Researchers have developed a novel framework for modeling psychological disorders in reinforcement learning agents, moving beyond single-run, hand-tuned approaches. This new method allows for dose-controllable manipulation of cognitive appraisal signals to induce seven distinct disorders, including anxiety, mania, and depression, each measured by a specific assay. Across over a thousand experimental runs, the induced disorders demonstrated a graded, dose-dependent response that controls did not replicate, suggesting a robust and controllable method for simulating affective phenotypes in AI. AI
IMPACT This research provides a new computational testbed for understanding psychological disorders and the failure modes of affective control in AI systems.
RANK_REASON Academic paper detailing a new computational framework for modeling psychological disorders in AI agents. [lever_c_demoted from research: ic=1 ai=1.0]
- Addiction
- anxiety
- impulsivity
- major depressive disorder
- Mania
- Miniworld Rotterdam
- obsessive-compulsive checking
- post-traumatic stress disorder
- Proximal Policy Optimization
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
