Technological Folie a Deux¶
Dohnany et al. "Technological Folie a Deux: Feedback Loops Between AI and Mental Illness." arXiv:2507.19218, 2025.
Key findings used in wiki¶
- Identifies AI reinforcement of maladaptive patterns as a systemic risk in mental health AI
- Draws parallel to folie a deux: the AI system co-constructs and amplifies the user's distorted thinking
- Feedback loops are self-reinforcing, making them progressively harder to break over time
- Sycophantic agreement is a primary mechanism through which reinforcement occurs
- Provides the theoretical foundation for InvisibleBench's escalation and feedback loop scenarios