Misleading AI-Generated Summaries Can Distort Human Memory (college.georgetown.edu)

New research from Georgetown University and the University of Washington highlights the potentially harmful effects of misleading AI-generated summaries on human memory. The study, "AI-Enabled Human Memory Manipulation: Misleading AI-Generated Summaries Distort Human Memory," demonstrates that incorrect AI summaries can significantly distort how individuals recall events, revealing a concerning trend as AI technology becomes more integrated into critical areas like law enforcement and corporate communications. By analyzing AI outputs from models like ChatGPT and Google’s Gemini, the researchers found that a striking 51.6% of central details were omitted in the summaries of a video depicting a car-pedestrian accident. Participants exposed to misleading summaries showed a dramatic drop in accurate recall, with only 44.8% able to correctly answer questions about the event, compared to 83.6% who read accurate summaries. This phenomenon occurred regardless of participants' beliefs about the AI or human origin of the summaries. The study underscores the need for careful evaluation of AI's role in information dissemination, particularly in high-stakes environments where distorted memories could lead to serious consequences, prompting calls for more rigorous audits and transparency in AI technologies.