CoPlan: A Contestable AI Interface for Care Planning
A novel AI system named CoPlan has been introduced in a paper on arXiv (arXiv:2608.05107) to enhance the reliability of AI-assisted care planning. This system enables human participants to review, contest, and modify AI-generated suggestions. It employs a multi-agent framework where specialized AI agents propose potential interventions along with supporting or opposing arguments. Care planners, including clinicians, patients, caregivers, and care teams, have the ability to accept, reject, alter, or supplement these arguments prior to the finalization of the care plan. CoPlan aims to address the shortcomings of many AI systems that offer fixed recommendations, which may clash with clinical judgment, patient preferences, or practical implementation. It is designed to facilitate the coordination of intricate decisions encompassing clinical, functional, psychosocial, and environmental aspects of care planning.
Key facts
- CoPlan is a co-intelligent and contestable interface for human-AI care planning.
- It uses a multi-agent workflow with specialized AI agents generating candidate interventions and arguments.
- Human care planners can accept, reject, modify, or add arguments before final plan generation.
- The system aims to address fixed AI outputs that limit stakeholder inspection and revision.
- It supports coordination across clinical, functional, psychosocial, and environmental needs.
- The paper is available on arXiv with identifier arXiv:2608.05107.
- The announcement type is 'new'.
- The system combines co-intelligence and contestability.
Entities
Institutions
- arXiv