Skip to Content

Support MinnPost

UCare generously supports MinnPost’s Second Opinion coverage; learn why.

People are very bad at detecting fake photos, study finds

People are very bad at detecting fake photos, study finds
One of five photographs, taken by Elsie Wright in 1917, showing Frances Griffiths with the alleged fairies.

If you’re concerned about the proliferation of fake news (and who isn’t), here’s some news (not fake) that will undoubtedly add to that concern: A new study reports that people are not very good at detecting fake images — photos that have been manipulated.

The study, which was published Tuesday in the journal Cognitive Research: Principles and Implications, found that people could identify a fake image of a real-world scene only six out of 10 times, on average.

And they could pick out the exact spot that had been manipulated in the photo less than half the time.

“Our study found that although people performed better than chance at detecting and locating image manipulations, they are far from perfect,” said Sophie Nightingale, the study’s lead author and a doctoral student in psychology at the University of Warwick in Coventry, England, in a released statement

“This has serious implications because of the high-level of images, and possibly fake images, that people are exposed to on a daily basis through social networking sites, the internet and the media,” she added.

‘Nearly anyone’ can do it

Fake photos are not new. One of the most famous — and earliest — cases was the Cottingley fairy hoax of 1917.  Two young English cousins took photographs of fairies dancing in their garden. The fairies were actually paper cutouts copied from a children’s book, but the young girls claimed — convincingly, apparently — that the creatures were real. The photos became a sensation, and the people fooled by them included even Sherlock Holmes’ creator, Arthur Conan Doyle.

Today, of course, digital photography makes the creation of fake photos much, much easier. Indeed, as the authors of the current study point out, “The growing sophistication of photo-editing tools means that nearly anyone can make a convincing forgery.” 

That development raises all sorts of concerns, for it “has consequences across almost all domains, from law enforcement and national security through to scientific publishing, politics, media, and advertising,” write the researchers.

A pair of experiments

Previous research has shown that people are poor at detecting geometrical inconsistencies within a scene, but little was known about whether this deficiency extended to real-world scenes.

To find out, Nightingale and her colleagues conducted two separate experiments. For the first experiment, they recruited around 700 online participants, mostly men, with an average age of 26. The participants were shown 10 separate color photos of people in ordinary scenes (such as a man standing on a residential street and a woman standing next to a British canal). Five were “real” photos, while the other five were “fake.”

The fake images were manipulated either in a plausible way (a person’s wrinkles were airbrushed away, for example, or a trashcan was inserted into a street scene) or in a physically implausible way (such as by changing shadows in ways that implied two suns). Some images included multiple manipulations. 

The participants were then asked, “Do you think this photograph has been digitally altered?’ They successfully identified the fake photos only 60 percent of the time. And when those who had correctly identified the fraudulent photos were asked to locate the precise place in the photo that had been manipulated, they succeeded only 45 percent of the time. The participants were somewhat better at identifying physically implausible changes in the photos than they were at identifying the plausible ones.

In a second experiment, a group of 659 online participants were asked to identify exactly what was manipulated in the photos whether or not they thought the image had been altered. They were able to pinpoint the problems in the photos about 56 percent of the time.

This finding suggests that people are better at recognizing something fake in a photo after they are told the photo has been manipulated. 

Needed: better detection tools

“People’s poor ability to identify manipulated photos raises problems in the context of legal proceedings, where photos may be used as evidence,” said Kimberley Wade, a study co-author and a cognitive psychologist at the University of Warwick, in the released statement. “Jurors and members of the court assume these images to be real, though a manipulated image could go undetected with devastating consequences.”

“We need to work to find better ways to protect people from the negative effects of photo manipulation, and we’re now exploring a number of ways that might help people to better detect fakes.”

In the meantime, the rest of us will need to stay alert for fake photos, particularly on social media. 

FMI: The study can be read in full on the website of Cognitive Research: Principles and Implications. The Washington Post has put up a short “fake or not-fake” quiz using the study’s photos, which you can take on that newspaper’s website.

Get MinnPost's top stories in your inbox

Related Tags:

About the Author: