At the turn of the year, most universities and higher education institutions analyse the previous year’s performance targets by degree level or by other selected criteria. During that period, many educational institutions are launching the following year’s academic degree structures and preparing the various teaching modules. In how many institutions have the results been predicted using the traditional approach, such as, for example, that 70% of all beginners will graduate on time? And should this prediction model continue to guide the implemention of teaching for the coming academic year, as usual?
It is somewhat baffling that, at the same time as the AnalyticsAI project is being implemented, migration to SISU and Peppi is underway, whereby part of the prediction model development is likely to be learning the features of the new systems rather than developing the actual analytics tool. Why don’t all have the same system in place? The most recently adopted Data Protection Act creates its own challenge into the overall picture. This in turn has led to the fact that even in the development of analytics we have had to be very cautious in developing “real” forecasting models.
Fortunately, in the project, we have jointly identified the risks described above and are able to look at the whole picture openly by first identifying and recognizing the actual prediction models already in place and at least in use at some universities. Only then will we develop new, appropriate analytics’ tools. In doing so, we will not come up with something that is already available from SISU or Peppi. To support the development of analytical tools, it is thus reasonable to think of a three-part timeline: the existing tools, tomorrow’s tools, and the tools for beyond tomorrow’s prediction model.
At the moment we can quite reliably extract from SISU, for example, student-specific PSPs, study load data by period, number of graduates by educational product, etc. However, these do not really predict anything, but describe what has already happened. However, with a bit of work, it is relatively easy to have the opportunity to retrieve the follow-up data for each student or group of students by year of study, and to estimate the number of degrees to be achieved by educational product, anticipating the development prospects of the next 3-4 years. These examples illustrate the attempt to present analytics results as trends that better support the choice of measures required both in the progress of students and their studies, and in the management of educational products. Using SISU’s features, it is also easy to generate system-based alerts for students and management alike if, for example, study time threatens to stretch or the degree goals do not appear to be met. In such cases, the counteractive action may be taken immediately as opposed to an a posteriori analysis of what might have gone wrong.
Towards the end of the project, we will probably be prepared already to present a tool to support the design of educational products featuring the ”day after tomorrow's” forecasting model, which will allow employer feedback and labor market skills needs to be incorporated into the productivity analysis of an educational product. As we have a strong view of the future of the project, we think it is more than justified to continue with the development work in the future, even beyond “the day after tomorrow”. Hopefully, the project sponsor accords with our vision when the ideation of follow-up projects becomes topical.
Katriina Mielonen ja Harri Eskelinen
Lappeenranta-Lahti University of Technology LUT