Skip to main content

Über dieses Buch

In October of 1992 an assembly of researchers in simulation and computer models for instruction convened in Bonas, France, to learn from one another in a non-automated environment. The event was the Advanced Research Workshop entitled The Use of Computer Models for Explication, Analysis, and Experiential Learning. Sponsored by the Scientific Affairs Division of NATO, this workshop brought together 29 leading experts in the field loosely described as instruction and learning in simulation environments. The three-day workshop was organized in a manner to maximize exchange of knowledge, of beliefs, and of issues. The participants came from six countries with experiences to share, with opinions to voice, and with questions to explore. Starting some weeks prior to the workshop, the exchange included presentation of the scientific papers, discussions immediately following each presentation, and informal discussions outside the scheduled meeting times. Naturally, the character and content of the workshop was determined by the backgrounds and interests of the participants. One objective in drawing together these particular specialists was to achieve a congress with coherent diversity, i.e., we sought individuals who could view an emerging area from different perspectives yet had produced work of interest to many. Major topic areas included theories of instruction being developed or tested, use of multiple domain models to enhance understanding, experiential learning environments, modelling diagnostic environments, tools for authoring complex models, and case studies from industry.



Building Domain Models to Support Instruction

An account is given of domain modelling in the ITS IE project. The project was concerned with simulation based training in industrial environments. In such situations, models are required not only of the technical systems being manipulated, but also of the operational expertise required to carry out tasks. Now the way something should be modelled is dependent on the context in which the model will be used — this is true whatever the nature of the model: mental, numerical, or otherwise. Consequently, multiple models are required along with a framework which allows their effective use. A number of modelling tools are described and related to a classification of models. Two demonstration systems are described for illustrative purposes.
K. Horne, J. Kwaan, D. Scott, W. Scott

Knowledge Structures for a Computer-Based Training Aid for Troubleshooting a Complex System

Turbinia-Vyasa is a computer-based instructional system that trains operators to troubleshoot marine power plants. It is implemented on Apple Macintosh II computers. The simulator,Turbinia is based on a hierarchical representation of subsystems, components, and primitives together with necessary physical and logical linkages among them.Vyasa is the computer-based tutor that teaches the troubleshooting task usingTurbinia. The simulator, an interactive, direct manipulation interface, and the tutor (with its expert, student, and instructional modules) comprise the architecture for the instructional system. Details of knowledge organization that supports the functions of the three major elements of the tutoring system are discussed.
Vijay Vasandani, T. Govindaraj

Authoring Interactive Graphical Models for Instruction

Although many intelligent tutoring systems have made use of interactive graphical models of topic matter, most have been developed using conventional programming tools, sometimes augmented by special interface editors. This approach permits a high degree of flexibility and control, but its successful application requires a great deal of technical skill and development time. It is therefore feasible for only a small subset of the training and instruction contexts in which an interactive simulation would be of benefit. Recently, several authoring environments have been developed that support the creation of interactive graphical models by direct manipulation. IMTS, RAPIDS n, and RIDES are three authoring tools for the production of intelligent tutoring systems centered around interactive graphical models. The experience of developing these three authoring environments and applying them to produce a number of tutors has helped to clarify the desirable features for authoring systems for interactive graphical models.
Allen Munro

Visualization in Educational Computer Modeling

This paper describes recent research on the design and use of visual models and simulations in science education. The focus of the work has been the development and use of a variety of model development tools and applications to demonstrate the power and instructional benefits of visualization methods. Modeling applications were made in biology, chemistry, mathematics, and physics. The models incorporate graphic interfaces designed to enable easy interaction by students without the need for a deep understanding of computers. Work is described with two distinctly different kinds of visual modeling systems: a generic environment for science simulations (Explorer Science), and a universal visual programming language (Function Machines). In classroom trials of these programs, high school students were successfully introduced to the use of model-based inquiry in problem investigations.
Wallace Feurzeig

Diagnosis of Intentions and Interactive Support of Planning in a Functional, Visual Programming Language

Based on a theoretical framework of problem solving and knowledge acquisition, criteria for intelligent knowledge communication systems and help design are described. The ABSYNT Problem Solving Monitor for the acquisition of basic functional programming concepts in a visual language is designed according to these criteria. It incorporateshypotheses testing of solution proposals, and a learner model is designed to supplyuser-adapted help.
Claus Möbus, Heinz-Jürgen Thole, Olaf Schröder

The Flexible Use of Multiple Mental Domain Representations

In this paper we present a framework which describes learning in physics domains as a succession of multiple levels of mental domain representations ordered along the dimension qualitative/quantitative. The emphasis is on the mental representation of functional relationships between physics variables. We exemplify our approach in the domain of elastic impacts, a subtopic of classical mechanics. In order to analyze the feasability of our framework an empirical study as well as three cognitive models are presented. MULEDS, a computerized multi-level diagnosis system, is capable of diagnosing correct, incorrect, and incomplete elements of students’ knowledge. It incorporates mechanisms for tailored testing as well as for active adaptation of instruction to diagnosed misconceptions. KAGE is a cognitive model of how students acquire knowledge about functional relationships between physics variables. It accounts for the question which knowledge states have to be expected when specified analysis-based learning mechanisms are applied to given instructional information. The Sepia model shows which and how qualitative physics knowledge facilitates quantitative physics problem solving. Sepia is also discussed with respect to its potential for supporting the design of physics instructions.
Klaus Opwis

