Slip Through the Chat: Subtle Injection of False Information in LLM Chatbot Conversations Increases False Memory Formation

Abstract

This study examines the potential for malicious generative chatbots to induce false memories by injecting subtle misinformation during user interactions. An experiment involving 180 participants explored five intervention conditions following the presentation of an article: (1) no intervention, (2) reading an honest or (3) misleading article summary, (4) discussing the article with an honest or (5) misleading chatbot. Results revealed that while the misleading summary condition increased false memory occurrence, misleading chatbot interactions led to significantly higher rates of false recollection. These findings highlight the emerging risks associated with conversational AI as it becomes more prevalent. The paper concludes by discussing implications and proposing future research directions to address this concerning phenomenon.