Making education more effective, equitable and available to every child in the world has long been the holy grail for education technologists and entrepreneurs, who have developed countless solutions, like massive open online courses. Now, with advances in artificial intelligence (AI), we can take this a step further and pursue adaptive education.
The core promise of adaptive education is an intelligent, 1:1 experience for every student. Research has long established that students who receive personalized, one-on-one tutoring from human teachers outscore their peers on final exams. As early as 1984, Dr. Benjamin Bloom, the late educational psychologist, showed that students who receive personalized tutoring score higher than almost all of their fellow students who attend standard classroom lectures.
The implications are dramatic: We can build a computer-based tutor for millions of students around the world.
Adaptive education represents a paradigm shift—from the conventional model (an instructor-centric, passive learning experience) to an intelligent one (a learner-centric, interactive, active learning experience). In the adaptive model, each student is paired with a virtual “coach.” It’s a concept that can be scaled to millions of students at a fraction of the cost of human tutors.
The full spectrum of technology innovation in education has long been understood using the SAMR framework—for substitution, augmentation, modification and redefinition. Adaptive education powered by AI, not surprisingly, fits into the final stage of “redefinition,” where transformative experiences truly happen and learning itself is redefined. (For more information on these perspectives, check out Rewiring Education, for which I co-wrote a Chinese version.)
How does an adaptive AI teaching system work? It performs three continuous functions:
- Assessment: It can diagnose or assess a learner’s knowledge on all or part of the target area of learning.
- Targeted content: It makes personalized recommendations such as different content, depending on the learner’s knowledge assessment.
- Real-time interactions: It leverages the strategic use of AI at the full process level (rather than at the task level) to make instructional and task selection decisions. An adaptive platform is more intelligent, for example, than AIs that simply find answers to scanned test questions using image recognition and pattern matching mechanisms.
AN EVOLVING MODEL
How far away are we from putting adaptive learning platforms to work? In order to deliver on their promise, companies must first build an adaptive learning engine that can acquire, store and analyze data. At my company, we have invested heavily in the most advanced machine learning algorithms and continue to conduct research that can improve not only our products but also the AI and science that drives them.
Derek Haoyang Li
For example, one machine learning model we developed—called a multidimensional probabilistic knowledge state—uses data from learners to measure their proficiency across a wide range of variables. The application module that uses an AI algorithm can measure a student’s approximate proficiency level on each of 500 knowledge factors using just 25-30 questions.
For years, conventional education has been defined, in part, by geography and the financial means to pursue college degrees. New types of online learning in recent years have been helpful in overcoming some of those limitations, but they haven’t solved the core challenge that adaptive AI can—to shift the teaching model from one-to-many to one-to-one.
Ultimately, I believe adaptive systems—with the combination of AI and human coaching— represent our best hope for giving all students the chance to reach their full potential.
CREDIT: Klaus Vedfelt/GettyImages; Courtesy of Derek Haoyang Li
Credit: Google News
