People Are Terrible at Identifying Manipulated Photos, Study Shows

People are awful at distinguishing between real images and those that have been altered, according to a study published Monday in Cognitive Research: Principles and Implications. This doesn’t bode well for the ongoing fight against hoaxes and other misleading information online, especially when such misinformation can be all the more insidious when paired with a convincingly doctored photo.

A research team from the University of Warwick gave participants a series of images, half of which had been altered in some way. The researchers found that people could tell when a photo had been changed only 60 percent of the time, and they were able to identify what was specifically wrong with the picture only 45 percent of the time.

“This has serious implications,” said lead author Sophie Nightingale in a statement, “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.”

The researchers compiled a set of 10 photos from Google Images and created 30 fake versions of them using various styles of alteration, including “physically implausible” and “physically plausible” changes. The 707 participants saw five of the original images, altered versions of another five, and were then asked to identify which of them had been changed and which hadn’t.

When participants correctly guessed that a photo was fake, they were asked to select the region of the image that contained the problem upon a grid overlay. People correctly identified the altered region only 45 percent of the time.

Try it yourself, as you may be more easily manipulated than you think. Can you tell what’s wrong with this photo?

Now here’s the original image, along with a progression of all the ways the researchers altered it.

That people are able to tell real from fake only a bit more than half the time is concerning, given that hoaxes and misinformation have influenced elections and inspired violence in the past year.

This study is particularly concerning because photos are often used as evidence in court. “Jurors and members of the court assume these [fake] images to be real,” said co-author Dr. Kimberley Wade, “though a manipulated image could go undetected with devastating consequences.”

The researchers completed a second, related study, in which they asked 659 participants to locate the altered region of a photo without first asking them to distinguish between doctored and undoctored photos. People were better at finding the problems in this scenario. They were successful 89 percent of the time.

Jumping from 60 to almost 90 percent accuracy may sound reassuring, but let’s consider what these results mean. What they may indicate is that people are better at finding a problem after they assume that one exists rather than deciding whether a problem exists in the first place. But that isn’t the situation for most users online, who typically have to rely on their own judgment to discern whether or not content is real and may well assume what they read is true unless specifically told otherwise.

Luckily, some platforms — including Facebook — are testing out methods for inhibiting the spread of such fake news, either by removing it completely, demoting it within algorithms, or presenting alternate, more reliable sources.

For the researchers, the stakes couldn’t be higher. “Images have a powerful influence on our memories,” said co-author Dr. Derrick Watson, “so if people can’t differentiate between real and fake details in photos, manipulations could frequently alter what we believe and remember.”