Metacognitive Artificial Intelligence

Unlock the future of AI with metacognitive systems that self-assess, adapt, and optimize for unparalleled reliability and innovation.

Paulo Shakarian (Edited by), Hua Wei (Edited by)

9781009522458, Cambridge University Press

Hardback, published 25 September 2025

308 pages
23.4 x 16 x 2.1 cm, 0.59 kg

'This book on metacognitive AI addresses a timely and critical question in the general field of artificial intelligence: how to make AI systems more reliable and self-aware. The book strikes a good balance between theories, methods, and applications. It is an invaluable resource for researchers and practitioners.' Hanghang Tong, University of Illinois Urbana-Champaign

This groundbreaking volume is designed to meet the burgeoning needs of the research community and industry. This book delves into the critical aspects of AI's self-assessment and decision-making processes, addressing the imperative for safe and reliable AI systems in high-stakes domains such as autonomous driving, aerospace, manufacturing, and military applications. Featuring contributions from leading experts, the book provides comprehensive insights into the integration of metacognition within AI architectures, bridging symbolic reasoning with neural networks, and evaluating learning agents' competency. Key chapters explore assured machine learning, handling AI failures through metacognitive strategies, and practical applications across various sectors. Covering theoretical foundations and numerous practical examples, this volume serves as an invaluable resource for researchers, educators, and industry professionals interested in fostering transparency and enhancing reliability of AI systems.

Part I. Introduction: 1. Metacognitive AI Hua Wei, Paulo Shakarian, Christian Lebiere, Bruce Draper, Nikhil Krishnaswamy, Sarath Sreedharan and Sergei Nirenburg
Part II. Taxonomy of Metacognitive Approaches: 2. An architectural approach to metacognition Christian Lebiere, Robert Thomson, Andrea Stocco, Mark Orr and Donald Morrison
3. Metacognitive AI through error detection and correction rules Bowen Xi and Paulo Shakarian
4. Mutual trust in human–AI teams relies on metacognition Sergei Nirenburg, Marjorie McShane and Thomas M. Ferguson
Part III. Neuro-Symbolic Models in AI: 5. Learning where and when to reason in neuro-symbolic inference Christina Cornelio
6. Assessment of competency of learning agents via inference of temporal logic formulas Zhe Xu, Nasim Baharisangari, Jean-Raphaël Gaglione and Ufuk Topcu
Part IV. Metacognition with LLMs: 7. Metacognitive intervention for accountable LLMs through sparsity Tianlong Chen
8. Metacognitive insights into ChatGPT's arithmetic reasoning Noel Ngu, Paulo Shakarian, Abhinav Koyyalamudi and Lakshmivihari Mareedu
Part V. Metacognition in Learning Agents: 9. Uncertainty quantification's role in metacognition Gavin Strunk
10. The role of predictive uncertainty and diversity in embodied AI and robot learning Ransalu Senanayake
Part VI. Assured Machine Learning in High-Stakes Domains: 11. Towards certifiably trustworthy deep learning at scale Linyi Li
12. Metacognition with neural network verification and repair using Veritex Xiaodong Yang, Tomoya Yamaguchi, Bardh Hoxha, Danil Prokhorov and Taylor T. Johnson
Part VII. Metacognition as a Solution to Handle Failure: 13. Reasoning about anomalous object interaction using plan failure as a metacognitive trigger Nikhil Krishnaswamy
14. Tractable probabilistic reasoning for trustworthy AI YooJung Choi
Part VIII. Applications of Metacognitive AI: 15. Robust and compositional concept grounding for image generative AI Yezhou Yang
16. mLINK: Machine learning integration with network and knowledge Sergei Chuprov, Raman Zatsarenko and Leon Reznik
17. Military applications of artificial intelligence metacognition Bonnie Johnson.

Subject Areas: Artificial intelligence [UYQ]