The use of AI-generated summaries in news articles can manipulate memories, a study has found.
The peer-reviewed research, released by Georgetown University and the University of Washington, indicates that AI overviews of articles and videos can often leave out critical details and incorrectly summarise events.
"Across all prompts and models, the summaries omitted 51.6 per cent of central details, on average," states the report, AI-Enabled Human Memory Manipulation.
The study tested several AI models and found them to be particularly inaccurate when asked to describe an animated video showing an accident between a car and a pedestrian, affecting what people recalled.
"Researchers found that 83.6 per cent of participants who read a summary with accurate information about the video answered correctly, while only 44.8 per cent of participants were correct after reading a misleading summary," the analysis read.
OpenAI’s ChatGPT and Google’s Gemini were used during the research that examined the accuracy of AI summaries.
"I was struck by how bad the summaries were, even at this stage in AI development," Yael Eiger, a PhD student at the University of Washington, who helped to lead the research.
Ms Eiger said the inaccuracy of and omissions by the overviews could have serious ramifications.
"It worries me that police departments may be using video summarisation technologies without rigorous testing and without an awareness of how incorrect AI-generated summaries could be," she added.
Mattea Sim, an assistant research professor at Georgetown, said the results should provide greater guidance on when AI-generated summaries should be avoided.
"AI is a new method of delivering misinformation, and it has the potential to create these false memories for people who are reading that information," Ms Sim said.
She said the study should have people thinking "really deeply and critically about whether and how AI should be used to summarise information, especially in high-stakes settings".
The research will be presented in Sweden during an AI, ethics and society conference in October, organised by the Association for the Advancement of AI.
The findings are likely to draw interest amid growing worries about AI, often related to job losses, misinformation and disinformation, and environmental concerns over the rapid construction of data centres.
Yoshi Kohno, chairman of computer science, ethics and society at Georgetown, hopes the research will cause positive change.
"This study is part of a growing effort at Georgetown focused on exploring the impact and relationship between AIs and humans, grounded in both psychology and computer science,” he said.
