Position: Artificial Intelligence Needs Meta Intelligence -- the Case for Metacognitive AI
For AI researchers, this paper argues for a new design principle but provides only a single case study, making it more of a proposal than a validated approach.
This position paper proposes metacognition as a design principle for AI to improve accuracy, security, and efficiency. It demonstrates improved learning efficiency and security in a Federated Learning case study.
This position paper argues for metacognition as a general design principle for creating more accurate, secure, and efficient AI. The metacognitive solution involves systems monitoring their own states and judiciously allocating resources depending on each problem instance's difficulty or cost of mistakes. Drawing inspiration both from past work on resource-rational AI and from well-documented metacognitive strategies in psychology and cognitive science, we identify specific challenges in embedding these strategies into AI design and highlight open theoretical and implementation problems. We showcase these principles through a tangible example of improved learning efficiency, effectiveness, and security in a Federated Learning (FL) case study. We show how these principles can be translated into practice with a novel software framework developed specifically to allow the community to design, deploy, and experiment with metacognition-enabled AI applications.