Computer-Based Support for Analogical Problem Solving and Learning

Despite the important role of specific examples for learning and problem solving, little support is given in computer-based learning and teaching environments to help students organize information about examples and problem solving episodes in a way that may enhance generalization and transfer. The main thesis of this chapter is that learning from examples can be improved — in particular, the transfer problem can be reduced — if students are supported in managing specific knowledge as it is acquired from worked-out examples and students’ own problem solving experiences. We sketch out the blueprint for a “Memory Assistant”, a computer program that helps students in the analogical problem solving process by reducing memory load, by providing semi-automatic remindings, and by pointing out differences and similarities between a new problem and the analogical source. After having identified some of the essential cognitive demands learning from examples imposes on students, we describe the interface features and functional requirements for a computerized tool that can help them to cope with these demands. It is suggested to use techniques developed in case- based reasoning systems to handle issues of case retrieval and modification, and combine them with a hypertext-based user interface, thus allowing for smooth case acquisition and retrieval. We illustrate these ideas with examples from the domain of mechanics problem solving.
Peter Reimann, Sieghard Beller

Modeling Actors in a Resource Dilemma: A Computerized Social Learning Environment

kis is a computerized knowledge-based model of how people act and interact in ecological-social conflicts. The acronym stands for knowledge and intentions in social dilemmas. The model is embedded into a computerized version of such a conflict. It reflects the interplay between motives and knowledge for the generation of actions. Ecological and social knowledge are modeled. Action knowledge is represented in form of action schemata. Intention formation and realization are simulated as central processes of motivated action. Action knowledge is seen to be built up from learning by doing, mental simulation and observing others.
The curricular function of the model is twofold. It sets up a framework to formulate learning goals in the domain of ecological-social dilemmas and to develop instructional measures to achieve them. The second aspect addresses a rather unusual function of cognitive modeling. Social learning is supported by equipping computerized learning environments of the microworld type with artificial actors. They are run by the model and behave according to instructional demands.
The kis model has been developed on the basis of previous empirical studies and it is fully implemented. But systematic tests of its validity and instructional practicability have not been carried out yet.
Andreas M. Ernst, Hans Spada

Basic Research on the Pedagogy of Automated Instruction

In this chapter we argue for the importance of basic research on the pedagogy of automated instruction, and outline a relevant research project currently underway at the Armstrong Laboratory. The goal of the project is to delineate general principles as well as specific guidelines for developers of automated instruction. The project seeks to contribute to knowledge of automated instruction in the following four ways. First, we are supporting the development and use of a task-decomposition taxonomy to promote and support synthesis of results across studies. Second, we are developing a set of standard criterion tasks that will allow benchmarked comparisons of instructional approaches. Third, we are developing automated instructional systems, including Intelligent Tutoring Systems (ITS), based on formal theories of knowledge and skill acquisition. Fourth, we are rigorously evaluating these instructional systems in a controlled laboratory setting. In this chapter, we briefly describe this approach to conducting basic research on automated instruction, and then present three examples of research findings from our laboratory: (a) An alternative approach to using computer resources in training environments that, for certain classes of tasks, quadruples the number of trainees that may be trained with fixed training resources; (b) An inexpensive instructional intervention that significantly reduces or eliminates the post-training gender gap in performance of a highly spatial, complex dynamic control task; and (c) A simple, inexpensive intervention that reduces post-training error rates by 50% and performance latency by 33% in a procedural console operation task. We conclude by briefly discussing the implications of these findings for developers of automated instruction.
J. Wesley Regian, Valerie J. Shute

Modeling Practice, Performance, and Learning

This chapter presents the results from a study examining the relationship between practice, performance, and learning. We compared two versions of an intelligent tutoring system differing only in the number of problems that needed to be solved per problem set (Abbreviated = 3 problems, Extended = 12 problems). Our hypotheses were that Abbreviated subjects, in comparison to Extended subjects, would: (a) take less time to complete the tutor because they had fewer problems to solve, (b) perform worse on the posttest measures (accuracy and latency), and (c) demonstrate poorer transfer of knowledge and skills across tutor problems given fewer practice opportunities. We found that, while Abbreviated subjects did take significantly less time to complete the tutor than Extended subjects, both groups performedequally across all outcome measures. Componential skill analyses enabled us to track the course of skill acquisition during practice, and predict the degree of skill transfer afterward. We conclude with suggestions for the development of efficient automated instruction.
Valerie J. Shute, J. Wesley Regian, Lisa A. Gawlick-Grendell

