Type to search

Artificial Intelligence for Personalized Learning

Curriculum Planning and Development

Artificial Intelligence for Personalized Learning

Higher education has long sought to balance its mass teaching model with personalized learning. The hope was that computers could provide the personalization at low cost through adaptive learning, but these efforts floundered due to the long hours required to populate adaptive learning systems with content and branching scenarios. Plus, it is much easier to program a computer to teach topics with objectively right answers (e.g., math and languages) than those that require students to describe abstract concepts (e.g., philosophy). As a result, adaptive learning systems never progressed much beyond souped-up, self-paced quizzing systems that were not really personalized in any way.

But now AI is providing institutions with a way to offer true personalized learning due to two features. One, an AI system can learn on its own and thus does not require extensive human programming time. Today’s systems can be fed course content and extract information from it to answer student questions. Two, AI systems interact with users in natural language, allowing them to interpret student questions and answer them in a conversational format. AI can identify knowledge gaps from a student’s reply and deliver the answers they need in a format that produces genuine understanding.

A number of institutions are developing AI tutoring systems that essentially act as virtual teaching assistants to relieve faculty of the burden of answering common student questions. While these programs are in their infancy, and the scope of topics they can cover is somewhat limited, their power is growing daily as companies and institutions invest billions of dollars to bring AI to all facets of life. We are on the cusp of an educational revolution on par with online learning, and faculty who are willing to experiment with these systems will lead their institutions forward into AI in learning.

Georgia Tech’s Jill Watson

In 2016, a team led by Ashok Goel at Georgia Tech created Jill Watson, an AI chatbot that answers student questions about course structure, deadlines, requirements, and more. It was built on the architecture of IBM’s Watson system, which famously beat two Jeopardy! champions on TV. The team introduced the system to students as a teaching assistant in an online course without informing them that it was a computer. While the first iteration was generally successful at answering student questions, it made some mistakes that led students to suspect that it was a computer. Moreover, it required 1,000 to 1,500 person hours to manually program with class information.

The team learned from Jill Watson’s shortcomings and has steadily improved the system over the years so that its answers better match those that would be given by a person. Plus, they added the Agent Smith program, which analyzes a course syllabus and questionnaire filled out by a developer, to create an AI assistant in only about 10 hours of human work. It has since been run in 17 courses, undergraduate and graduate, online and face-to-face. While it is still limited to answering questions about course structure rather than course content, it saves instructors and TAs many hours of work responding to mundane questions, freeing them up to spend more time on helping students with conceptual problems in course content.

Goel’s team also developed SMART (Student Mental Model Analyzer for Research and Teaching). SMART analyzes students’ summaries of course material and gives them feedback on how well they explained the concepts in the material. This is an ideal way to help students better understand course readings. The instructor first assigns a reading and then loads a summary of the reading into the SMART system. The system analyzes the instructor summary to extract the important concepts that the students must understand. Then students log in to the system, find the assignment on their dashboard, and write a summary of the reading. The system compares the student response with its own understanding from the instructor summary to provide feedback on any student misconceptions.

Goel says that the team wants to take its systems to other institutions and has developed the National AI Institute for Adult Learning and Online Education to foster the growth of AI in learning. They welcome inquiries from faculty and others who want to partner with the team.

Walden University’s Julian

In 2019 Walden University began work on Julian, an AI tutor that takes Jill Watson a step further by helping students understand course content. Faculty load course content into Julian and specify the learning outcomes they want from the class. The system then develops a variety of assessments and activities that will help students learn the concepts. While some activities are simple multiple-choice or fill-in-the-blank questions, others require students to paraphrase text or provide a short answer to a question. The system analyzes the answers to determine knowledge gaps and follows up with information and new questions to guide the student through the topics. In essence, it creates a dialogue with the student to both measure understanding and teach, thus acting like a true tutor.

AI and you

Online education initially met with resistance from faculty, but soon nearly all institutions adopted it because it allowed millions of nontraditional students to get an education; it also created new jobs for faculty who were willing to learn how to teach online. Once college presidents and provosts saw the power of online education to bring in more students, they searched for faculty at their own institutions who could develop the programs. The early adopters became the program directors and then deans and vice presidents of distance learning.

We are in the same position with AI. Academic leaders today are being inundated with articles and presentations about institutions developing AI programs. They see how these programs not only improve student learning but also give institutions a competitive advantage in the fight for students. Once again, these academic leaders will start looking for people at their own institutions to develop similar initiatives. Faculty and administrators who experiment with AI will find themselves sought after when academic leaders ask who can lead the institution’s program. Will you be that person?

John Orlando, PhD, writes and edits articles related to online learning and teaching with technology for The Teaching Professor.


You Might also Like