The use of an expert system to identify pupils' misconception in science: a prototype and evaluation

In this research, the author proposes a development which contributes towards a knowledge of linking research in diagnosing student misconception in science education and the expert systems technology. Specifically, the thesis will describe the development and evaluation of a prototype diagnostic system to become a supportive tool for classroom teachers. Three topics of electricity, speed and motion graphs, and floating and sinking were selected to explore the use of expert systems technology in diagnostic testing in science. For each topic, the strategy for building the rule-based diagnostic knowledge representation is discussed. The main steps are analysis of past research literature in pupil misconceptions, building a matrix table consisting of various parameters and logical relationship between these parameters, designing the questions for eliciting the understanding and building the rule base. Finally the rule base has to be organised for encoding into a format suitable for inclusion into a generic expert system shell (Leonardo). In general, the two forms of rules contained in the knowledge base are diagnostic rules and the question sequence rules. The diagnostic rule consists of if-then statements which describes the patterns of typical science misconceptions found in the literature. Detection of a specific pattern results in descriptive diagnostic feedback. The question sequence also consists of if-then rules which are used to support the branching of questions according to previous responses. In the topic of floating and sinking, the diagnostic rule makes use of the certainty factors feature of the shell in making a decision. Both school pupils and teachers were used to validate the program. The analysis of pupils' responses suggests that the program is capable of diagnosing pupil's misconception and that new diagnosis rules can be added to the program to cater for new patterns of understanding detected by the system. The teachers responded favourably to a questionnaire regarding the user interface, the accuracy and outcomes of the questions used in the program and the accuracy of the diagnostic feedback provided by the program. In conclusion, within the limitation of the scope of the diagnosis rule base contained in the program, the research shows that such a methodology for using the available expert knowledge is feasible.