arXiv:2607.07753v1 Announce Type: new Abstract: Modelling psychological disorders in artificial agents offers both a testbed for computational psychiatry and a lens on the failure modes of affective control. Prior work induces one or two disorders in a reinforcement learning (RL) agent by hand-tuned reward shaping, labels the behaviour post hoc, and reports single runs. We recast disorder modelling as dose-controllable manipulation of cognitive appraisal signals in an appraisal-guided PPO agent, expressing seven disorders (anxiety, mania, obsessive-compulsive checking, depression, impulsivity,
The paper leverages recent advancements in Reinforcement Learning (RL) and computational psychiatry to create more nuanced models of psychological disorders, moving beyond previous limited approaches.
This research provides a novel testbed for understanding the computational underpinnings of mental health conditions and the failure modes of AI, offering insights for both human and artificial intelligence.
Disorder modeling in AI shifts from simple, hand-tuned inductions to dose-controllable manipulation of cognitive appraisal signals, enabling the study of multiple 'disorders' within a single agent.
- · AI ethicists
- · Computational psychiatrists
- · Developers of robust AI systems
- · Mental health researchers
- · Developers of simplistic AI models
AI agents can be designed with 'psychological profiles' that mimic human conditions more closely.
Understanding AI 'psychopathology' could lead to new diagnostic or treatment frameworks for human mental health.
The development of 'mentally healthy' AI could become a design criterion for general artificial intelligence, influencing future regulatory frameworks.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG