aspects. For example, one person checks whether the location is correct, another looks more closely at the context, and a third tries to research who actually published it. Then we puzzle over it from different angles. The videos themselves are often not wrong, but taken out of context. That’s why the chronological order is also important.

WDR: Can you give a current example?

Dustin Haack: We received some videos about this in August 2024, when there were street riots in Great Britain. Right-wing extremists, but also Muslims, were there and there were also some riots. Then we had to find out whether the videos were really current. One of these videos, for example, was already ten years old.

First, we try to find the place where the incident is said to have taken place. If we have the street and the location, we can use various map tools to trace it and find the rioters’ route ourselves. And we can see what it looks like there today. The images you see on Google Maps, Apple Cards or Mapillary were all taken at different times. This way, you can find out how likely it is that it is a recent photo and whether the location is really correct.

The special thing about the images of the riots was that there were twelve videos, but not twelve riots. There were only three, but they were always filmed from different perspectives. And then at the end you find out: It’s a group of maybe 20 people who are screaming through the city center. There are an incredible number of videos circulating of this, which suggest that thousands of people are tearing the city apart.

WDR: Fake images are now also created using artificial intelligence. Are you actually working with AI to expose AI?

Dustin Haack: We sometimes use it to help. But you can’t just throw in a video and get a clear result of “fake or not”. But there are tools where you can upload photos, images or text, which then indicate how likely it is that a piece of content is man-made or not. The tools are getting better and better, but those who are up to no good are already working on the next fake and are always one step ahead.

WDR: Speaking of fake images – in 2023, two AI-generated images won first prizes in renowned photo competitions. If professionals can’t recognize such images, how can normal users?

Dustin Haack: The images and videos we come across can usually be identified if we know which details to look for. At some point we will no longer be able to do that. That is why my tip is to always critically examine the content of images and videos.

For example: I see a video of Chancellor Olaf Scholz announcing his resignation or a picture of him in swimming trunks on the Spree. Then you should ask yourself: How likely is that actually or what agenda could be behind it? As a rule, the fakes are pretty easy to spot.

WDR: What would you like to tell users when they watch and share?

Dustin Haack: No matter what I see on the Internet, I’m always skeptical, especially if it fits my own narrative. People usually have their own stance or are on a certain side in a conflict. And then I see something that fits. I would take a closer look – especially if I don’t know the sender. And that’s usually the case when things are published by some no-name account.

The interview was conducted by Katja Goebel

On the series “Our life with AI” shows that WDR television will broadcast the program “How artificial intelligence deceives us” on September 25, 2024 at 10:15 p.m.