An Investigation of Accuracy of Metacognitive Judgments during Learning with an Intelligent Multi-Agent Hypermedia Environment

Successful learning with advanced learning technologies is based on the premise that students adaptively regulate their cognitive and metacognitive processes. However, research suggests that students are rather dysregulated in their learning. One major source of dysregulation is based on inaccurate metacognitive judgments made during learning. This study investigated learners’ accuracy and confidence in metacognitive judgments made in the context of learning about the human circulatory system with MetaTutor, a multiagent intelligent hypermedia learning system. 83 college students took part in this study, and their interactions within MetaTutor in the two-hour learning session provided data for this study. In general, the results revealed that learners were overconfident to differing degrees in ratings of their judgments of learning (JOLs) and feelings of knowing (FOKs). It was also found that receiving timely prompts and adaptive feedback from the artificial agent in MetaTutor improved the accuracy of metacognitive judgments. Learners in the Prompt and Feedback condition (PF) were overconfident to a lesser degree than those in other conditions (Prompt Only [PO] and Control). Finally, one-way ANOVA and Tukey post-hoc results indicated that learners who received prompts and feedback attained significantly (p < .05) better learning efficiency scores than learners in Control and Prompt Only conditions.