4 Ways Machine Learning Can Improve Online Education
Machine learning offers some exciting possibilities for online education, but they may be a while in coming.
Any year now, machine learning is poised to take online education by storm. If you’re unfamiliar with the term, machine learning is a form of artificial intelligence that helps computer systems analyze data sets while performing certain tasks, and then gradually improve performance based on that data feedback without need for further programming.
It’s an upgrade on current automation systems. While it’s easy enough to set up an automation system that performs the same tasks over and over again, creating one that self-improves is another matter. And yet, this is exactly what machine learning aims to do.
Wondering how this technology might fit in with online education? Predictions are, as yet, hard to make. But machine learning certainly has quite a few promising applications. Here are just a few.
1. Adjust course delivery to match the pace and capacity of the learner.
One of my favorite learning tools is a “spaced repetition” flash card program called Anki. You design and set up your own flash cards, and it delivers them using an algorithm that gradually lengthens the time between card reviews. If you keep answering cards correctly, it waits longer and longer to show you that card again. This prevents learners from over-investing time in cards they already know, and keeps the focus on the material they have to learn.
Now, this system is based on my input, so it’s not what you would call machine learning. I tell the program whether I know the card, and how hard or easy it was for me to remember the answer. Spaced repetition systems have been around for a while, and they have proven marvelously effective in memorizing information. How could that be applied to machine learning?
What if a program could tell, just from how I interacted with content on the screen, whether I understood the material? And what if that program could adjust the content to match my learning, either by speeding it up or slowing it down? Like the spaced repetition system, this would help learners concentrate on areas that need more practice without losing time on material they had already mastered.
2. Get rid of administrative busywork through smart automation.
One common complaint about automation is that it removes the human element from education. Students do best with genuine teacher interactions—not AI ones that merely mimic them. Yet the more important teaching roles are turned over to AI, the more this is likely to happen.
However, the picture changes when machine learning is applied to administrative tasks. Now, instead of replacing teachers, AI is freeing them from the burden of tedious busy work, and leaving them with more time to devote to the classroom.
3. Personalize content for a learner’s strengths and weaknesses.
Like the previous example in which machine learning software adjusted pace, there’s no reason it couldn’t also adjust the content itself. This could become especially useful in scenario-based learning, or in gamified elements.
For instance, let’s say you’ve created a customer service program to help your learners respond to unhappy customers. Scenario-based learning is a great way to tackle this subject, but it can take a lot of planning, and options for learner responses can quickly spiral out of control. However by turning this over to a machine learning program, learners would have a more robust way to practice their customer management skills.
It’s clear to see how this might also apply to gamification. As the program learns how to respond to the learner, it can adapt the work to be more or less challenging, based on learner needs.
4. Respond to learner questions and provide intelligent feedback.
When learners are struggling with subject matter, they may need extra help to work their way through. But this individualized feedback can be time consuming and if there are a lot of learners on a course, instructors may not have the resources they need to adequately respond to students.
But what if, instead of talking to an instructor, machine learning software were capable of understanding the question and responding at a level the learner could comprehend? To achieve this, machine learning would need to be advanced enough to parse the content of the question, no matter how the learner phrased it, while also delivering a response that is neither too simplistic nor too complex.
This is certainly an exciting opportunity for machine learning. Although natural language processing still has a ways to go before this technology could provide a helpful user experience. The biggest danger is that it might become the equivalent of an automated telephone answering system, where learners feel barred from speaking with the instructors, but also can’t make the system answer their questions.
However, unlike the answering system, the machine learning program could offer better answers over time. The user input and feedback could quickly improve the service, making it a real boon to learners and instructors alike.
Machine learning is on its way, but it won’t replace the roll of educators.
Undoubtedly, machine learning and advanced AI will open a lot of opportunities for educators and learners alike. But it’s important to recognize the limits of this technology as well.
Personally, I am most skeptical of suggestions that machine learning will be able to generate content—or even whole courses—without a great deal of supervision. At best, such courses could only be derivatives of other courses, and more likely than not, learners would end up with material that wasn’t as well-matched to their needs as that which an educator might provide.
However, machine learning can help match content to learners, while also relieving educators or many administrative burdens. It’s hard to say when this technology will become accessible enough to hit the mainstream of online education, but it remains an exciting development to keep an eye on.