In academia, instruction usually stops at the (real or virtual) classroom door. When students enter the workforce, they’re typically on their own if they have questions about how to apply what they learned in a real-world situation. Certainly, it’s unreasonable to expect instructors to field emails, texts, and calls from former students during the workday, but a lot of learning happens while actually on the job, where people apply the skills learned in a classroom.
Learning by doing
I and many others believe learning works best when it’s driven by the learner’s questions in an interactive and personalized setting. People learn best through doing. And even within the context of online learning, we think the most value comes from using knowledge to complete a real-world task.
Certainly, many topics can be learned by reading a book or watching a video outside of a class. But when you get confused, you may need an interaction to help you understand where your train of thought diverged or to show you a different way of learning you’ve not seen before.
But providing that interaction outside of the classroom when questions arise is no simple task. How can we take that rich content prepared for a course and make it accessible in context in vivo? Great teachers don’t scale. We can’t copy their minds and hand them out to former students to help them use their skills on the job.
AI and real-time learning
This is a perfect application for artificial intelligence (AI), but the solution isn’t to replace teachers. Rather, AI can help facilitate the learning process with the help of subject-matter experts. People often think of AI as a sort of Magic 8 Ball that provides answers to questions. But I firmly believe that “black boxes” don’t belong in education, because it’s hard to learn when we don’t know know how the AI arrived at its answers. Google’s AlphaGo, for example, beat the best Go players in the world, but its moves were bewildering, even to professional players, and the AI can’t explain why it made the plays that it did.
Related: 7 ways AI will shape the future of education and work
When AI looks at and explains things in a more human way, learning is easier. In academia, we can leverage AI to understand student questions in the way humans understand it and then partially automate the process to provide people with answers that have been validated by a human expert. Understanding the meaning of a question is much more difficult than it might seem at first glance. To provide relevant answers, the AI needs to understand not just the literal meaning of the words, but also the context from which the question originates to help disambiguate what’s not understood.
Once the AI understands the question, it can then dive back into the course material to find an answer. But course materials must be structured by a subject-matter expert, who approves answers before the AI responds. This human participation is absolutely necessary, not only because highly regulated industries such as healthcare won’t allow an algorithm alone to provide life-or-death answers to a healthcare provider, but also because, without an expert, you risk the “Dr. Google” problem. Any physician can tell you how frustrating it is when patients come to their offices with wrong information they’ve gathered from a search engine, which simply shows them the most popular answers. After approval, however, the expert need not review answers to that question until the material changes, and AI can help there too, notifying the expert when content appears to have changed.
What if a student asks a question to which there isn’t an answer in the course materials? In this case, the AI becomes part of a virtuous feedback loop, alerting instructors about gaps in the course content, which can help the class better address the needs of future students.
I’ll be talking a lot about this at Open edX 2019 in March, when I will be a keynote speaker. The open-source platform is now working toward establishing waypoints in content to help learners and, potentially, AIs, understand what the content is for, how it’s used and applied. Most of the work is done after a course, when you want to apply on the job what you’ve learned.
Related: AI can humanize teaching—if we let it
AI and lifelong learning
Applying AI to constant education is not about replacing teachers. It’s a wonderful way to expand the reach of education to more employees. The trick is to balance that reach with the same rich learning experience as if the instructor were right there. Doing that requires making the platform sufficiently flexible to meet learners’ needs.
Constant learning is increasingly the norm for more and more jobs. To scale education beyond the classroom and into daily life, we need to move the learning process into work itself. That only works and scales with AI navigating, knowing when we need to know and where we need to go and scaling up mentoring.
Walter Bender is the former executive director of the MIT Media Lab and was co-founder and president of One Laptop Per Child. He is a keynote speaker at Open edX 2019.
- 8 top trends in higher education to watch in 2024 - April 16, 2024
- Defining a path to equitable AI in higher education - April 12, 2024
- Leveraging AI-driven edtech for continuous improvement in higher ed - April 11, 2024