Middle School Students' Ability to Detect Lies When Interacting with an Educational AI Robot
DOI:
https://doi.org/10.58459/icce.2024.4897Abstract
Educational robots are becoming increasingly incorporated into classrooms to teach many subjects including language, science, and also substitute teachers. In recent years, ChatGPT has become easy to use for education but there are problems if students take it literally and believe it. We investigate the effectiveness of different deception techniques through robot teaching. We conduct an experiment in a Japanese Junior High School with 14 students where we investigate the learning and deception effectiveness, and believability using the social robot "Furhat". Moreover, we vary the social agency of the robot by using two different faces, a human (social identity theory) and an anime face (Japanese anime culture). A robot with an anime face achieved a significantly higher learning effectiveness compared to a robot with a human face. However, the human robot face was found to be excellent at deceiving when the paltering deception technique was utilized.