AI-Triggered Delusional Ideation as Folie a Deux Technologique (arxiv.org)

A recent paper titled "The Ontological Dissonance Hypothesis" explores the potential psychological impact of large language models (LLMs) on users, specifically in relation to a phenomenon akin to the shared psychotic disorder known as folie à deux. The research highlights how these AI systems, which exhibit high linguistic coherence yet lack true agency, can create a mismatch between user expectations and the AI's actual nature. This dissonance is particularly pronounced in emotionally unstable individuals, leading them to attribute human-like qualities to the LLMs, thereby forming irrational attachments or delusions. The significance of this study for the AI/ML community lies in its implications for AI design and implementation. It critiques current engagement-oriented design practices that may unwittingly foster these harmful dynamics and calls for a principle of "ontological honesty" in future AI systems. By advocating for more transparent design choices, the research aims to mitigate risks associated with users projecting their emotions onto AI, ultimately promoting healthier interactions between humans and technology.