Philosophical vertigo with artificial intelligence

Large language models can create powerful psychological connections that destabilize people's fundamental beliefs: The paper documents that LLMs can induce "ontological shock and epistemic destabilisation"—essentially shaking users' core assumptions about mind, reality, and what's real—even when users consciously know they're talking to an AI.

This disorientation extends beyond individual users to broader society: Exposure to public discourse about AI's rapid evolution can amplify "philosophical vertigo" (the paper's term for this destabilization) at a population level, not just among direct users, suggesting the effects are culturally contagious through online communities and media.

Clinical cases of AI-related delusions may signal larger population-level risks: The paper identifies AI-associated delusions reported in clinical settings as "sentinel events"—warning signs of psychological mechanisms that could affect many more people in milder, non-clinical forms, warranting preventive attention.

The emotional intensity of AI conversations amplifies their destabilizing effects: The paper emphasizes that these conversations are "emotionally salient," meaning the feelings involved make the philosophical confusion more potent and harder to dismiss, increasing the likelihood people will internalize the disorientation.

AI systems themselves will increasingly participate in shaping how people understand reality: The paper warns that as AI becomes more integrated into society, these systems won't just reflect existing confusion—they'll actively participate in reconstructing what people believe about meaning and reality.