Abstract:

Artificial intelligence applied to computer-assisted instruction aims, among others, to optimize and diversify automated differentiated learning capacities. This includes tailored lessons and dynamic adaptation of pedagogical interventions to meet current individual students’ interest and needs. Thus, computer-assisted tutored learning "ideally" takes place through intelligent tutoring systems (ITS): particularly via knowledge tracing tutors. However, due to the high costs of building explicit cognitive models characterized by the hyper-specification and the quasi-inextensibility of learning knowledge spaces and models closed to users, these systems are not much used by teachers and often remain at the prototype stage.

As a part of an overall objective that covers the different aspects (conceptual, architectural, pedagogical, etc.) contributing to the extensibility of curriculum within knowledge tracing tutoring, this communication focuses on designing an implicit generic domain model exploitable to operationalize tutored learning activities based on the knowledge tracing paradigm.

Despite unflattering criticisms, only generic shells enable the reuse of whole systems (just giving any learning content) among all the solutions aiming to reduce ITS’ development costs and facilitate the reuse of its components. Thus in this paper, we propose a design of an integrated didactic workshop for a generic shell that should include principal ITS’ functions.

Computer-Assisted Instruction tutors are specific types of educational software that, focused on predefined learning goals, include all the functions usually devolved to the teacher regarding learners: transmission and assessment of knowledge and skills, dynamic adapted and contextualized feedback (reinforcement/remediation). When using artificial intelligence principles and technics to design and implement these features, we talk about “intelligent” tutors.

Note that artificial intelligence describes and models knowledge through explicit or implicit representations. So, for a narrow array of knowledge components, the explicit representation enables knowledge tracing tutoring to trace step-by-step problem-solving processes and so knowledge raised by learners at each step. Unlike curriculum sequencing tutors, through their implicit and structured models opened to users can manage an extensible wide range of knowledge components: Which contributes to a better acceptance of these despite a lower quality of learners' assessment.

It seems that opening a knowledge tracing model (namely Constraint-based logic) and its coupling with principles of contents organization and scheduling inherent in contents sequencing logic could solve the problem of scalability of tutoring and inclusion of multiple domains in ITS. The use of micro-content enhances and facilitates the implementation of direct instruction tasks.

The result expected is an entirely reusable and generic ITS shell that supports the logics of knowledge tracing and content sequencing. Currently, we have designed the architectural components of this shell, defined its principal functionalities and implemented the integrated didactic workshop. We will present details about knowledge domains we could model and represent, the nature and scopes of contents (like micro-contents) we can manage, and features provided by that workshop to our shell, as well as the pedagogical approach implemented.

Keywords:

Intelligent Tutoring System, Didactic workshop, Micro-content, Knowledge Tracing, Content Sequencing.