Automatic Acquisition of Knowledge for Constraint-based Tutors

Intelligent Tutoring Systems (ITS) assist students in learning by adaptively providing pedagogical assistance. Numerous empirical studies have shown that students learn more effectively by interacting with ITSs in comparison to traditional classroom based teaching [1, 2]. Although ITSs are highly regarded as effective tools for education, developing an ITS is a time and labour intensive task requiring programming skill as well as knowledge engineering skills. A major proportion of the time and effort for building an ITS is spent on acquiring the domain knowledge required for providing adaptive assistance. Anderson and co-workers estimated that ten hours or more were required to produce a single production rule [3]. Our main goal is to automate the knowledge acquisition process to drastically reduce the time and effort require for building an ITS. Researchers have been exploring ways of conquering the knowledge acquisition bottleneck ever since the inception of ITSs. Previous research including KnoMic (Knowledge Mimic) [4], Disciple [5] and Demonstr8 [6] have focussed on acquiring procedural knowledge by recording the actions of a domain expert and generalising the recorded trace using machine learning algorithms. Although these systems are well suited for inherently procedural domains like simulated environments, they fail to acquire declarative knowledge required for non-procedural domains. Our goal is to develop an authoring system that is capable of acquiring knowledge for procedural as well as non-procedural domains. Constraint based modelling (CBM) [7] is a student modelling technique that somewhat eases the knowledge acquisition bottleneck by using a more abstract representation of the domain compared to other popular domain modelling techniques [8]. However, building a constraint base still remains a major challenge. Mitrovic reported that, she took just over an hour to produce a constraint for SQL-Tutor, which currently contains more than 650 constraints [9]. Our research is focussed on automating the process of acquiring knowledge for constraint-based tutors. The authoring system will be an extension to the web-based tutoring shell, named WETAS [10], that facilitates building constraint-based tutors. WETAS provides all the domain-independent components for a text-based ITS, including the user interface, pedagogical module and student modeller. The main limitation of WETAS is its lack of support for authoring the domain model. We propose a four-stage process to infer constraints automatically. During the first phase the domain expert composes an ontology of the instructional