A brief history of Intelligent Tutoring Systems and their necessary educational functions which have already been realized and not yet been realized are presented separately, then problems to be solved within the framework of ITS and problems that transcend the framework of ITS are discussed. Lastly, it is indicated that the problems will be solved by an amalgamation of an open-end system like a micro world and a discovery system with direct manipulation into ITS and that the central problem to realize the amalgamation is a discovery learning by a machine itself.

This paper proposes an effective comprehension support by an assortment of multiple comprehension support methods. Each comprehension support method has strong and weak points. However, comprehension support methods can play complementary roles. These methods can be combined so that these methods may play complementary roles and more effective comprehension support can be achieved by this combination. We constructed a compound comprehension support system by assorting multiple comprehension support methods. The evaluation of this system is given in this paper.

A pattern is a string of constant symbols and variables. The language defined by a pattern p is the set of constant strings obtained from p by substituting nonempty constant strings for variables in p. In this paper we are concerning with polynomial time inference from positive data of the class of unions of a bounded number of pattern languages. We introduce a syntactic notion of minimal multiple generalizations (mmg for short) to study the inferability of classes of unions. If a pattern p is obtained from another pattern q by substituting nonempty patterns for variables in q, q is said to be more general than p. A set of patterns defines a union of their languages. A set Q of patterns is said to be more general than a set P of patterns if for any pattern p in P there exists a more general pattern q in Q than p. Clearly more general set of patterns defines larger unions. A k-minimal multiple generalization (k-mmg) of a set S of strings is a minimally general set of at most k patterns that defines a union containing S. The syntactic notion of minimality enables us to efficiently compute a candidate for a semantically minimal concept. We present a general methodology for designing an efficient algorithm to find a k-mmg. Under some conditions an mmg can be used as an appropriate hypothesis for inductive inference from positive data. As results several classes of unions of pattern languages are shown to be polynomial time inferable from positive data.

A tree pattern is a structured pattern known as a term in formal logic, and a tree pattern language is the set of trees which are the ground instances of a tree pattern. In this paper, we deal with the class of tree languages whose language is defined as a union of at most k tree pattern languages, where k is an arbitrary fixed positive number. In particular, We present a polynomial time algorithm that, given a finite set of trees, to find a set of tree patterns that defines a minimal union of at most k tree pattern languages containing the given set. The algorithm can be considered as a natural extension of Plotkin's anti-unification algorithm, which finds a minimal single tree pattern language containing the given set. By using the algorithm, we can realize a consistent and conservative polynomial time inference machine that identifies the class of unions of k tree pattern languages in the limit from positive data for every k > 0.

This paper presents a direct manipulation environment for learning intelligent tutoring systems. The environment provides views of internal states of a intelligent tutoring system and facilities for manipulating the states in order to study structures and functions of constituents of ITS. The target users of the learning environment are school teachers, students of AI courses and ITS researchers. School teachers get deep understanding about ITS by observing not only the external behaviors but also the inner states. AI students learn about design issues of ITS, roles of knowledge and characteristics of knowledge processing. ITS researcher can use the environment for a test bed of analyzing different methods of ITS constituents, because it provides facilities to replace its constituents.

This paper proposes a support system for analysis of student’s motivation in group learning. A teacher may select a learning form, where students talk mutually about exercises with a chat system. In order for students to learn it more effectively, the teacher should analyze the learning conditions (e.g. the learning goal, the exercise, the group member, and so on), evaluate the learning effect, and improve the learning conditions. However, it is a very time-consuming and difficult job. We consider that student’s emotions toward the learning conditions have some relations to his/her motivation. The system visualizes students’ emotions under various learning conditions. According to these visualizations, the teacher can analyze and improve the learning conditions.

Reading a program, learners must understand syntax, semantics, and usage, which are characteristics of each individual programming language. Moreover, they must acquire typical writing art through understanding various good programs. In order to obtain good programming methods, learners must also know the execution flow of programs, which is difficult to explain statically. In this paper, as a learning environment for Prolog programs, an animation tree is proposed, which allows learners to interactively manipulate program execution back and forth in a stepwise manner. The effect of the animation tree in the problem solving process and the contents of finding through comparison of two programs with the same structure are investigated.

This paper presents a method to construct both an intelligent tutoring system and its authoring system for Japanese on reading mathematical expressions in English. Because mathematical expressions can be described easily in DCG, the remaining problems are mainly identification of error origins in a learner's answer and judgment of his/her state of understanding. The paper proposes that it is possible to construct a practically solid intelligent tutoring system if a domain and suitable technologies of artificial intelligence are selected carefully, though it is often held that a tutoring system using technologies of artificial intelligence is of no practical use. © 2000 Scripta Technica, Syst Comp Jpn, 31(10): 56–64, 2000

This paper presents methods for modeling and assisting students' understanding of causal relationships between physical quantities based on comparative reasoning. Our tutoring system discusses causalities in an object system with the student by choosing dialogue strategies according to the student's understanding state. This state is represented by the causal network and table. Dialogue strategies are characterized by the following four elements of dialogue: control of topic, progress of dialogue, granularity of dialogue, and style of utterance. This paper also presents a method of deepening the student's understanding of the object system by comparing it with a similar system, which is generated by partitioning or simplifying the original system. The partitioned system has a causality different from that of the original system, and the simplified system has a causality similar to that of the original system. © 1998 Scripta Technica, Syst Comp Jpn, 29(6): 1–15, 1998

