The trope of artificial intelligence (AI) systems rising up to make human beings obsolete is a common one in science fiction. But University of Alberta educational psychology professor Jason Harley says its dominance in the public imagination does a disservice to the potential for AI to help human learners succeed in educational contexts—and to ensure access to educational supports for all students.
“The question I most often get as an AI researcher is ‘are you planning on replacing teachers?’ That’s problematic,” said Harley, director of the Computer-Human Information: Technology, Education, and Affect (CHI-TEA) lab at the U of A. “What we should be asking is: ‘what can artificial intelligence do to improve learning? How can we use AI to open up possibilities for teachers, to let them do new things—even in crowded classrooms, and in situations where there continue to be economic challenges?’ Replacing teachers is not part of the research agenda.”
Setting realistic learning goals
Harley’s own research recently looked at how learners can work with artificial intelligence “pedagogical agents” to define the objectives of a given tutoring session—in the case under study, learning about the human circulatory system. This kind of collaborative interaction not only makes for a more effective use of study time, but may have positive impacts on motivation and later learning, Harley said.
“Collaborative learning is an area of research that’s emerged over the last decade, but all that research is focused on interactions between humans. My colleagues, Drs. Taub and Azevedo [of North Carolina State University], and Dr. Bouchet [of Sorbonne Université and Laboratoire d'Informatique de Paris 6], and I wanted to see whether these new frameworks that have strong empirical foundations can be applied to human-AI interactions,” Harley said. “When students have a chance to set learning goals and engage in learning opportunities that are more customized to their interests and backgrounds, this can help make learning meaningful and has motivational benefits.”
Leveling the playing field
While further study is required to optimize interactions between students and AI tutors, Harley said early indications are that learners don’t mind negotiating an educational objective with a non-human agent.
“There’s been some concern about whether people are going to be willing to listen and take tutorial suggestions from a program employing AI. What we found in our recent study was that, overwhelmingly, people were really ready to listen to and act upon these suggestions.”
Harley said that, rather than making teachers obsolete, AI pedagogical agents have the potential to help fill in gaps where students need extra help that may not be available in a typical classroom.
“One of the most interesting areas for AI is exploring what it can do to level the playing field where students have different access to tutors, different levels of prior knowledge, or learn at different rates,” he said. “There’s all kinds of things AI can do—even provide emotional support.” Toward this end, Harley has also published a taxonomy of approaches and features for developing emotion-ware advanced learning technologies with colleagues at McGill University and University of Montreal.