On July 14, Environment Canada issued a severe thunderstorm watch for New Brunswick — which means atmospheric conditions are favourable for the development of thunderstorms.
The watch mentioned that there was a tornado risk associated with the potential conditions.
Neither severe thunderstorms nor a tornado materialized, but a social media storm ensued leading to the spread of misinformation, unreliable images and a lot of panic.
Extreme weather, natural disasters and health emergencies are prime events for attracting misinformation, according to Sander van der Linden, a psychology professor at the University of Cambridge in the United Kingdom and an expert in misinformation.
“It’s real time,” he said. “There’s urgency. There’s uncertainty. So people are going to try to fill in the void.”
An Environment Canada “watch” is not the same as a “warning.” A watch advises people to get ready for potential severe weather, while a warning tells people to act immediately as severe weather is either happening or will happen.
Jill Maepea, a Fredericton-based Environment Canada meteorologist, told CBC News on the day that the severe thunderstorm watch was issued that the risk of a tornado was low.
Despite this, people on social media were already worked up about a potential severe thunderstorm prior to the watch from Environment Canada with Facebook pages posting, days in advance, warning of the event.
At the time, Maepea said while meteorologists try to predict convective weather in advance, it can develop very rapidly and it isn’t the same as tracking “a large winter storm.”
She said that these storms can also dissipate very rapidly.
“It’s very hard to predict and also track.”
People took to social media to criticize Environment Canada for not putting out a notice about the storm sooner. And even when the watch was posted on that morning, posts then began circulating misinterpreting the information, telling people that tornado warnings were in effect.
Still, there was no tornado warning, or tornado watch for that matter.
Van der Linden said it’s good that there was official communication happening, but sometimes, that information can be interpreted in an ambiguous way if people don’t necessarily know what the information means.
“The initial message that was cautious and scientific might diffuse to being quite alarming,” he said.
When the watches were over, the social media fury didn’t end. Some people were angry that there was no storm to witness at all, while others began sharing images of a storm that never happened.
In one case, an AI-generated image shared on Facebook of a funnel cloud in Fredericton had an “AI content” label attached. Hundreds shared the post and commented on it, with many, but not all, pointing out that the image was AI-generated.
Another photo, posted by an anonymous account in a Facebook comment section, included a huge waterspout seemingly pictured off the bow of someone’s boat with the caption, “Kent county got it hard.”
Another, commenting under a post that referred to the lacklustre nature of the storm, said that Quebec didn’t get so lucky, posting a menacing tornado photo.
Quebec also had widespread weather alerts on the same day as New Brunswick. But this photo, that the commenter claimed was taken in Drummondville, Que., was actually taken during a fatal EF4 tornado that struck Fairdale, IL., in 2015.
Van der Linden said hyperrealistic AI is making it more difficult for people to discern what is real or fake online. He also said that when someone sees an image of a severe weather event on social media, their brain might believe it even if it isn’t the most realistic because of other factors.
Aaloak Jaswal, the head of innovation at the University of New Brunswick’s Data Science and AI Research Institute, said it’s important to recognize that every new communication technology has increased the ease of spreading information. But AI, he said, is the first that has “made it effortless to manufacture … convincing information.”
Jaswal said there have always been rumours, going back to chain letters, then chain emails, then manipulated photographs.
“AI hasn’t invented misinformation … but it has dramatically lowered the barrier,” he said.
“Today, anyone with no technical background, they can just simply type a sentence into a free AI tool and generate a remarkably realistic image in under a minute.”
At the same time, he said he doesn’t believe people are consciously deciding whether an image is real or fake when looking at it, especially during an emergency.
For example, he said if there has been information spreading about a possible tornado, and someone sees an image of a tornado, their first thought isn’t to question the photo’s authenticity but rather to ask whether everyone is OK and whether their family and friends were affected.
“We’re moving from a world where we once verified images to one now … where we increasingly need to verify what the source is, and I think that’s a much more sustainable habit because the technology is going to continue to improve,” Jaswal said.
Jaswal also said that there’s a risk if people are repeatedly exposed to misinformation and convincing AI-generated images.
When there’s a genuine public safety issue, he said official sources are all trying to communicate timely and accurate information, but if people start to lose confidence in what they’re seeing and hearing, they may ignore legitimate warnings or make decisions based on unverified information.
“Nature, it creates enough uncertainty during an emergency that we shouldn’t be adding artificial uncertainty on top of that,” Jaswal said. “Fundamentally, this is not a technology problem, … I think it’s a trust problem.”
Prof. van der Linden said people might also apply a “better safe than sorry” heuristic to the news that they’re hearing in emergency situations.
And while he said this heuristic might make sense in some situations and the trade-off of being overprepared for a tornado as opposed to underprepared might be worth it to some, online fearmongering could have negative consequences if done too often.
For example, he said if there is a real environmental or public health emergency, people might not believe it because of past experiences where misinformation clouded the real information.
“The difficulty becomes that people don’t really know, you know, what to believe anymore,” van der Linden said.
He said one of the best things someone can do is something called “pre-bunking,” which involves warning friends, neighbours and family in advance about misinformation spread.
“You gotta … say, ‘Look, there’s misinformation spreading, here’s what’s coming from official sources, don’t be fooled,'” van der Linden said.
“That’s a public service that everyone can contribute to, so that we all have a healthy information diet and that we’re keeping each other safe as a community-led strategy.”
