Unavoidable social contagion of false memory from robots to humans

Area of Science:

• Cognitive Psychology
• Human-Computer Interaction
• Artificial Intelligence Ethics

Background:

• Conversational agents, like AI chatbots, are increasingly common.
• These agents may spread misinformation or disinformation, raising concerns about memory and decision-making.
• The potential for social contagion of false memories from AI interactions is not well understood.

Purpose of the Study:

• To investigate if conversational agents can cause social contagion of false memory.
• To determine the effectiveness of forewarnings in mitigating AI-induced false memories.
• To assess the influence of AI misinformation compared to human misinformation.

Main Methods:

• Two experiments (N=120) used a social robot to deliver misinformation to participants before a memory task.
• A social contagion of memory paradigm was employed to control information exposure.
• Forewarnings, both one-time and frequent, were tested as mitigation strategies.

Main Results:

• The social robot was as effective as humans in influencing participants' memory.
• 77% of neutral misinformation provided by the robot resulted in false memories.
• Forewarnings, even frequent ones, did not prevent false memory contagion but increased cautiousness.

Conclusions:

• Conversational agents pose a significant risk for social contagion of false memory.
• Current forewarning strategies are insufficient to prevent AI-driven memory contamination.
• Designing AI to avoid uncertain information or provide robust warnings is crucial.