Intelligent and Adaptive Student Support in FLIP - Early Computer Programming

The general purpose of this work is to develop techniques by which a significant proportion of teaching of elementary computer programming can be shifted from human tutors to intelligent agents. An extensive literature review has been made and a concrete research objective has been formulated. The techniques that have and will be developed as part of this work are being delivered and tested through a computerised tutoring system. A framework has been designed for this purpose and a first prototype of an Exploratory Learning Environment (ELE) has been deployed. This ELE is called FLIP (FLexible, Intelligent, Personalised) and it has already been used and tested in the classroom. Currently FLIP integrates a combination of off-the-shelf and own components to provide intelligent support to students of Javascript programming. This is task-independent support provided in the context of open-ended exploratory programming sessions. FLIP’s natively supported AI component is a rule-based reasoner (Karkalas and Gutierrez-Santos, 2014a) that is based on a concept inventory (CI) of student initial misconceptions (Karkalas and Gutierrez-Santos, 2014b). The development of this CI was based on previous research projects (Goldman et al., 2008; Kaczmarczyk et al., 2010) and original exploratory field reseach. The data elicitation process took place in University computer laboratories and involved collection of primmary data through direct observations and face-toface interviews. The system is also adaptable and can adjust the level of support depending on previous student activity. This is based on a learner model that takes into account the amount of support that has already been provided by the system. An issue that is inherently problematic in all rulebased systems is the method of conflict resolution between competing rules that get activated concurrently. This is an important aspect of the system’s automated support as it is directly related to the human-computer interaction component and may significantly influence the applicability and effectiveness of the technique. This is therefore an emerging problem that needs to be addressed and resolved in the remaining part of this research. In parallel the techniques employed in the system should be assessed both with formative and summative evaluations. If these techniques prove to be effective and time permits, then the same system can be tested for task-dependent support.