This paper proposes an intelligent support method for solving arithmetic word problems and a method to build a learning environment on a network. Our system solves arithmetic word problems written in Japanese in the same way as humans do, by syntax analysis, mathematical semantic analysis, formulation, and calculation. Using data generated during these processes, computers support learners by adapting to each individual learner's comprehension level through Q & A with animation. The fundamental ideas and a few examples of solving word problems, supporting learners by using animation, and building the environment on a network are described. © 1998 Scripta Technica, Syst Comp Jpn, 28(10): 30–39, 1997

Qualitative reasoning is an effective method for intelligent tutoring systems. It can provides causal explanation of behavior that cannot be achieved by numerical simulation. The causal explanation is obtained based on a set of differential equations. If a student doesn't understand the explanation, we should explain the reason why the equations hold. Qualitative reasoning cannot answer this question because the equations are predefined for qualitative reasoning mechanisms. This is a problem of system modeling. In order to explain why equations or relations of system parameters hold, a tutoring system is required ability to derive the relations from a system structure.This paper proposes a method of deriving relationship among forces and elements of a given kinetic systems. The relations are represented by a network called causal relation model, which shows causes of force occurrences. Because the derivation method is based on human intuition, it provides a natural explanation for a student. A tutoring system based on the causal relation model is also presented in this paper In order to brush up student's ability to model kinetic systems, the tutor asks the student to illustrate forces that act in a given System. Student's interpretation of the system is modeled by a network called student's kinetic model, which has the same structure as the causal relation model. The tutor diagnoses the student by comparing the causal relation model and the student's kinetic model to give adaptable advice to the student.

This paper presents methods for modeling and assisting students’ understanding about causalities between physical quantities based on comparative reasoning. Our tutoring system discusses causalities in an object system with the student by choosing dialogue strategies according to student’s understanding states. The student’s understanding state is represented by the causal network and table. Dialogue strategies are characterized by the following four elements of dialogue; selection of topic, progress of dialogue, granularity of dialogue, and style of utterance. This paper also presents a method of deepening student’s understanding of the object system by comparing it with a similar system which is generated by partitioning or simplifying the original system. The partitioned system has a different causality from the original system, and the simplified system has a similar causality to the original system.

This paper proposes a method of learning assistance based on a set of student models in group learning environments. Our system makes student models in the same way as intelligent tutoring systems do and gets pseudo students to join in learning to assist learners. By referring to student models, pseudo students induce learners to revise incorrect part and to repair missing part of other learners’ knowledge. Pseudo students also assist learners to join in discussion. We also propose the minimum number of the necessary pseudo students to cover the proposed assistance. Outline of Group Learning Environment Figure 1 shows the outline of our group learning environment. It has aspects of both groupwares and intelligent tutoring systems. It provides learners the following facilities as a groupware system. 1 Private Windows: Learners should get their ideas into shape before they discuss about a given problem. Private windows are work spaces, in which learners visualize their ideas. Because a private window is a private space for each learner, learners aren’t allowed to access other learners’ private windows. 2 Shared Window and Discussion Window: The shared window and the discussion window are used for collaborative learning. The shared window is a work space shared by all learners in a group. Drawings and statements in the shared window are recorded with names of learners who enter the information to clarify responsibility. Learners aren’t allowed to change the information without permission of drawers. The discussion window is used for discussion among learners. It records a sequence of statements with speakers’ names. Our group learning environment has domain knowledge and student models, and assists learners as intelligent tutoring systems do. However the method of assistance is different. A learning goal of the environment is that all learners in a group acquire knowledge or skills to solve given problems by discussion. In order to achieve the goal without disturbing a collaborative learning style, we introduce pseudo students to assist learners. Pseudo students join in discussion through the shared window and the discussion window. domain knowledge bug knowledge

Abstract In this chapter, we present a polynomial time algorithm, called a k-minimal multiple generalization (k-mmg) algorithm, where k ≥ 1, and its application to inductive learning problems. The algorithm is a natural extension of the least general generalization algorithm developed by Plotkin and Reynolds. Given a finite set of ground first-order terms, the k-mmg algorithm generalizes the examples by at most k first-order terms, while Plotkin’s algorithm does so by a single first-order term. We apply the k-mmg algorithm to several learning problems in inductive logic programming, and knowledge discovery in databases.

The k-minimal multiple generalization (k-mmg) is a natural extension of the least generalization (lg) given by Plotkin in 1970. The k-ming generalizes given first order terms by at most k-terms, while the 1g does by a single term. Thus, k-mmg gives a more precise approximation of a given set of examples. In this paper, we extend the algorithm for as more abstract class of objects by abstracting a generalization structure of first-order terms. We present a general design of a polynomial time k-mmg algorithm for the wider classes of objects, and prove the correctness. Using the algorithm, we prove the polynomial time inferability from positive data of unions of at most k languages in a subclass of pattern languages. One class is the class of one-variable pattern languages, and another is the class of regular pattern languages with a bounded number of variables. We also discuss the use of refinement operator and NC-learnability from positive data.

In this paper, we consider the polynomial time inferability from positive data for unions of two tree pattern languages. A tree pattern is a structured pattern known as a term in logic programming, and a tree pattern language is the set of all ground instances of a tree pattern. We present a polynomial time algorithm to find a minimal union of two tree pattern languages containing given examples. Our algorithm can be considered as a natural extension of Plotkin's least generalization algorithm, which finds a minimal single tree pattern language. By using this algorithm, we can realize a consistent and conservative polynomial time inference machine that identifies unions of two tree pattern languages from positive data in the limit.

This paper describes an empirical study on in-house teacher training for the use of computers in primary school lessons. In the training a common format is given to a trainee in order to construct a teaching program and to make the program known widely in the school, so that other teachers may join in discussing the program regardless of their experiences concerning computers. Discussions on constructing the teaching program by using the common format are given both from the epistemological viewpoint and from the teacher's standpoint.