Investigating the roles of feedback categories within an intelligent tutoring system

Tutorial dialogue is an important component of one-to-one tutoring environments. AutoTutor, an intelligent tutoring system, implements tutorial dialogue in natural language. One of the core components of tutorial dialog is feedback, which carries the primary burden of informing students of their performance. This dissertation addressed the effectiveness of two types of feedback while college students interacted with AutoTutor: content feedback and progress feedback. Content feedback is a qualitative assessment of the domain content being covered in a tutoring session (i.e., accuracy of knowledge and completeness of answers). Progress feedback is a quantitative assessment of the student's advancement through the material being covered (i.e., how far the student has come, how much farther they have to go, points scored). These two types of feedback were manipulated at local and global levels within the AutoTutor system, on the topic of Newtonian physics. Students interacted with different versions of AutoTutor that varied the type of feedback (content versus progress feedback) at both local and global levels. The impact of these independent variables on learning outcomes was explored, along with indices of motivational and attitudinal measures. The results showed that the presence vs. absence of content feedback had the largest effect on both learning and motivational measures, whereas the impact of progress feedback was small or nonsignificant.