New Study Reveals AI-Generated Summaries Can Distort Human Memories and Omit Critical Facts

A recent study has revealed that the use of AI-generated summaries in news articles can manipulate human memories. Released by Georgetown University and the University of Washington, the peer-reviewed research indicates that AI overviews of articles and videos frequently omit critical details and incorrectly summarize events.

According to the report titled AI-Enabled Human Memory Manipulation, the summaries omitted an average of 51.6 per cent of central details across all prompts and models. The study tested several AI models and found them to be particularly inaccurate when asked to describe an animated video depicting an accident between a car and a pedestrian, which in turn affected what people recalled.

The analysis noted that 83.6 per cent of participants who read a summary containing accurate information about the video answered correctly, while only 44.8 per cent answered correctly after reading a misleading summary. The research examined the accuracy of AI summaries using tools such as OpenAI’s ChatGPT and Google’s Gemini.

Yael Eiger, a PhD student at the University of Washington who helped lead the research, expressed surprise at the poor quality of the summaries. She warned that the inaccuracies and omissions could have serious ramifications, expressing particular concern that police departments might utilize video summarization technologies without rigorous testing or awareness of potential errors.

Mattea Sim, an assistant research professor at Georgetown, stated that the results should provide greater guidance regarding when AI-generated summaries ought to be avoided. She emphasized that AI serves as a new method of delivering misinformation with the potential to create false memories for readers, urging people to think critically about whether and how AI should be used for summarization in high-stakes settings.

The findings are scheduled to be presented in Sweden during an October conference on AI, ethics, and society, organized by the Association for the Advancement of AI. These insights arrive amid broader concerns regarding AI, including job losses, misinformation, disinformation, and the environmental impact of rapid data centre construction.

Yoshi Kohno, chairman of computer science, ethics, and society at Georgetown, expressed hope that the research would prompt positive change, noting that the study forms part of a broader effort at Georgetown to explore the relationship between humans and AI using both psychology and computer science.