eLearning and its associated online resources, particularly analytics for eLearning, have become a source of educational big data, as reams of information are gathered charting the progress of learners on courses and training schemes. The study and interpretation of all this output and feedback has emerged as the field of learning analytics.
Predictive analysis of educational Big Data seeks to take this study a step further, anticipating what a learner requires in future based on their performance now, and tweaking educational experiences to create truly personalised learning. Personalisation itself is set to be the tech buzzword of 2016 as more and more companies look to how they can apply it to the shopping, learning and wearable technologies.
As they proceed through a course, information is collected from each participant: how often they log into their learning platform, which exercises they went through today, their responses to quiz questions, comments in discussion forums, etc.
Special programming algorithms and machine learning tools are applied to this data, to help project the learner’s likely performance in future exercises of this type, what information should be made available to help them perform better, which activities are likely to engage them more, and so on.
A Matter of Interpretation
Traditionally, several forms of data have been collected at the end of an eLearning course, such as assessments of the knowledge taken in and retained, the impact of the training on business operations, and feedback giving reactions to the course content.
All these data types are descriptive snapshots measuring things that have already occurred. But by shifting your perspective, it’s possible to apply these metrics in ways that can augment and improve future eLearning endeavours.
Having identified data pointing to changes that might be made to improve matters, you’ll then have to test these changes out, on the next run of the training programme. And again, the feedback you receive after that next course will reveal if those changes were successful – in addition to providing ideas for further improvement.
With a future-looking mindset to the data being collected from ongoing courses, you’ll then have to analyse that information to see what useful patterns emerge, then create a plan of action to address all those issues.
Broad Goals of Predictive Analysis
Predictions based on Big Data analysis have for some time been an integral part of business operations, particularly in the field of customer relations, where information is analysed to predict the behaviour of buyers, based on their past consumption patterns, and reactions to past promotional campaigns.
Predictive analytics have also become a mainstay of talent management, where organisations study data on the job performance of their staff in efforts to predict which ones have the greatest potential to be future managers, team leaders, and star performers.
In an eLearning context, predictive analysis efforts have been mainly targeted at creating the personalised learning experience. The goal is to produce truly intelligent learning systems that can respond in real time to a learner’s specific needs or aspirations, due to predictions gleaned from the data collected on their activities.
Such a learning system would be a responsive one, capable of changing the content or direction of a course to better suit the skill level, personal traits and performance of the person taking it. All of this without imposing a huge design and monitoring overhead on the course creators – or demanding too much of the predictive analytics software.
Ideal Features of the Software Platform
The learning platform should ideally split a course into brief and logically connected modules, each of which are reusable, and set out clearly enough for easy data gathering and reporting. WIth cloud-based authoring tools, it should be possible to create relationships between course modules, and to track their usage (e.g. when data is shared on social media, or between Learning Management Systems).
The analytics should be powered by an engine such as the xAPI programming interface, with the capability to store data records offline, and to bring together and study data sets from several learning systems. With the ability to house data from different LMSs, it’s possible to create analytical comparisons based on the performances of a wide range of users.
So, an eLearning platform based on such predictive analytics could for example suggest to learners what their next course step should be, based on what other learners in similar situations have done. Or omit parts of a course altogether, if the learner already possesses that knowledge. Or suggest repeating certain exercises, if the necessary skills haven’t been taken on board. And so on.
For the User…
Personalised training modules based on Big Data come with the obvious questions as to how much information is being gathered about learners, who has access to it, and how it could possibly be abused. So confidentiality and privacy need to be built in, from the start.
Users need to be given the power to opt out if they choose, and have laid down methods of redress and complaint, if they feel that personal information has been abused.
For the Course Designer…
An adaptive learning system may give useful feedback not only to learners, but to course designers as well. For instance, if a training module simply isn’t hitting the mark with users, the system may be configured to alert the course facilitator to this fact – and may suggest removing that module entirely.
And for privacy purposes, any data collected on the users should have an agreed expiration date and lifespan, together with built-in mechanisms for learners to request that data be removed, if certain conditions are met.
Some allowance for random variation should be made in the course recommendation tools. In this way, modules may be tweaked in terms of presentation, style and content to make the personalised training experiences of each learner truly memorable.