Research News

Social audiences disrupt ‘learning by teaching’

Towers of Hanoi
The "Tower of Hanoi" task (Courtesy Dan Levin)

“Learning by teaching,” a method in which teaching facilitates the tutor’s own understanding, may be improved when the audience is not human, new research from Vanderbilt University’s Peabody College finds. The study, based on research that suggests a person learns best when teaching a concept to someone else, compared teaching a person against teacher a computer. The researchers  found the teacher learned less when teaching the human audience then when teaching the computer.

The study authors asked participants to solve a basic, three-ring version of a Tower of Hanoi puzzle and demonstrate their solution to a computer or person. The audience was then removed and participants were asked to solve a more complicated, four-ring version of the task.

The researchers predicted better learning in the human audience condition, assuming that demonstrating actions to people would lead to better results because it would induce deeper, more goal-focused reasoning. However, in two separate experiments, they found participants were significantly worse at solving the task after showing the more basic task to a human audience.

“Demonstrating the solution to a social audience depleted participant’s capacity to subsequently transfer their understanding toward solving the more difficult problem,” said lead study author and Peabody graduate Jonathan Herberg. “[rquote]Social highlighting led to longer solution times and less optimal solution paths, which suggests that engaging in social teaching is not always beneficial and can potentially result in costs to learning.”[/rquote]

Another implication of this research is that computers should only imitate people when there is good reason to expect strong benefits from the added social element, professor of psychology and co-author of the study Dan Levin explained.

“In paradigms where you are trying to use technology for students to learn things, sometimes the technology may be better if it doesn’t involve human-like agents,” he said.

Herberg and Levin worked with associate professor of psychology and Vanderbilt Kennedy Center member Megan Saylor to complete this study, which recently published in the Journal of Experimental Social Psychology and was supported by a grant from the National Science Foundation.

The authors will continue this research by testing different tasks to explore the types of activities that suffer from a social audience and conversely tasks that benefit from a social audience.