Page - 237 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

Figure 7. Example of use of the editor for a compound question. 5. Conclusions and Future Work Our survey of current students’ preferred approaches and resources for studying mathematical topics showed that the proportion of students preferring web-based learning resources has increased since our previous study in 2015. This emphasises the need for extensive high-quality on-line teaching and learning materials for more advanced mathematical topics, including self-test tutorial exercises which provide students with appropriate feedback on their answers. Our CalculEng system makes some progress to achieving this requirement, our new editing tool allowing us to create a more comprehensive range of exercises for students, including more useful multi-part structured questions with inter-dependencies between the answers to successive sections. This tool will make the task of creating new questions, or editing existing ones, easier, less tedious and time consuming, which should encourage more teachers to make use of CalculEng and increase the range of topics covered and the number of exercises available to students, greatly enhancing the utility of the system. Our system has attracted interest from other Higher Education institutions, and we hope to extend both the applicability and evaluation of CalculEng to a wider range of students, subject disciplines (e.g. Chemistry, Physics, Economics or Business subjects) and institutions. However, at present, CalculEng is essentially an “Expert System”, with the mathematical knowledge it uses encoded by expert teachers, rather than a genuinely intelligent system. In the future, we hope to integrate Machine Learning and/or Deep Learning approaches with CalculEng, enabling it to learn from students’ responses to the questions, and identifying what errors students actually make, rather than relying on those errors which teachers anticipate students will make when doing the teachers’ exercises. We aim to achieve this by logging all the student users’ various types of interactions with the system, analyse these statistically, noting which lead to positive outcomes and which to negative ones, and try to “learn” patterns and generate appropriate feedback from these. A methodology for such a logging of interactions, and a preliminary analysis thereof, has been carried out in the context of a system, M.Davis etal. /Developing“Smart”TutorialTools toAssist StudentsLearnCalculus 237