We rely more than ever on images to evaluate claims about the world, from images published on social media and news websites to images of research results published in the scientific literature. We come across so many images per day, so how can we fact-check images in a way that gives us the most bang for the buck? How do we know images are legitimate and have not been deceptively manipulated? What are some of the ways that images can be manipulated?
This article covers these questions and also provides several different methods that anyone can use to fact-check stories with misleading photographs, debunk viral images and help increase the intellectual integrity of image use online.
How can images be deceptive?
Images are powerful forms of media. They can provide the smoking gun, capture an ongoing situation or give emotional context to a news story. However, they can also be misused in deceptive and equally powerful ways. What are some of the ways that images can be created, manipulated or altered to give a misleading impression of the situation?
Some images can be staged by using actors that look like famous people or other people with power in order to create false associations between those people and some large context that is deemed beneficial or harmful. These photographs are genuine and not digitally altered, but the content of the photos and the context in which they are presented are deeply misleading. For instance, there is a photo going around that depicts someone resembling Donald Trump hugging extremists from the Ku Klux Klan in front of a burning cross. In reality, this is a staged photograph taken by Alison Jackson for her book Private that contains staged photographs of actors portraying different powerful people in compromising situations. She is also behind a staged photo appearing to show Bill Clinton naked on a bed while having his back rubbed by a woman. Both of these photos are taken of actors merely pretending to be Donald Trump and Bill Clinton. None of them show the real politicians involved.
Other images can have important elements removed in an effort to push historical revisionism. Communist dictators such as Vladimir Lenin or Joseph Stalin and others engaged in this technique frequently to make people such as Nikolai Yezhov, Leon Trotsky or Lev Kamenev disappear.
Yet other images are digitally altered to create attention by e.g. combining two other images to create a misleading experience. A classic example of this is the infamous shark attack where someone put together a photo of a rescue helicopter and a shark jumping out of the water to make it appear as if the shark attacked the rescue worker on a helicopter ladder.
Image manipulation can also be weaponized to push bigotry against ethnic and sexual minorities. An image of two hairy and almost naked men in a Pride parade was digitally altered to include two small children in an effort to spread fear and disgust over gay people and pride parades. The image includes a hand signal for stop and the French text “do not touch”.
Some scientists engage in scientific misconduct by manipulating images to make their results appear more credible. This involves duplicating, rotation, zooming or other image alteration techniques to create results that do not actually exist in reality or improve existing results so that they look better than they really are.
Step #1: Know when to fact-check images
People are bombarded by hundreds or thousands of images online every day when they scroll through their social media feeds, visit news media, meme websites or blogs. There is virtually no hope in being able to manually fact-checking all of them and this is probably not even desirable since it would take an incredible large amount of time and many images are stock photos or not relevant to the claims being made in the story.
Some stories have images that are only tangentially relevant. Some stories show images of buildings, locations, maps and so on that accompany stories just to provide some visual content because just posting text might seem too boring. Other stories have images that are part of the evidence for a claim, but supplemented with further arguments and evidence inside the text. Other stories rely exclusively on the alleged validity of an image. So how can you tell which images are worth fact-checking and which ones are not? Which ones are crucial to explore further, and what investigations are largely a waste of time?
How do you get as much bang for the image fact-checking buck as possible?
One reasonable approach is to create a method for identifying those stories whose validity rely exclusively or strongly on the validity of the associated image. That way, it is enough to fact-check the image to fact-check the story. If the image is fake, staged or manipulated, the story itself is largely bogus. As the images associated with a story becomes less and less relevant for the truth of the claims, it becomes less and less useful to fact-check the images.
Step #2: Check previous fact-checking material
If it is an image that is vital for the story and the story has already gone viral days or weeks ago, chances are that someone has already fact-checked the image. Make sure you check credible fact-checking organizations and major news media to see if the image has already been debunked and double-check their work so that you do not fall for a botched fact-checking job. There are many fact-checking organizations out there that regularly debunk fake images, such as Snopes, PolitiFact and FactCheck.org.
How do you fact-check a fact-checker? Simple. They are usually transparent with their methods and sources, so you just repeat what they have done.
Step #3: Reverse Image Search
Some search engines such as Google allows you to use an image to search for similar images. Go to images.google.com and click the little camera on the right of the search field.
This opens up a small menu where you can choose between uploading an image to search with or linking the URL from an existing image that you have found online.
The Google algorithms will make a best guess on what the image is and show a list of search results that include that image or images that are visually similar to it. This is a quick way to find original images if you find an image you suspect is digitally altered from preexisting images that exist online. You can also restrict the search by other parameters such as year. This allows you to step back in time year by year or month by month until you find the place where the image was first published online. This does not just allow you to find the original image, but also the context in which the original image appeared. Even when the image itself is not digitally manipulated, the original context in which the image first appeared can be completely different. The reverse image search method allows you to find out if this is the case.
Step #4: Forensic image tools
Some image formats, such as JPEG, save metadata on what has happened with an image, from what camera and resolution was used to what things have been done in image editing software. Inspecting image metadata can provide some clues about how the image has been edited and potentially manipulated in a deceptive way. However, this tool is of limited utility since images are often spread around the Internet, saved to different formats, transformed into screenshots. However, it is a quick method that will sometimes work. There are many image metadata viewers online, such as Jeffrey’s Image Metadata Viewer.
More sophisticated tools will allow you to see differences in compression levels in different parts of an image. Basically, when someone edits a photograph and adds or removes content, the file is typically saved with different compression. So the manipulated areas will stand out compared with the original parts of the image that has not been altered. This is called error level analysis. One tool that can do this is called Forensically and can be used online. A tutorial for how to use this program can be found here.
Another tool available for forensic image analysis is Droplets. It is produced by the Office for Research Integrity (ORI) at the U. S. Department of Health and Human Services. It allows you to spot areas of a photo that has been made brighter or darker, see if different images come from the same source and compare to images that may contain similar features. ORI has even more advanced tools available for the professional users. A crucial insight is that image manipulation can be detected not just on the items in the foreground, but also probed by comparing patterns in white or noisy background. There is a fascinating interview with a data integrity analyst who spot manipulated images for a living available at Nature News and Comment.
You can also read an entire textbook on digital image forensics written by Hany Farid if you want to improve your background knowledge on the subject. A cached version of the textbook is available here.
Identify stories that crucially rely on the validity of some photo to get as much bang for the image fact-checking buck as you can. Check existing fact-checking sources, do reverse image searches and forensic image tools to check them. Write up your analysis and post in on the Internet so that others also can benefit from it. While you are at it, submit a tip or a link your analysis to more prominent organizations and see if they pick it up.