elearning machine learning

Improving eLearning with Machine Learning is possible


Machine Learning, put to simplification, is the application of statistical prediction models that, from many and rapid observations, draw more or less accurate conclusions. From these observations, automatic actions are executed, without human intervention.

Machine learning is not something new in mathematics. What makes it new is the access to technology and the availability of lots of data. This makes viable calculations that were previously infeasible.

Machine Learning has many applications that are not exactly in the learning industry, but I will talk about just that, the application in the learning industry.


An almost obvious example: automating answers to recurring questions is possible
elearning machine learning

For example, suppose I have an elearning platform (that is, an LMS). There are many people enrolled in a program and students ask questions in the forum to the tutor or to their own inquiry box. The system could be able to interpret that similar questions have a certain answer. Or more simply, detect when a question must be answered "manually" or automatically.

Obviously, for a system to associate that a similar question leads to the same answer, it should be trained and supervised. One way to do this is to take a reference tutor and compare the answer the system would give with the one he would give. Starting from the assumption of tutor infallibility, when 99% of the answers coincide with those that he would give, we could consider that the system is effective.


And where are the savings for the business?

Let's be clear. In saving the tutor, in its reactive tutoring modality. So that either the commercial margin will be higher at the same price as the educational product, or it can be sold cheaper while maintaining margin.


And what do we do with the tutors?

Now the tutors will be trainers of machine learning systems in elearning, and there will be profitability if the training process is short, the courses are very recurrent, with a considerable volume of students and affordable machine learning technology.


Don't you think that AI can replace tutors? Well, chatbots in education are a reality

Recently, the results of a tutorial carried out with Whatson, the super system devised by IBM, were presented, in which it was stated that the students did not realize that the tutor was "non-human" and that they also rated it very well. You can't take the merit away from Whatson, or the IBM talents that made it possible. But is this really useful?

The truth is that more than 50% of the questions asked in a recurring training program are always the same (I dare say that "I have lost the password" is still the star). This is something that the tutors or, in your case, the facilitators know perfectly well. That is why in general forums, FAQs are often “posted”. What happens is that the list of FAQs is very long and it also "sucks" to use the search engine (which in addition, in those of free code often does not even work). As it becomes easier for you to ask Siri if tomorrow is a holiday in Madrid rather than to google it, and Siri is right, the automatic reactive tutor makes sense in training, either as a chatbot or as a service in a call center.

By the way, Hubtype is a much more affordable solution than Whatson with plenty of off the shelf cases for one that might interest you.


Okay, this is about substituting tutors. So AI is for low value-added applications, right?

False! AI and Big Data in educational technology are the paradigm shift.

elearning machine learning

Watch out! Let's not stay in the simple of reactive tutorials, those of if I have a question, I ask you. This is something deeper.

In my opinion, the impact is broader and here I share 5 examples that are easily understood:


1. Individual content curation

From the searches that I do both on google and on specialized educational or thematic portals or on my corporate collaborative platform, I can get to curate content on demand. It will be enough if system learns from me, from if the contents that it shows me are relevant to me, and it knows that based on whether I read them or not or if I tell the system that they are relevant or not.

Today content curation makes sense for large groups, but it is hardly "payable" individually for each employee. This would make it possible. Flipboard has been working reasonably well for years, the model is applicable within a corporate LMS.

elearning machine learning
2. Suggesting microlearning in your "dead time"

How many times have you seen a super interesting course at UDEMY that you don't do because you can't find the time or the moment to do it? Fixed and mobile devices know perfectly well whether you're on the phone, on the way to work, doing something with Excel, smoking on the doorstep of the office, or watching Netflix.

It’s enough to suggest, in those valley moments, that you dedicate time to a microlearning of value or a longer program occupying certain spaces. Your elearning platform connected to content hubs and MOOCS can help you as long as it is mobile.

elearning machine learning
3. Anticipate training needs from your peers' interests

People tend to think that we are unique, but in reality we are not so unique. That's why virals and kitten videos are so popular. In corporate environments our same role may be repeated by dozens and even hundreds of people, just look at the commercial network of a bank, insurance or automotive.

When a car dealer starts looking for information on a certain type of competing vehicle, it is likely that they are preparing to work through objections to an undecided customer. When more than 50 do, there is an urgent training alert, sales are in jeopardy.

elearning machine learning

Raising the alert to the learning community manager automatically and suggesting content of interest to disclose or "bless", after automatic disclosure, is of great value. Of course, recommending training done by your peers is an interesting clue, but this is going much further.


4. Identify valuable content inside and outside your systems

LMS cannot continue to only be the company's approved and controlled formal and informal knowledge library. If we want relevant content or valuable learning to flow to employees, we must combine several content sources and make them converge. But for content to be relevant to you, data must flow and to some extent be shared, to get other data in return.

If I know that an employee is looking for a lot of training in specialized portals, outside of my catalog of competences, it is possible that they have a deficiency in my catalog of competences.

elearning machine learning

If I have a medical case with certain X-rays or CT and determine the existence of a strange pathology, I may want to immediately see information on treatments for that pathology and specific courses. Moreover, I want to see other cases with the same identified pattern that were not diagnosed with this pathology. Support of machine learning in diagnosis and big data in the search for cases and similar documentation.


5. Identify behavior patterns in systems that help retain talenta retener el talento

Sharing data within with other systems is important: if a model detects that those who regularly do training suddenly cancel online training or do not complete it, they leave the company within two months, should an alert go off to try to retain them?

elearning machine learning

Obviously, with the LMS information I don't have enough to draw a conclusion, I need data analytics from other HR systems to establish patterns, but it triggers a call from my mentor or HR, right?

This is an important point, both if we anticipate the students who can fail, as well as if we detect the content of greatest interest, as we anticipate the possible flight of an employee, the important thing is the action we take based on the AI ​​algorithm: improve teaching, monitor more specific students, improve employee conditions ...

AI without management is useless!

And before finishing, I wanted to remind you of something fundamental that pedagogues often remind to us business managers:


Just a moment! AI doesn't make you learn what you don't want to learn. There is no learning without motivation.

Yes, having Siri or the training chatbot suggest things to you or answer your questions all day is fine. But as Darth Vader said, you shouldn't be dazzled by the technological terrors we build. There is nothing comparable to the power of force. If we understand strength as “learning”, no one learns what they do not want to learn.

elearning machine learning

If I do not want to call the customer, even if I have the sales argument to refute the competing model, if I want to play the drums instead of doing that UDEMY blockchain course, or if I am definitely going to leave the company because I am fed up … Neither Whatson, SIRI, or SKYNET are going to be stronger than my personal motivation.

And by the way, there are also algorithms to know the state of motivation of the staff, which they analyze from the words that are used in emails, chats, instant messages, etc.

Everything is to be done and everything is possible with AI!

Carles Roca

Carles is a Senior Account Manager at SEIDOR Opentrends. He leads the financial sector and is an expert in operations management and BPM, CRM, LXP&LMS technologies. Previously, he has held management positions of talent management and digital learning in consultancy firms, insurance companies and banks. His aim is to develop technology-based value propositions that help improve customer experience, increase sales or reduce costs in a way sustainable in time.