Abstract: the modeling process of metacognitive functions in Intelligent Tutoring Systems (ITS) is a difficult and time-consuming task. In particular when the integration of several metacognitive components, such as self-regulation and metamemory is needed. Metacognition has been used in Artificial Intelligence (AI) to improve the performance of complex systems such as ITS. However the design ITS with metacognitive capabilities is a complex task due to the number and complexity of processes involved. The modeling process of ITS is in itself a difficult task and often requires experienced designers and programmers, even when using authoring tools. In particular the design of the pedagogical strategies for an ITS is complex and requires the interaction of a number of variables that define it as a dynamic process. This doctoral thesis presents a metamodel for the personalized adaptation of pedagogical strategies integrating metamemory and self-regulation in ITS. The metamodel called MPPSM (Metamodel of Personalized adaptation of Pedagogical Strategies using Metacognition in intelligent tutoring systems) was synthetized from the analysis of 40 metacognitive models and 45 ITS models that exist in the literature. MPPSMhas a conceptual architecture with four levels of modeling according to the standard Meta- Object Facility (MOF) of Model-Driven Architecture (MDA) methodology. MPPSM enables designers to have modeling tools in early stage of software development process to produce more robust ITS that are able to self-regulate their own reasoning and learning processes. In this sense, a concrete syntax composed of a graphic notation called M++ was defined in order to make the MPPSM metamodel more usable. M++ is a Domain-Specific Visual Language (DSVL) for modeling metacognition in ITS. M++ has approximately 20 tools for modeling metacognitive systems with introspective monitoring and meta-level control. MPPSM allows the generation of metacognitive models using M++ in a visual editor named MetaThink. In MPPSM-based models metacognitive components required for monitoring and executive control of the reasoning processes take place in each module of an ITS can be specified. MPPSM-based models represent the cycle of reasoning of an ITS about: (i) failures generated in its own reasoning tasks (e.g. self-regulation); and (ii) anomalies in events that occur in its Long-Term Memory (LTM) (e.g. metamemory). A prototype of ITS called FUNPRO was developed for the validation of the performance of metacognitive mechanism of MPPSM in the process of the personalization of pedagogical strategies regarding to the preferences and profiles of real students. FUNPRO uses self-regulation to monitor and control the processes of reasoning at object-level and metamemory for the adaptation to changes in the constraints of information retrieval tasks from LTM. The major contributions of this work are: (i) the MOF-based metamodel for the personalization of pedagogical strategies using computational metacognition in ITS; (ii) the M++ DSVL for modeling metacognition in ITS; and (iii) the ITS prototype called FUNPRO (FUNdamentos de PROgramación) that aims to provide personalized instruction in the subject of Introduction to Programming. The results given in the experimental tests demonstrate: (i) metacognitive models generated are consistent with the MPPSM metamodel; (ii) positive perceptions of users with respect to the proposed DSVL and it provide preliminary information concerning the quality of the concrete syntax of M++; (iii) in FUNPRO, multi-level pedagogical model enhanced with metacognition allows the dynamic adaptation of the pedagogical strategy according to the profile of each student.
展开▼
