Abstract. In designing and building tutorial dialogue systems it is important not only to understand the tactics employed by human tutors but also to understand how tutors decide when to use various tactics. We argue that these decisions are based not only on student problem-solving steps and the content of student utterances, but also on the meta-communicative information conveyed through spoken utterances (e.g., pauses, disfluencies, intonation). Since this information is often infrequent or unavailable in typed input, tutorial dialogue systems with speech interfaces have the potential to be more effective than those without. This paper gives an overview of the Spoken Conversational Tutor (SCoT) that we have built and describes how we are beginning to make use of spoken language information in SCoT. Specifically, we describe a study aimed at using meta-communicative information to gauge student uncertainty and respond accordingly. In this study, we identify linguistic devices used by human tutors when responding to utterances containing signals of uncertainty, integrate these response strategies into two versions of SCoT, and evaluate their relative effectiveness. Our main hypothesis—that tutors are more effective if they use these linguistic devices in response to student uncertainty—was not confirmed, but our secondary hypothesis—that tutors using these linguistic devices are more effective than tutors that do not use them—was supported by the results.

In contrast to current intelligent systems, which must be laboriously programmed for each task they are meant to perform, instructable agents can be taught new tasks and associated knowledge. This thesis presents a general theory of learning from tutorial instruction and its use to produce an instructable agent. Tutorial instruction is a particularly powerful form of instruction, because it allows the instructor to communicate whatever kind of knowledge a student needs at whatever point it is needed. To exploit this broad flexibility, however, a tutorable agent must support a full range of interaction with its instructor to learn a full range of knowledge. Thus, unlike most machine learning tasks, which target deep learning of a single kind of knowledge from a single kind of input, tutorability requires a breadth of learning from a broad range of instructional interactions. The theory of learning from tutorial...

Abstract: KERMIT is an intelligent tutoring system that teaches conceptual database design using the Entity-Relationship data model. Database design is an open-ended task: although there is an outcome defined in abstract terms, there is no procedure to use to find that outcome. So far, constraint based modelling has been used in a tutor that teaches a database language (SQL-Tutor) and a system that teaches punctuation and capitalisation rules (CAPIT). Both systems have proved to be extremely effective in evaluations performed in real classrooms. In this paper, we present experiences in using CBM in an open-ended domain. We describe system’s architecture and functionality. KERMIT has also been evaluated in the context of genuine teaching activities. We present the results of an evaluation study with students taking a database course, which show that KERMIT is an effective system. The students enjoyed the system’s adaptability and found it a valuable asset to their learning. 1.

Click here to download paper in PDF format This article presents a vision of the near future in which computer interaction is characterized by natural face-toface conversations with lifelike characters that speak, emote and gesture. These animated agents will converse with people much like people converse effectively with assistants in a variety of focused applications. Despite the research advances required to realize this vision, and the lack of strong experimental evidence that animated agents improve human computer interaction, we argue that initial prototypes of perceptive animated interfaces can be developed today, and that the resulting systems will provide more effective and engaging communication experiences than existing systems. In support of this hypothesis, we first describe initial experiments using an animated character to teach speech and language skills to children with hearing problems, and classroom subject and social skills to children with autistic spectrum disorder. We then show how existing dialogue system architectures can be transformed into perceptive animated interfaces by integrating computer vision and animation capabilities. We conclude by describing the Colorado Literacy Tutor, a computer-based literacy program that provides an ideal test bed for research and development of perceptive animated interfaces, and consider next steps required to realize the vision.

