AutoTutor Improves Deep Learning of Computer Literacy: Is It the Dialog or the Talking Head?

摘要

AutoTutor is a tutoring system that helps students construct answers to deep-reasoning questions by holding a conversation in natural language. AutoTutor delivers its dialog moves with an animated conversational agent whereas students type in their answers. We conducted an experiment on 81 college students who learned topics on computer literacy (hardware, operating systems, Internet) with AutoTutor or control conditions, and were assessed on learning gains. We designed the experiment to assess the impact of learning condition (AutoTutor, read-text control, versus nothing) and the medium of presenting AutoTutor's dialog moves (print only, speech only, talking head, versus talking head + print). All versions of AutoTutor improved performance in assessments of deep learning, but not shallow learning. Effects of the medium were subtler, which suggests that the message (the dialog moves of AutoTutor) is more important than the medium.