The Mandela effect tricks our brains with false memories. Is AI making it worse?

National Geographic

7 days ago

Darth Vader never said, 'Luke I am your father.' So why do we all think he did? Here's the science behind this strange phenomenon—and how AI is poised to send it into overdrive. Cape Town street art depicts former South African President Nelson Mandela. A phenomenon known as the Mandela effect originates from the false memory of Nelson Mandela passing away during his imprisonment in the 1980s. Photograph by Nardus Engelbrecht, AFP/Getty Images
Darth Vader never actually said 'Luke I am your father.' Your favorite children's book series was the Berenstain Bears, not the Berenstein Bears. And the cow on Laughing Cow cheese never actually had a nose ring.
These are some of the most famous examples of a phenomenon known as the Mandela effect —an experience where the public collectively misremembers an image, event, or phrase.
It's possible that modern advances in technology, such as generative artificial intelligence, could lead to similar confusion but with potentially negative consequences. Exactly what role AI might play in the creation of our memories is something that experts in both human memory and AI misinformation are interested to find out.
Here's how the Mandela effect explains the science of misremembering. What is the Mandela effect?
The Mandela effect is a kind of false collective memory in which many people remember the same incorrect details about an event, phrase, or image.
'When we think of false memories, we usually think of them in an individual way, like, 'Oh I remember my second birthday party being a certain way' but when you look back at the photos [it's different],' says Wilma Bainbridge, assistant professor of psychology at the University of Chicago who has studied the Mandela effect. 'What's really striking about the Mandela effect is that it is a form of false memory that occurs across people.' Limited Time: Bonus Issue Offer Subscribe now and gift up to 4 bonus issues—starting at $34/year.
The Mandela effect was coined in 2009 by paranormal researcher Fiona Broome when she noticed that many people shared a false memory about the death of South African President Nelson Mandela. Mandela died in 2013 from a lung infection, but many people falsely remember that he died while in prison in the 1980s for fighting against apartheid.
(This ancient technique may help you remember almost anything.)
Since then, examples of the Mandela effect have popped up in different forms with help from social media to bring together people to uncover these shared experiences. Bainbridge says millennial childhoods are often the breeding ground for this effect because they are some of the heaviest social media users on sites like Instagram or Reddit where these effects have been seen. That said, this effect can still be seen in some older iconography as well, she says.
These altered memories are typically harmless. Yet at its core, the Mandela effect makes us doubt our own memories and even our sense of reality. Research behind the Mandela effect
Scientific studies of the Mandela effect are still relatively new, but scientists have already been studying the creation of false memories for decades. This is part of the work that Aileen Oeberst does as a professor of social psychology at the University of Potsdam in Germany.
Part of what makes memories so fallible is that the brain uses the same area—the hippocampus—for both imagination and memory storage, says Oeberst.
'That already suggests some important consequences for false memories,' she said. 'We know from research that if people imagine something repeatedly, they tend to believe at some point that they actually experienced it and that it is basically a memory.'
(How using your senses could help you make stronger memories.)
When you recall a memory, your brain doesn't play it back like a video but rather reconstructs it, which makes it susceptible to misremembering. For example, an individual might fill in the gaps in their memory with details they might expect to be true based on stereotypes. Or they might filter a memory through an emotionally charged lens—ultimately remembering what happened as having been better or worse than it really was.
Memories that are novel, emotional, and self-relevant are more susceptible to these changes because we tend to remember and talk about them often, Oeberst says.
(Yes, you can teach yourself to forget. And here's why you should.)
Interestingly, however, these explanations don't totally fit what happens in the Mandela effect, says Bainbridge. In her 2022 work, Bainbridge and co-author Deepasri Prasad found that a Mandela effect memory can be formed even in opposition to a stereotype.
To better understand how a Mandela effect is created, Bainbridge and Prasad studied people's reactions to different iconography, such as Curious George or Pikachu's tails, the Monopoly Man's outfit, or the Volkswagen logo.
Take the Fruit of the Loom logo for example, Bainbridge says. The brand's logo is a collection of grapes with an apple at the center, and it floats on white background.
'The common false memory is that there's a giant cornucopia around the fruit [in the logo],' Bainbridge says. 'But we see fruit so often in our daily lives and when do we ever see a cornucopia?'
Bainbridge and Prasad even seeded another false image of the Fruit of the Loom logo—this time placing the fruit on a plate instead of a cornucopia—but participants still chose the cornucopia more often than both the plate version and the actual logo.
This research did not conclusively determine exactly what makes different icons or events susceptible to the Mandela effect, but Bainbridge says they could determine that simple images with just a couple of interesting quirks seem to be the most memorable. The researchers also found that what people misremember about these images is remarkably consistent and becomes stronger with repetition. In the future, Bainbridge is interested in extending this work to study exactly what makes an image memorable and even reverse engineering that to create new Mandela effects.
This idea of solidifying false memories through repeated exposure is part of what could make false AI images such a risk, says Prasad, who is now a graduate student of psychological and brain sciences at Dartmouth.
'I think misinformation, especially when presented as the truth repeatedly, could definitely lead to the creation of false memories, or at least, doubting the validity of your own memories,' she says. False Memory and AI
If the Berenstain Bears are a standout example of the Mandela effect, then the viral image of the late Pope Francis wearing a giant Balenciaga puffer jacket is a standout example of what this phenomenon could look like in the world of AI.
'The pope in a fluffy coat was one of the first [generative AI images] that went viral,' says Jen Golbeck, a professor at the University of Maryland's College of Information who studies AI, social media, and trust online. 'And there's probably people who saw that image and didn't realize that it was [AI] generated. We're in such an interesting time now where people know they have to be suspicious, but some people just don't care.'
There are a number of headwinds that have made the perfect storm for this kind of misinformation to spread, Golbeck said, including the rise of 'fake news' sites, the erosion of institutional trust throughout the COVID-19 pandemic, and the rise of increasingly convincing content from gen-AI. The risk of being tricked by AI comes not only from human-prompted creations, but from believing AI hallucinations as well.
Even for those who pride themselves on being good at sniffing out AI images, it's becoming harder and harder to do so. One of the biggest risks, says Oeberst, is that our brains are predisposed to forget source information more quickly than content, meaning that we may remember what a false AI image looked like but forget that we're meant to distrust it.
Because generative AI is new and evolving, researchers don't know about its future risks. However, they are eager to explore this technology's influence on both individual false memories and the Mandela effect.
Both Oeberst and Prasad are interested in whether false AI images will be believed more readily if they reinforce someone's beliefs or opinions, and Golbeck is interested in whether AI can play a role in reinforcing false memories.
The researchers agree that long-term studies will be important to understand the extent of these interactions between memory, images, and AI.
As for what we can all do right now to keep our memories safe from corruption, Golbeck says it's important to lean on community.
'One important step is to really establish a cohort of people that you do trust,' Golbeck said. 'Like journalists, scientists, politicians, who you've really evaluated and are going to tell you correct information, even if it's not what you want to hear. I think that's critical.'