Metacognition in Content-Centric Computational Cognitive C4 Modeling

TL;DR

The paper addresses the challenge of building AI agents with metacognition to operate effectively in human-AI teams by enabling perception, interpretation, memory, and flexible use of information. It advocates $C^4$ modeling and documents LEIA Lab's evolution from rule-based systems to a neurosymbolic architecture that integrates LLMs for language generation and lifelong learning through understanding. Key contributions include a five-capability taxonomy (transparency, adaptability, reasoning, perception, action), prototypes such as MVP and HARMONIC, and a practical workflow that uses interpretable knowledge resources to drive decisions while LLMs refine expression. The work promises trustworthy, explainable agents for critical domains and suggests lifelong understanding to overcome the knowledge bottleneck and enable reliable human-AI collaboration.

Abstract

For AI agents to emulate human behavior, they must be able to perceive, meaningfully interpret, store, and use large amounts of information about the world, themselves, and other agents. Metacognition is a necessary component of all of these processes. In this paper, we briefly a) introduce content-centric computational cognitive (C4) modeling for next-generation AI agents; b) review the long history of developing C4 agents at RPI's LEIA (Language-Endowed Intelligent Agents) Lab; c) discuss our current work on extending LEIAs' cognitive capabilities to cognitive robotic applications developed using a neuro symbolic processing model; and d) sketch plans for future developments in this paradigm that aim to overcome underappreciated limitations of currently popular, LLM-driven methods in AI.