The CARINA Metacognitive Architecture

The CARINA Metacognitive Architecture

Manuel Fernando Caro, Darsana P. Josyula, Dalia Patricia Madera, Catriona M. Kennedy, Adán A. Gómez
DOI: 10.4018/IJCINI.2019100104
(Individual Articles)
No Current Special Offers


Metacognition has been used in artificial intelligence to increase the level of autonomy of intelligent systems. However, the design of systems with metacognitive capabilities is a difficult task due to the number and complexity of processes involved. The main objective of this article is to introduce a novel metacognitive architecture for monitoring and control of reasoning failures in artificial intelligent agents. CARINA metacognitive architecture is based on precise definitions of structural and functional elements of metacognition as defined in the MISM meta-model. CARINA can be used to implement real-world cognitive agents with the capability for introspective monitoring and meta-level control. Introspective monitoring detects reasoning failure (for example, when expectation is violated). Metacognitive control selects strategies to recover from failures. The article demonstrates a CARINA implementation of reasoning failure detection and recovery in an intelligent tutoring system called FUNPRO.
Article Preview

1. Introduction

Cognitive Informatics is a multidisciplinary research area that investigates the internal information processing mechanisms of the brain and natural intelligence shared by almost all science and engineering disciplines (Wang, 2007). Cognitive Computing is an emerging paradigm of Artificial Intelligence (AI) based on Cognitive Informatics, that implements computational intelligence by autonomous inferences and perceptions mimicking the mechanisms of the brain and natural intelligence (Wang et al., 2010). Computational metacognition in in Cognitive Computing refers to the capability of Intelligent Systems (IS) to monitor and control their own learning and reasoning processes (Caro, Josyula, & Jiménez, 2014). Computational metacognition distinguishes reasoning about reasoning from reasoning about the world (Cox, 2005).

A metacognitive architecture provides a concrete framework for detailed modeling of mechanisms for an AI agent's high-level reasoning about itself, through specifying essential structures, divisions of modules, relations among modules, and a variety of other essential aspects (Sun, Langley, Laird, & Rogers, 2009; Sun, Zhang, & Mathews, 2006). Metacognitive architectures differ from cognitive architectures in that the agent itself is the referent of the cognitive processing but share a commitment to formalisms for representing knowledge, memories for storing this domain content, and processes that utilize and acquire the knowledge (Langley, Laird, & Rogers, 2009; Sun et al., 2009). Nowadays increasingly complex AI agents that make decisions based on multiple variables are developed. The complexity increases the probability that reasoning failures occur in such AI agent (Conitzer & Sandholm, 2003; Schmid, Ragni, Gonzalez, & Funke, 2011). A reasoning failure is defined as an outcome other than what is expected or a lack of some outcome or appropriate expectation (Cox, 1997). Reasoning failures are generated mainly by unfinished tasks or unexpected results in the performance of a task. Metacognition provides introspective monitoring and meta-level control mechanisms for an AI agent to detect and correct its own reasoning failures.

Some implementations exist and have made progress in the integration of metacognitive functions in cognitive architectures. The ACT-R (Anderson, 1996; Anderson, Matessa, & Lebiere, 1997; Borst & Anderson, 2015), a cognitive architecture based on the theory of rational analysis, can provide a description of the processes from perception through to action for a wide range of cognitive tasks (Borst & Anderson, 2015). Although ACT-R theory was not developed with specific metacognitive mechanisms, several metacognition researches have been developed based on this architecture. For example, the work of (Bogunovich & Salvucci, 2011) demonstrate through the use of ACT-R that subjects can dynamically control their problem-solving process based on the knowledge of the relative demands of tasks and the mental resources needed to complete it. Reitter (Reitter, 2010) designed several metacognitive layers in ACT-R that implement several strategies to address the basic control task, as well as the means to classify and select those strategies according to their suitability in a given situation. However, the modeling process has concentrated on multitasking behavior and has not clarified the specific relationship between multitasking and metacognition.

Complete Article List

Search this Journal:
Volume 18: 1 Issue (2024): Forthcoming, Available for Pre-Order
Volume 17: 1 Issue (2023)
Volume 16: 1 Issue (2022)
Volume 15: 4 Issues (2021)
Volume 14: 4 Issues (2020)
Volume 13: 4 Issues (2019)
Volume 12: 4 Issues (2018)
Volume 11: 4 Issues (2017)
Volume 10: 4 Issues (2016)
Volume 9: 4 Issues (2015)
Volume 8: 4 Issues (2014)
Volume 7: 4 Issues (2013)
Volume 6: 4 Issues (2012)
Volume 5: 4 Issues (2011)
Volume 4: 4 Issues (2010)
Volume 3: 4 Issues (2009)
Volume 2: 4 Issues (2008)
Volume 1: 4 Issues (2007)
View Complete Journal Contents Listing