Social media users are transforming the information world, especially the growing legions armed with photo and video capabilities on their smartphones.
But inauthentic and/or manipulated images have emerged as a real problem. Fortunately, new tools are available to spot these fakes, and squash their propagation before they take on a life of their own.
The root of the problem
Whatever story words may tell, images often far better accomplish the task.
It is becoming more and more rare for interesting phenomenon near any populated area to go undocumented, and real-time accounts from the hardest hit regions of a disaster are increasing.
Unfortunately, there are built-in incentives for disseminating illegitimate photos.
Posting a viral breaking news photo or nature scene can mean a whole lot of marketability when it comes to retweets and shares.
Hurricane Sandy and beyond
— Andy Carvin (@acarvin) February 19, 2013
Last year, Hurricane Sandy created a social media storm as it barreled toward the Northeast megalopolis, targeting the most populated part of the United States with its most extreme impacts.
As it was a type of a storm that hadn’t been experienced in recent memory in the region, it was all the more difficult to make quick judgment calls on fake or misleading photos being circulated including by The Washington Post.
Advanced tools may eventually solve the problem. A paper published earlier this year by Aditi Gupta et al. claimed that up to 97 percent of Sandy fakes could have been identified automatically.
Among other findings, Aditi Gupta et al. highlight one glaring issue when it comes to fakes running rampant: 86 percent of non-legitimate Sandy imagery tweets came from retweets.
Getting the false imagery into the hands of an influential user is typically the catalyst to making a fake go viral. In the study, the top 30 users out of a sample of 10,215 resulted in 90 percent of the retweets of Sandy fakes.
Additionally, Gupta and company wrote “… in cases of crisis, people often retweet and propagate tweets that they find in Twitter search or trending topics, irrespective of whether they follow the user or not.”
The researchers hope to get a working form of their model available as at least a Web plug-in at some point in the future. Certainly a much-needed application in the breaking — and non breaking — news environment.
Why? Nature is in our nature.
Multiple Earth picture accounts on Twitter have upwards of 1 million+ followers, with each tweet featuring old — often mislabeled and unattributed — pictures that receive hundreds or thousands of retweets. Sharing the raw beauty is part of enjoying it, but many sources are unknown and fakes get plenty of play.
Combating fakes without a special algorithm
The Associated Press has been a leader in the verification process when it comes to social media. They have shared several versions of a multi-step process which consists of, among other things, confirming with the original source, comparing with their own reporting, and checking the source’s social media history.
Time is of the essence when it comes to the rapid pace of the social superhighway, but to a valued news organization like AP, getting it right is more critical than being first.
Still, in the end, your average social media user doesn’t often have the time, capabilities or interest to go through a rigorous verification process.
Examining weather photos, there are often clues a careful observer can find. It can be done without spending a lot of time, and while not as fool proof as the process employed by AP, it’s better than nothing.
Some of the most simple things to spot and match up are:
Time of day (location of sun, shadows); Season (are the trees full of flowers, dark green, turning red, bare?); Geographical features (would a waterspout in Tampa have mountains behind it?); Type of weather “expected” (tropical systems aren’t known for their beautiful structure as much as their squally grayness); Other evidence (including corroborating photos, video, reports).
When it comes to the source of the fakes themselves, there are often some clues: Did the account just pop up? Any signs pointing to the person being the originator or directly connected? Sometimes it’s as simple as checking a profile on Twitter, Facebook or YouTube to gather a few facts.
Part of the verification process is recognizing quickly if it’s just an old re-circulated photo. Google Images (click the camera icon in the search bar) or TinEye are two examples of sites that can be used to analyze the host image URL or direct file upload. This method is not fool proof, as it can still miss imagery not widely used or heavily modified.
In a vast majority of cases, the consequences of a fake have not been serious. Sharing a fake is likely something that most of us who use social media frequently have accidentally partaken in. With a little extra effort, we might be able to cut down on the occurrence while waiting for computers to better solve the problem for us.