Facilitated by Dr Kirsty Kitto, Institute for Future Environments & Faculty of Science and Engineering, Queensland University of Technology
Learning Analytics is fast moving from a technology on the horizon to an institutional must have. We are in the middle of a rapid take up of technology that increasingly relies upon computational methods such as machine learning to perform rapid analysis of large datasets that could not traditionally be understood in anything but the most basic form. This enables us to develop a much finer understanding of our students. For example, we can classify students who are at risk of failure; automatically detect traces of cognitive presence in communities of inquiry; find underlying themes that a large class is talking about; understand which students are well connected with their peers etc. The accuracy of these algorithms is rapidly increasing, but it is worth considering whether it will continue to do so indefinitely. Context effects mean that many of the traits we can identify in one data set do not necessarily carry across to another one, and even within a data set obtained from one institution we often see that differences in teaching style, student preferences and underlying learning goals make it almost impossible to obtain the kind of accuracies common in many other fields of data analysis. Care is required in analysing learning data. Furthermore, we might ask whether we should want our algorithms to be as accurate as possible for the purposes of learning. Machine learning can help to make a very effective scaffold of student learning, but some of the most interesting applications in this area rely upon slightly inaccurate classifications to get a student to think about their own behaviour, and to challenge the machine. This puts the student at the core of a human centred analytics system, which gives them much more control of their own self-directed learning. This talk will explore these issues of automation, building up an argument that a very real class of imperfect learning analytics solutions exist which use inaccuracy as a lever to enhance both formal student learning and their digital data literacy.
Kirsty Kitto is a Senior Research Fellow in the School of Mathematics and the Institute for Future Environments at Queensland University of Technology. She models the way in which people interact with complex information environments using both mathematical and computational approaches. She is currently leading the OLT funded project “Enabling Connected Learning via open source analytics in the wild: Learning Analytics beyond the LMS” which is using Experience API (xAPI) to unify data obtained about students learning that occurs within pre-specified social media environments and present them with dashboards to help them understand their own learning processes. She is currently serving a term as an inaugural member of the board of directors for the Data Standards Interoperability Consortium (DISC), which is shepherding the xAPI standard as it is released to the worldwide community by ADL who developed it for the US Department of Defence as the new educational data standard to replace SCORM.
If you would like to join this research group or would like more information please contact Lorenzo Vigentini firstname.lastname@example.org
Room 1025, Level 10, Library Tower