One of the topics that cropped up several times at this year’s Educause conference was analytics. This has been something of an emerging topic for some time and something that I don’t think we have really developed here in the UK, although next week’s CISG Conference may prove otherwise.
In the first case, an institution had identified a specific problem and has made use of the data it had to offer a solution to its student population. The problem was that many students select modules or pathways that are not suitable for them. This affects the completion rate as students who are on courses that they are not suited to often drop out or fail. As the level of completion was one of the criteria used to establish funding for the institution concerned, finding a solution would have a financial benefit.
One reason for students selecting the wrong options is that the language used to describe modules is often opaque – the suggestion was that the descriptions might just as well be in a foreign language for the amount of sense they made to the undergraduate trying to select their degree pathway. The solution was a course selection programme that made use of historic and personal data to recommend the modules that would best suit the students and so offer them the best chance to succeed. The grade information from all students over time was combined with the entry grades for the individual student to predict the grades for the student’s chosen path. The system then uses the course plans and predicted grades to make recommendations to the students to assist them to choose. However the student is not left to his or her own devices to make the decision. Rather the system supplements face to face advice and helps inform the discussion with advisors. There was some concern that such a system would lead to greater standardisation in the course pathways but, in the same way that Amazon or the iTunes Genius bar offer unexpected choices, it has sometimes broadened horizons. The predictions made as part of the recommendations have proved to be very accurate and that accuracy will improve further as more and more data is collated and used. And the system is delivering results – those students taking the modules recommended to them are getting better grades which in turn are providing greater incentive to those students to return. The student is also given advice on the potential careers that their major may lead them to which may also influence the student choice.
The system builds on the recognition of the fact that students have particular strengths and weaknesses and as such are going to be better suited to some paths over others. This was a problem identified in the Italian higher education system a few years ago; one of the main reasons for the high drop out rate in Italian universities was simply the fact that students were taking courses that were wholly inappropriate for their personal traits. The introduction of pre-admissions profiling resulted in students taking more suitable courses for their personalities and analytic capabilities with a consequent improvement in retention rates.
Another institution had created a student dashboard that facilitated evaluation of both a cohort and individual students’ learning. Again the system made use of both historic and current data to help guide the student towards the resources they needed according to their learning preferences and styles. The dashboard was used by the student to identify areas where they need further work in order to improve their grades, and by faculty to identify areas where the lessons delivered in their teaching had not been absorbed by their students. Consequently they were able to make changes to their courses to improve understanding and hence student success. The intention was to move to a personalised environment where a body of historic and personal data is used to understand an individual’s learning style and so better guide them to what they need to succeed.
Finally, one institution made use of a wide range of current data to assess which students were at risk of dropping out. This is common application in the UK where data is pulled from diverse sources such as the coursework submission system, library access gates, and VLE login data to pick up those who are not engaging with their university. There was a difference in the application however. The US institution analysed the data and put the students into three categories. Interestingly rather than focus on the lower tier (ie those most at risk of failing), efforts were concentrated on the middle tier to ensure they remained and improved their results. This contrasts with the approach adopted by UK institutions where support is given to the most at risk group to try to ensure they continue in their studies. This perhaps reflects the different drivers – in the US, some institutions are funded according to the numbers of successful graduates so it makes sense to put more effort into those who are likely to stay but need support to graduate. In the UK, institutions are penalised for high drop out rates and so the emphasis is on retaining those students at risk of dropping out.
Students paying higher fees are likely to expect far more in terms of support. The data institutions have built up over the years can play a role in supporting students but there is a cost in developing systems to analyse that data and tailor it to individual circumstances. The driver for institutions to invest may well emerge as the quality of student support becomes an area to seek competitive advantage.