Abstract: E-learning is becoming more and more popular with the widespread use of computers and the Internet in educational institutions. Current e-learning courses are nearly always developed using course management systems (CMS), such as WebCT or Blackboard. Although CMS tools provide support for some administrative tasks and enable instructors to provide online instructional material, they offer no deep support for learning: students have access to on-line material, simple multi-choice quizzes and chat tools, but there is no ability to track student’s progress and adapt the learning material and instructional session to the individual student. In this paper we present our experiences with three Web-based intelligent tutoring systems in the area of databases. SQL-Tutor teaches the SQL query language, NORMIT is a data normalization tutor, and KERMIT teaches conceptual database modelling using the Entity-Relationship data model. All three tutors in DB-suite have been used and evaluated in the context of genuine teaching activities. We present the most important features of these systems, as well as evaluation results. The DB-suite tutors have proved to be very effective in supporting deep learning, and are well liked by students. Keywords: Web-based intelligent tutoring systems, computational intelligence in learning and authoring tools, student modelling in Web-based education, evaluation of intelligent Web-based teaching and learning systems

This article presents five principles of learning, derived from cognitive theory and supported by empirical results in cognitive psychology. To bridge the gap between theory and practice, each of these principles is transformed into a practical guideline and exemplified in a real teaching context. It is argued that this approach of putting cognitive theory into practice can offer several benefits to statistics education: a means for explaining and understanding why reform efforts work; a set of guidelines that can help instructors make well-informed design decisions when implementing these reforms; and a framework for generating new and effective instructional innovations

What are good ways of using natural language dialog in intelligent tutoring systems? A role with high potential payoff is to support the meta-cognitive process of selfexplanation. In previous experiments involving the PACT Geometry Tutor, we found that students learn with greater understanding, when they are required to explain their solutions steps "by reference", that is, by naming the rule that was used. However, the tutor may be even more effective if students explain their solution steps in their own words and if the tutor helps them, through dialog, to improve their explanations. An exploratory experiment with a tutor version that did not do any natural language processing, strongly suggested the need for natural language dialog. Without feedback from the tutor, students provided few free-form explanations in response to the tutor's prompts. Of the explanations that they did provide, only a small portion were correct and complete. During the experiment, we also i...

Abstract: One of the promises of ITSs and ILEs is that they will teach and assist learning in an intelligent manner. Historically this has tended to mean concentrating on the interface, on the representation of the domain and on the representation of the student’s knowledge. So systems have attempted to provide students with reifications both of what is to be learned and of the learning process, as well as optimally sequencing and adjusting activities, problems and feedback to best help them learn that domain. We now have embodied (and disembodied) teaching agents and computer-based peers, and the field demonstrates a much greater interest in metacognition and in collaborative activities and tools to support that collaboration. Nevertheless the issue of the teaching competence of ITSs and ILEs is still important, as well as the more specific question as to whether systems can and should mimic human teachers. Indeed increasing interest in embodied agents has thrown the spotlight back on how such agents should behave with respect to learners. In the mid 1980s Ohlsson and others offered critiques of ITSs and ILEs in terms of the limited range and adaptability of their teaching actions as compared to the wealth of tactics and strategies employed by human expert teachers. So are we in any better position in modelling

Abstract. In this paper, we explore the problem of selecting appropriate interventions for students based on an analysis of their interactions with a tutoring system. In the context of the WHY2 conceptual physics tutoring system, we describe CarmelTC, a hybrid symbolic/statistical approach for analysing conceptual physics explanations in order to determine which Knowledge Construction Dialogues (KCDs) students need for the purpose of encouraging them to include important points that are missing. We briefly describe our tutoring approach. We then present a model that demonstrates a general problem with selecting interventions based on an analysis of student performance in circumstances where there is uncertainty with the interpretation, such as with speech or text based natural language input, complex and error prone mathematical or other formal language input, graphical input (i.e., diagrams, etc.), or gestures. In particular, when student performance completeness is high, intervention selection accuracy is more sensitive to analysis accuracy, and increasingly so as performance completeness increases. In light of this model, we have evaluated our CarmelTC approach and have demonstrated that it performs favourably in comparison with the widely used LSA approach, a Naive Bayes approach, and finally a purely symbolic approach.

Studies aimed at understanding what makes human tutoring effective have noted that the type of indirect guidance that characterizes human tutorial dialogue is a key factor. In this paper, we describe an approach that brings together sociolingusitic research on the basis of linguistic choice with natural language generation technology to systematically produce tutorial feedback appropriate to the given situation.