/data/photo/2026/03/24/69c27e9e09215.png)
Why it matters
The finding matters because it moves the debate about synthetic media beyond the question of whether a fake is convincing. According to Kompas.com, the concern is what happens to memory after people repeatedly encounter fabricated material. Human recollection is not fixed; it can shift when new information arrives, including information that is wrong. Psychology researchers have known this for a long time. What is different now is that AI can produce images, video and text that look credible even when their content does not match reality.
Kompas.com notes that when people are exposed to false information again and again — seeing it, reading it, taking it in — their memory of the original event can be affected or even replaced. That makes the spread of realistic synthetic content a problem not just for sorting fact from fiction in the moment, but for what people carry away afterward.
How the MIT experiment worked
The study, as described by Kompas.com, ran in stages. Participants first looked at a set of photographs prepared by the researchers. They were then split into two groups. One group was shown the original photos again. The other group was shown versions of those photos that had been altered using AI.
The researchers then tested both groups on how accurately they recalled the images they had seen at the start. Those who had viewed the AI-manipulated versions made more errors about the content of the originals. Kompas.com reports that they were also inclined to feel strongly that their mistaken answers were correct — a combination that makes the errors harder to correct, since the people holding them do not experience doubt.
Earlier evidence on deepfakes and voting
The Kompas.com report points to a separate 2025 study that examined how synthetic content affects memory. In that research, participants developed false memories of a political scandal that never took place after being shown deepfake videos and AI-written text. The report adds that the fabricated information was also linked to participants' stated intention to vote in an election.
That earlier work suggests the effects of manipulated content do not stop at whether someone believes a specific claim. In some circumstances, Kompas.com reports, false information can shape a person's decisions. The MIT findings and the 2025 study point in the same direction: realistic synthetic media can leave a trace in memory that outlasts the moment of viewing.
What to watch
The Kompas.com article does not detail the sample size, the specific manipulations used, or a publication venue for the MIT work, so the scope of the findings is not fully clear from the report. What happens next will likely depend on how platforms label or limit synthetic content and on whether researchers can show how long these memory effects last and whether they fade with correction. For now, the practical takeaway from the reporting is straightforward: the images people scroll past are not always remembered as images. Sometimes they are remembered as events.
