Artificial Intelligence and Machine Learning

AI Tutoring: Genius on My Shoulder

A humanoid robot assisting a teacher.
Automated tutoring products driven by artificial intelligence are poised to become a significant presence in classrooms within the next decade.

Already used by hundreds of thousands of students, automated tutoring products driven by artificial intelligence (AI) are poised to become a significant presence in classrooms within the next decade, forever changing the way learning is done.

Key to that proliferation is AI tutoring's ability to personalize education.  Essentially, with an AI tutor at their side, a student has an incredibly smart digital friend with unlimited patience and 24/7 availability, skillfully programmed to sense the best way to impart knowledge to that particular student.

"The advantage of an AI system is that it can give 100% attention to each student, and that it can tailor material to an individual student's abilities and needs," says Kevin Leyton-Brown, a computer science professor at the University of British Columbia.

The companies providing AI tutors have found their greatest success in targetting highly specific, well-defined disciplines in which the answers students seek are clearly right or wrong.

For example, Assessment and LEarning in Knowledge Spaces (ALEKS) by McGraw-Hill Education uses adaptive questioning to assess what a student knows and does not know, then tailors lesson plans for instruction in math and the sciences. ALEKS assesses a student's prowess on any given subject without the need for input from a teacher by asking 20-30 questions (its choices of questions can change, based on each answer provided).  The responses allow it to assess the student's mastery of a particular subject, and then to present specific topics it believes the student is ready to learn, offering practice problems that teach the topic. To ensure topics learned are retained in long-term memory, ALEKS periodically reassesses the student.

AI tutors that work in this way include products from Thinkster Math, Whizz Education, Third Space Learning, and Carnegie Learning 's Cognitive Tutor product line.

"Hundreds of thousands of students a year already use AI-based systems like Cognitive Tutor and ALEKS, and existing computer-based blended learning systems are increasing their degree of use of AI year-by-year," says Ryan S. Baker, a professor specializing in intelligent tutoring and educational software at the Graduate School of Education at the University of Pennsylvania.

ALEKS has been used by more than 15 million students since its initial release, and is being used by 3.5-million K–12 and college students in 2017, according to Wil Lampros, ALEKS chief product officer at McGraw-Hill Education. "ALEKS has been built upon the principles of widely accepted learning science, and institutions across the country have documented how it has helped improve grades, retention and graduation rates," Lampros says. "Many institutions have seen pass rates increase by as much as 20-50%."

According to Lampros, ALEKS has a 94% success rate with students.

Still other AI tutors like Duo Lingo and Carnegie Speech aim to teach students foreign languages; again, an educational discipline in which there are clear right and wrong answers.

"It's not far away at all from when a majority of students in the U.S. will be using AI-driven tutoring systems for at least one of their classes," Baker says.

As with most things AI, many of the technology's enthusiasts see AI tutors becoming much more generalized as the technology matures. In the longer term, AI prognosticators anticipate the emergence of the 'lifelong learning companion': a single AI device that one will consult for specific areas of knowledge throughout one's life. The beauty of such a system is that with AI as its bedrock, a learning companion would grow ever more perceptive about how you learn best and, as a result, would grow ever more personalized about how knowledge is imparted to you.

"I do believe that 'lifelong learning companions' will be a reality," says Rose Luckin, a professor of Learner Centered Design specializing in AI/educational technology at University College London. "We have all the technology we need to build such companions; admittedly, not particularly sophisticated companions, but the sophistication of the companion will grow over time and with their learner."

Chao-Lin Liu, a professor specializing in AI at National Chengchi University, Taiwan, agrees. "In principle, an AI system that observes/accompanies a person for a long time will gather more information about the person, and so can offer the best personalized services possible."

Liu is concerned that AI-driven tutoring systems are too often limited to providing learning in subjects where answers are clearly right or wrong,  such as math. AI-driven systems (at least so far) are less effective in many courses in the humanities, where "discussion is the process of learning, and there is no perfect answer."

Still, says Liu, "An ideal AI-based tutoring system offers individualized assistance to the learners, which is hard and very expensive to achieve if only human teachers are available."

The University of Pennsylvania's Baker insists the technology will transform classrooms into hybrid learning environments, where AI tutors will handle some of the teaching tasks, while human counterparts are freed up to work with individual students.

"When AI-driven tutoring systems are used in classrooms, it allows the majority of students to proceed at their own pace, while enabling the teachers to focus their time on working one-on-one or with small groups of the students who are struggling right at that moment," says Baker.

On the other hand, says Pierre Dillenbourg, a professor specializing in learning technologies at the Federal Institute of Technology in Lausanne, Switzerland (and a former elementary school teacher), "The main pitfall of learning technologies has always been to promise too much; too many gurus, disconnected from empirical research."

Even so, it's hard to look at what the technology has already delivered—along with the promise of what could be—without saying, "I want one of those."

Joe Dysart is an Internet speaker and business consultant based in Manhattan, NY. 

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More