In our previous work [9, 10], we highlighted the importance of incorporating learners' pedagogical features in making paper recommendations and proposed the pedagogical-oriented paper recommender. In this paper, we report our studies in designing and evaluating a six-dimensional paper recommender. Experimental results from a human subject study of learner preferences suggest that in the e-learning domain where there are not enough papers as well as learners to start up the recommender system (RS), it is imperative to inject other factors such as the overall popularity of each paper, learner knowledge background, etc., although these factors are less important for making recommendations on movies, books, CDs.
We are developing a conversational agent called VIBRANT to provide adaptive support for brainstorming in pairs in a scientific inquiry context. Our previous experimental study indicated that although adaptive support was effective for learning, it did not mitigate the negative influence of social interaction on idea generation. In this paper, we present a process analysis that makes the effect of the adaptive support on the idea generation process more apparent and suggests directions for future work.
We present a publicly available tool called TagHelper that can be used to support the analysis of conversational data using automatic text classification technology. The contribution of this paper is to explore the limitations of the current simple approach to text processing employed by TagHelper tools with respect to identifying context-sensitive categories of conversational behavior. TagHelper can be downloaded from http://www.cs.cmu.edu/˝cprose/TagHelper.html.
Amali Weerasinghe, Antonija Mitrovic, Brent Martin
665 - 667
We present a project with the goal of developing a general model for supporting explanations, which could be used in both well- and ill-defined instructional tasks. We have previously studied how human tutors provided additional support to students learning with an existing intelligent tutoring system. Analysis of the interactions by human tutors indicates that they have helped the students to improve their understanding of database design. This paper presents the explanation model developed based on these findings.
Atomic Dynamic Bayesian Networks (ADBNs) combine several valuable features in student models: students' performance history, prerequisite relationships, concept to solution step relationships, and student real time responsiveness. Recent work addresses some of these features but has not combined them. Such a combination is needed in an ITS that helps students learn, step by step, in a complex domain, such as object-oriented design. We evaluated ADBN-based student models 49 human students, investigating their behavior for different types of students and different slip and guess values. Holding slip and guess to equal, small values, ADBNs are able to produce accurate diagnostic rates for knowledge states over each student's learning history.
Rob Weitz, Neil Heffernan, Viswanathan Kodaganallur, David Rosenthal
671 - 673
The little previous research comparing student errors across schools indicates that student “bugs” do not transfer – that is, the distribution of students' systematic errors in one school does not significantly match those in other schools. The issue has practical implications as cognitive (or “model-tracing”) tutors rely on the modeling of student errors in order to provide targeted remediation. In this study we examine the responses of students at three schools to a middle-school mathematics problem. We find the same error is the most common error across all schools, and this single error accounts for some half of all incorrect responses at each school. The top five errors are similar across schools and account for some 2/3 of errors at each school. We conclude that in this example, there appears to be considerable overlap of student errors across schools.
In this paper, we present an ontology-based approach, The Knowledge Puzzle approach, that aims to exploit principles from the AIED and Intelligent Tutoring Systems fields to produce e-Learning resources more tailored to learner's needs. We present a semi-automatic process to annotate learning material from different points of view: domain, structural and pedagogical. We then use this knowledge to generate dynamically learning knowledge objects based on instructional theories.