An Intelligent Tutoring System (ITS) view on AOSE

Abstract

The creation of Intelligent Tutoring System (ITS) applications involves tasks that require the modelling of complex agents (i.e., human students); therefore, a cognitive agency approach is more appropriate than a weak agency approach. ITS applications need to know, plan and understand several aspects of the other intelligent entities that inhabit the same environment. This is the viewpoint not considered in Agent-Oriented Software Engineering (AOSE) methodologies that this work intends to cover. This aspect is the main innovation of this work, and also presents a critical requirement that should be fulfilled by applications before someone tries to use the AOSE methods proposed here. In this context, the paper presents a set of AOSE methods derived from ITS research, which defines applicability criteria, designs principles and implementation guidelines to be applied in the software analysis, design and development process. The application of such methods is exemplified by several ITSs developed in our research group.

MASRML - A Domain-specific Modeling Language for Multi-agent Systems Requirements

“…An agent is a process situated in an environment, designed to achieve a purpose through an autonomous and flexible behavior. The environment is the application domain where the agent will work [11]. We adopt the BDI (Belief, Desire, Intention) cognitive model of [13], who states that the purpose of an agent can be fully specified by the definition of its beliefs and desires, and that its behavior is implied by its intentions.…”

Applying AUML and UML 2 in the Multi-agent Systems Project

“…All the same, the AUML language currently supports only weak notions of agency, representing agents as objects, while employing state-machine diagrams to model their behavior and extended Interaction diagrams to model their communicative acts [1], not support cognitive or social abstractions. Caire in [4] comments that, although this notation is useful, it bears no agent concept in its core, stating also that specifying the behavior of an object in terms of interaction protocols will not change such an object into an agent.…”

“…The representation of agents as objects conflicts with the definition of agents for, according to [1], 'an agent is a computational process located within an environment and designed to achieve a purpose within said environment by means of autonomous, flexible behavior'.…”

“…The justification for the employment of actors instead of objects, besides that of the concept of agent itself described in [1], is based upon the concept of an actor as described in [11], where is stated that an actor represents a role played by some entity and that an entity may play the role of several different actors and an actor may be played by multiple different entities. Although the concept of role used by UML might be different than the concept of role used in multiagent systems, they seem to be, at first sight, similar to each other.…”

A UML Profile Oriented to the Requirements Modeling in Intelligent Tutoring Systems Projects

“…Despite many AOSE methods having been created for multi-agent systems (MAS) projects, Vicari [1] states these are not totally adequate for the modeling of Intelligent Tutoring Systems. Likewise, some attempts to adapt and extend UML (Unified Modeling Language) to the multi-agent systems projects were made; nonetheless, those attempts we have studied did not concern themselves into extending and applying some of the UML resources, such as the use-case diagram, which is mainly employed for requirements collecting and analyzing, an essential phase for the achievement of a good system project.…”