Data, data everywhere

Technology is becoming an extra pair of eyes in the classroom, helping teachers anticipate when pupils might struggle or fall behind. Laura McInerney considers the revolution in learner analytics that’s helping to predict student behaviour.

Learner analyticsImagine if data existed detailing how likely it is that a primary school pupil will daydream in class, or be careless in their work, or fail to ask for help if they didn’t understand. If such data existed, should teachers use it to mitigate against pesky off-task behaviours and introversion, or is it every child’s right to daydream?  Well, we don’t need to imagine: the technology that provides this data is already out there.

Academics in the field of learning analytics use learner behaviours to predict how students will do in the future and to try and change any less-than-positive behaviours.  Incentivised by the need to keep fee-paying students, US universities are leading the analytics revolution.  Using popular e-learning platforms such as Blackboard and Signals, analysts have found that specific patterns of online behaviours correlate with student success.  Using the patterns, analysts create algorithms for student ‘success’ and then map current student behaviour against the formulae. If a student is exhibiting behaviours far outside the norm, the system flags their likely failure, both to the student and to specialist tutors deployed to provide additional services – e.g. mentoring, remediation classes or financial aid. So far, the evidence suggests the system is having a positive impact on student retention and performance.

Though so far analytics is mostly focused in higher education, the field also holds promise for schools and colleges. Ryan Baker has spent the past decade using student’s online behaviours to monitor their willingness to ask for help, the speed at which they tackle problems and their level of distraction.  One intriguing finding is that students in urban schools are ‘off-task’ twice as often as students in suburban schools suggesting that the local environment strongly impacts student behaviour. But more impressive is the finding that certain online behaviours also predict a student’s ability to learn in future unrelated tasks.  A willingness to ask for help when stuck, and taking a reasonable amount of time to use that help, is vitally important in the learning process. Where students answer very quickly after being helped, or if they take a long time to answer, they demonstrate a lower propensity for future learning. This shows that asking for help and using that help mindfully is a very important predictor of how well students will perform in the future. Even more impressively, after working on a computer for a short period, the data can show how likely a student is to engage in either of these behaviours.  Armed with this information a teacher can think carefully about how they will structure activities and ensure extra attention is given to students with a lower propensity to ask for help.

However, even where teachers have information they can be lazy about acting on it. In one analytics evaluation project researchers analysed online discussions completed by students and their teachers. Though teachers were shown clear evidence that they spoke more often to learners with the highest ability at the expense of lower ability students, teachers still tended to favour ‘top’ students in subsequent conversations.  This is a shame, as the evidence from learning analytics suggests that where lower ability learners are engaged in conversations this correlates with higher achievement and course satisfaction.

At present teachers are like turn-of-the-century doctors limited in terms of the tools we have for understanding what is actually happening in the minds of our ‘patients’. In the 20th century doctors became able to screen our hearts, bones, blood and brains and in doing so made better decisions about how to treat us.  The next step for the field is asking: What data do teachers want in order to help students more?  What do we want to know more about? Learning analytics can provide us with many answers, the trick will be finding out what are the most important questions we should be asking.

Laura McInerney was a teacher in East London for six years. She is currently a Fulbright-sponsored PhD student at the University of Missouri and is a Policy Development Partner at LKMCo.

  1. Arnold, K. E. & Pistilli, M. D. (2012). Course Signals at Purdue: Using learning analytics to increase student success. Proceedings of the 2nd International Conference on Learning Analytics & Knowledge. New York: ACM. (Retrieved 10 September 2012).
  2. Baker, R. S. J. &  Gowda, S. M. (2010). An Analysis of the Differences in the Frequency of Students’ Disengagement in Urban, Rural, and Suburban High Schools [PDF]. In Baker, R.S.J.d., Merceron, A., Pavlik, P.I. Jr. (Eds.) Proceedings of the 3rd International Conference on Educational Data Mining. (Retrieved 10 September 2012).
  3. Ali, L., Hatala, M., Gašević, D. & Jovanović, J. (2012). A qualitative evaluation of evolution of a learning analytics tool, Computers and Education, Volume 58, Issue 1, January 2012. (Retrieved 10 September 2012).

Share this page