Our world as we know it is running on artificial intelligence. Siri manages our calendars. Facebook suggests our friends. Computers trade our stocks. We have cars that park themselves, and air traffic control is almost fully automated. Virtually every field has benefited from advances in artificial intelligence, from the military to medicine to manufacturing.
However, almost none of the recent advancements in artificial intelligence have advanced the education industry. Why is education lagging behind? Why has the momentum for artificial intelligence in education seemed to have largely faded in the past few years?
Woolf, et al., (2013) proposed some “grand challenges” that artificial intelligence in education should work to address, including:
Virtual mentors for every learner: Omnipresent support that integrates user modeling, social simulation and knowledge representation.
Addressing 21st century skills: Assist learners with self-direction, self-assessment, teamwork and more.
Analysis of interaction data: Bring together the vast amounts of data about individual learning, social contexts, learning contexts and personal interests.
Provide opportunities for global classrooms: Increase the interconnectedness and accessibility of classrooms worldwide.
Lifelong and lifewide technologies: Taking learning outside of the classroom and into the learner’s life outside of school.
Over the last decade, applications of artificial intelligence have addressed several challenges of learning, including language processing, reasoning, planning, and cognitive modeling (Woolf, 2009). Known as Intelligent Tutor Systems, computer software is able track the “mental steps” of the learner during problem-solving tasks to diagnose misconceptions and estimate the learner’s understanding of the domain. Intelligent Tutor Systems also can provide timely guidance, feedback and explanations to the learner and can promote productive learning behaviors, such as self-regulation, self-monitoring, and self-explanation. Furthermore, Intelligent Tutor Systems can also prescribe learning activities at the level of difficulty and with the content most appropriate for the learner (Azevedo & Hadwin, 2005; Shute, 2008; VanLehn, 2006). These systems are also able to mimic the benefits of one-to-one tutoring, and some of these systems outperform untrained tutors in specific topics and can approach the effectiveness of expert tutors (VanLehn, 2011). Noteworthy examples of these intelligent tutor systems include Tabtor, Carnegie Learning and Front Row. A meta-analysis comparing learner outcomes using Intelligent Tutor Systems to learner outcomes using other instructional methods found that over a wide aware of conditions, learning from Intelligent Tutor Systems led to higher outcome scores (Ma et al., 2014).
In another application to learning, artificial intelligence can help organize and synthesize content to support content delivery. Known as deep learning systems, technology can read, write and emulate human behavior. For example, Dr. Scott R. Parfitt’s Content Technologies, Inc. (CTI) enables educators to assemble custom textbooks. Educators import a syllabus and CTI’s engine populates a textbook with the core content.
Progress in artificial intelligence and machine learning has been impressive, but there is still much work to be done to advance learning science. While some progress is being made to bring artificial intelligence to the education space as described above, these efforts pale in comparison to advancements in the non-education space. Most of the exciting breakthroughs in 2015 were in fields outside of education. For example, companies such as Amazon and UPS have been piloting the use of drones to deliver packages and other goods to customers. Google recently purchased an AI software company, DeepMind, from a British startup for half a billion dollars. Google has dedicated more than 140 computer scientists to DeepMind, and the software recently taught itself how to play 49 retro video games so well that it consistently outperforms human players. Google has also been testing its driverless cars. PR2, a robot from Cornell University, learned how to perform various small tasks, and then “taught” Baxter, another robot from Brown University, how to perform the same tasks in an alternate setting. Another robot, ConceptNet 4, took an IQ test with tasks in vocabulary, comparisons and comprehension and was found to have the intelligence of a 4-year-old.
I believe that artificial intelligence could play a role in the growing field of learning analytics, evaluating the quality of curricular materials, and in adaptive learning and recommendation engines. There is also the potential for artificial intelligence to create unique learning pathways for individual learners in MOOCs and blended and online learning(Chaudhry, et al., 2013). (For more information about the potential for artificial intelligence in these areas, check out a recent special issue of AI Magazine.)
The possibilities for artificial intelligence to make significant contributions in any field are tremendous, and education shouldn’t be left behind.