Guide to the AI Driving Innovations in Online Learning
The Industrial Revolution brought huge changes to the education system. Math and reading skills became more important as factory jobs replaced agricultural jobs. State education shifted to a factory model, where students would learn the skills that would serve those factories best. The coming revolution in artificial intelligence (AI) could usher in a similar tectonic shift in how we teach—and how we learn.
Since the mid-20th century, education has focused on specialization: learning more about less. But in a world of growing automation, that specialized knowledge becomes obsolete quicker. Students will increasingly come from all different age groups and backgrounds. They will need to learn new skills and relearn old ones over the course of varied careers. In the future, online education is going to be for everyone, and AI could help bring it there.
Online education has already gotten a huge boost from AI, if only by association. Major online learning resources like Udacity, edX, and Coursera were born out of the nation’s top AI labs and founded and/or headed by AI experts: Sebastian Thrun at Stanford (Udacity), Andrew Ng at Stanford (Coursera), and Anant Agarwal at MIT (edX). Now, that association could become a firm relationship, as AI-powered modules are coming online in every area of education.
Delivering courses online has already lowered costs, reduced inequality, and improved graduation rates in education. The AI revolution could, in turn, make online education smarter, faster, and cheaper still. It’s already started.
The Three Main Types of AI in Online Education
Adaptive learning is education software that’s customized to each student individually, such that concepts are presented in the order a particular student finds easiest to understand and are able to be completed in a self-set pace. Currently, adaptive learning models function optimally when a large cohort of students have to learn the same material, thus allowing a large amount of comparable data to be collected simultaneously. Their progression could see them introduced on a much more granular level.
Geekie, an adaptive learning startup in Brazil, is already delivering aspects of the high school curriculum to thousands of schools at once. Apps like Cram101, from Content Technologies, use AI to break down a textbook into a digestible study guide, with chapter summaries, practice tests, and flashcards.
Overall, adaptive learning tools will continue to make learning quicker, smarter, and more individualized.
Intelligent Tutoring Systems
Intelligent tutoring systems are AI-powered solutions tailored to match each individual student’s needs and abilities. MATHiaU, from Carnegie Learning, mirrors a human coach in its ability to provide feedback, rephrase questions, and give detailed assessments of a student’s progress. MATHiaU is focused on remedial math courses for college students, which, when pursued in the traditional manner, cost $6.7 billion for what is only a 33 percent success rate.
Intelligent tutoring systems can cut that cost dramatically while boosting the success rate at the same time. Bartleby, developed by Barnes & Noble Education, includes an AI-powered writing module that corrects grammar, spelling, and punctuation, while also searching for plagiarism and encouraging proper citation; it even provides a preliminary grade on a paper.
These systems, at their best, don’t replace teachers but rather shift teachers closer to the role of mentor.
Imagine if video games and chatbots had a baby, only the goal wasn’t a high score or a customer service query’s resolution but advancing a user’s education. That’s the soon-reality of virtual facilitators.
USC’s Institute for Creative Technologies has a head start here: already well-versed in creating AI-powered 3D environments and realistic virtual characters, they’ve also developed prototypes in virtual counseling for the US Army. Further work on Captivating Virtual Instruction for Training (CVIT) aims to merge live classrooms with intelligent tutors, augmented reality, and virtual instructors.
Meanwhile, IBM’s Watson has a virtual teaching assistant, Jill Watson. First introduced at Georgia Tech in a course titled “Knowledge-Based Artificial Intelligence,” Jill engages on a classroom’s online discussion forum, answering student questions alongside other, human teaching assistants. In many cases, she even outperforms her human colleagues by answering more quickly. In 2016, Georgia Tech students couldn’t discern which of the teaching assistants was an AI program.
The future’s not as far away as we might think.
The Bottom Line: The Future Applications of AI in Remote Learning
It’s still the early days for AI-powered online education—and self-improvement is baked into the process. AI systems are more transparent than human teachers: one can more easily retrace and audit the ‘thought process’ behind a teachable moment.
As more AI systems are deployed, they’ll have more data to draw upon—though that data must be collected in an ethical, secure, and transparent manner. Over time, these tools will become more fluid, more natural, and more effective. With improved AI-enabled online education, students will be able to learn faster than ever before, and they’ll be increasingly able to learn skills that will serve them in an automated future.
Pitfalls, however, remain. In 2009, the IMPACT program deployed AI and machine learning processes across all of Washington DC’s 109 public schools. The aim was to measure teacher performance, provide curated feedback, and improve the overall standard of education. The results were poor. Many teachers spoke out against it. Others gamed the system to receive better marks. The system itself had glitches that incorrectly lowered the scores of nearly four dozen teachers. The Washington Post called it a scandal.
Future innovations in AI-education must focus first and foremost on meeting the needs of the people who use it: teachers and students. And further inroads must be made to integrate these new innovations into the existing system. In a survey by the Economist Intelligence Unit, only 38 percent of responding teachers felt their training equipped them to use digital technology for instruction.
In the future of AI-enabled education, teachers, students, and AI developers still have much to learn.