Development of cognitive models for the metacognitive architecture CARINA
Jerónimo Montiel, Alba Judith | 2020
Cognitive modeling is a methodology of cognitive sciences that allows the simulation of human cognitive processes in a variety forms, commonly in a computational and mathematical way. The cognitive modeling aims at understanding cognition basis by designing cognitive models based on mathematical or computational processes, mechanisms and representations. A cognitive model is a verbal-conceptual computational and mathematical description of some mental processes, whose main purpose is to understand and/or predict human or animal behavior. Cognitive models developed for a cognitive architecture are characterized by being executables and producing a set of specific behaviors. CARINA is a metacognitive architecture to create artificial intelligent agents derived from Metacognitive Metamodel MISM. CARINA is a metacognitive architecture structured by two cognitive levels called object-level and meta-level. The object-level has the model of the world to solve problems. The meta-level represents the reasoning of an artificial intelligent agent.
Furthermore, the meta-level has the components, the knowledge and the mechanisms for an intelligent system to monitor and control its own learning and reasoning processes. The main objective of this research project is to develop cognitive models as knowledge acquisition mechanisms for the metacognitive architecture CARINA, through the following specific objectives: i) to represent formal, semantic and computationally cognitive models for the CARINA metacognitive architecture, ii) to build a functional prototype of a framework for the creation of cognitive models in the metacognitive architecture CARINA and iii) to create cognitive models in several knowledge domains using CARINA based intelligent systems. The methodology used for this research project was part of the research methods (R+D) used in computer science, called modeling, structured by five steps: i) Formal representation, ii) Semantic representation, iii) Computational representation of a cognitive model, iv) Creation of a functional prototype for build cognitive models and v) Prototype testing and maintenance. The developed research project allows simplifying the developing intelligent agents process and the easiness to enable any programmer to uses CARINA to solve cognitive tasks, focusing only on descriptions of cognition and relationships with algorithms and programs based on computer science and technology, using a functional prototype (MetaThink version 2.0). As a result, an open standard file format, simplifying the complexities of detailed descriptions of cognitive mechanisms of brain functioning was created.