Understanding the ‘Science of Learning’ at a deeper level has been a thorny undertaking despite copious amounts of papers being published in educational research over decades.
In most instances the investigations are centred around qualitative data based on user perceptions and surveys with relatively little evidence originating from quantitative data sets. Consequently it is suggested that our current level of knowledge is not sufficient to allow us to understand the ‘learning process’ across individual learners with the ultimate objective of ‘learning process optimisation’. The ‘Holy Grail’ of such research would be to generate some level of confidence in predicting outcome-level performance of a particular learner based on quantifiable learning-related inputs.
Is there a precedent for this elsewhere?
A useful example on how this may be achieved can be gleaned from the application of molecular genetics and genomics when investigating complex human traits relating to widespread lifestyle-related predispositions affecting people in later life such as obesity, hypertension or food intolerances.
All these are caused by a variety of inputs and influences, partly genetic, partly environmental, partly based on individual life experiences the disentanglement of which is a complex task.
In the field of genetic research striding progress has been made through a combination of high throughput data mining of the human genome, combined with studies on stratified human populations including ‘normalised’ data from mono-zygotic (identical) twins and large scale bioinformatics analysis of huge molecular datasets in order to detect any correlations between causes and their effects in a statistically significant way.
Although we are talking about highly complex multifactorial traits progress has been made over the past few years to predict with some confidence the likelihood of certain medical conditions arising well before they manifest themselves thus allowing the individual to undertake appropriate remedial lifestyle choices to pre-empt such onset (Y. Han et al. 2012).
So where is the application to learning?
Analogous to life-style ‘ailments’ learning can also be seen as being multi-factorial involving intelligence, personal traits such as self-organisation and discipline, social contexts and numerous environmental factors all contributing to the overall learning process and success.
Given the current costs to the learner to complete a degree-level course in the UK it is likely to be in their interest to obtain advance warning of any risks to them not being successful in their studies thus providing them with an opportunity to make the necessary adjustments to counteract that risk. Given the loss of direct fee income to the Higher Education Institution (and indirect costs due to KIS data returns) when a student is forced to withdraw would also justify a considerable interest and investment from a University perspective.
As a first step one could envisage a learner-specific traffic-light warning system alerting the student to a particular risk and probable outcome before it had actually occurred, for example
(i) being predicted to achieve an above average grade in one or all of the modules taken based on their current performance
(ii) being predicted to falling behind compared with an ‘average student’ based on their current study pattern (or even identifying that (s)he is studying the wrong subject area)
(iii) being predicted to significantly under-perform in or worse fail one or several modules without immediate remedial action being taken
What is holding us back?
Learning Analytics itself is a subset within the larger field of Business Intelligence (BI), a process which provides meaningful corporate data for senior executives to arrive at appropriate strategic decisions. No large organisation can compete effectively without a comprehensive armoury of BI data.
The consensus definition of Learning Analytics (according to JISC) is the measurement, collection, analysis and reporting of data about the progress of learners and the contexts in which learning takes place.
It consists of two elements (i) data arising from the direct actions of the learner undertaken as part of their study behaviour (which obviously raises concerns about data privacy) and (ii) organisational data centred around parameters such as environmental/ infrastructural/level of support given, class sizes, mode of study) in which learning takes place.
The complexity of the endeavour to obtain a deeper understand of ‘learning’ should not be under-estimated. First we have to define a baseline of what defines ‘good’ learning behaviour taking into account the variability of intelligence and multi-faceted patterns of human behaviour. For this one needs sound quantitative data accumulated from the learning behaviour and contextual background collected from several successful students, followed by developing rules which define the parameters of ‘successful learning’ at a certain level of granularity. Data mining based on Learning Analytics database queries will provide the raw data set, and pattern recognition and machine-learning will define the parameters and boundaries for any predictions made.
Currently most business-related data collected at Higher Education Institutions is held within different ‘isolated’ databases, and only retrieved at an ad hoc basis, often at the end of the academic or financial year as part of a one-off exercise. What is missing is an ongoing database interrogation/query which generates regular real-time reports (daily/weekly) and is visually displayed via a Web-based ‘dashboard’ interface to any stakeholder who is involved in the learning and management process. This is what learning analytics aims to achieve.
Sorting and analysing the multitude of data based on the different business levels (institutional, programme level, course level, module level, service level, mode of study, class or student level) would be the next step from which appropriate business responses and actions would emerge.
This short overview into Business Intelligence and Learning Analytics aims to illustrate the benefits to both the learner and the ‘learning provider’ to make the best of a symbiotic relationship between the customer and the service provider. Being able to predict whether this relationship will work effectively will undoubtedly be of mutual benefit to both stakeholders.