Teaching and Learning Diagnostic Skills in a Simulation Environment

Simulation environments can be employed in a variety of ways to enhance a learner’s diagnostic skills and knowledge. Novices can receive instruction in critical subskills including device manipulation, symptom detection, and symptom interpretation, and they can practice those subskills with close automated support. Intermediate level learners can practice applying those subskills in a realistic diagnostic environment in which the difficulty of the learning environment is individualized through the use of problem selection and learner support functions. Advanced learners can utilize the simulation to meet personal learning objectives by exploring a wide range of normal and abnormal conditions and by controlling the introduction of simulated faults.
A host of alternatives face the instructional development in producing a training system to support such learning requirements. This chapter attempts to delineate some of the most critical alternatives for creating the device simulation and for managing the learning environment.
Douglas M. Towne

Environment Design and Teaching Intervention

This paper presents the first stages in devising a method for extending the principled design basis of Guided Discovery Tutoring systems. In particular, it clarifies the way in which ideas from didactique can be related to Guided Discovery architectures. This leads to an illustration of the mechanisms by which a domain model and a didactic model of the domain can be used in conjunction to derive aspects of the interface of simulation systems, and certain teaching interventions required by a linked tutor.
Mark T. Elsom-Cook

A Model to Design Computer Exploratory Software for Science and Mathematics

The aims of this paper are: 1) to characterize computer exploratory software; 2) to identify the roots of this kind of software; 3) to present a model to design computer exploratory environments for science and mathematics; 4) to discuss some of the basic issues of the model; and 5) to analyse some programs developed in the framework of the model. The model is based on findings in learning and in recent developments of computer graphic environments. It as-sumes that: 1) learning is a process of enculturation, a process of becoming familiar with ideas and representations; 2) exploratory software should be integrated with other resources; 3) exploratory software should allow direct manipulation of concrete-abstract objects and the exploration of multiple representations of a phenomenon.
Vitor Duarte Teodoro

Exploring a Domain with a Computer Simulation: Traversing Variable and Relation Space with the Help of a Hypothesis Scratchpad

Computer simulations provide a challenging opportunity to create learning environments in which learners are free to explore the domain and discover the domain properties themselves. Contemporary theories of learning state that knowledge acquired and constructed within such an exploratory environment will be rooted more deeply within the learner’s knowledge structures. On the other hand, it has also become clear that offering a simulation to the learner without offering any additional support may result in learners getting “lost” in the simulation environment, not learning very much. Therefore, additional support is deemed necessary for simulation learning environments.
The current chapter describes a software instrument (a “hypothesis scratchpad”) that can be offered to the learner in order to support the process of hypothesis formation. The design of this instrument is based on an analysis of hypothesis space, one of the two search spaces from the theory of Klahr and Dunbar [14], who describe discovery as search in two related spaces, a hypothesis space and an experiment space. A study was performed in which the hypothesis scratchpad was used to influence the learners search processes in variable space and relation space, the two subspaces of hypothesis space. The moment of stating hypotheses (before doing experiments or during a series of experiments) and the guidance for variable space search (starting with instances or directly going to general variables) was varied in this study. It appeared that the learners who were prompted to state hypotheses before doing experiments stated more hypotheses while doing experiments with the simulation. Learners who were searching varibale space only at the level of general variables, chose relations at a more precise level than learners searching variable space, starting with instances before going to the general concepts.
Wouter van Joolingena, Ton de Jong

Supporting Exploratory Learning by Offering Structured Overviews of Hypotheses

Exploratory learning with computer simulations is an approach that fits well within the current emphasis on viewing the learner as an active, constructive person. In previous studies we concluded that a valid performance of exploratory learning processes was a bottleneck and especially the process of hypothesis generation posed difficulties to learners. The major objective of the present study was to evaluate the effect of supporting hypothesis generation by offering structured overviews of predefined hypotheses. Subjects were 88 Mechanical Engineering students working in pairs, with a computer simulation program for control theory. Two experimental groups and one control group received an open-ended assignment for exploring a given modelled system. The major means of support that the experimental groups received was a structured overview of hypotheses. These overviews offered a list of, basically, the same set of eight predefined hypotheses from which subjects could choose. Two variations were designed: the controller structure followed types of controllers of increasing complexity and the concept structure organised the hypotheses according to fundamental domain concepts. The control group received the same assignment, but no support measures. Prior knowledge of all subjects was measured and at the end of the lab they were given a posttest that intended to measure ‘deep’ knowledge. Subjects worked on so-called ‘fill-in forms’ and their notes were used for analyzing their learning processes. Results showed that the Controller group scored higher on the posttest than the Concept group and subjects’ level of prior knowledge influenced the posttest scores. Analysis of statements on the fill-in forms showed that among others the Controller group designed better (more complete) experiments than the Concept group.
Melanie Njoo, Ton de Jong

Exploration Strategies in an Economics Simulation Game

Dealing with knowledge application is rarely perceived as part of education. However, there is much evidence that even very knowledgeable subjects often fail to adequately apply their knowledge. In order to provide learning environments where utilizable knowledge can be acquired, the analysis and induction of preferable learning strategies is important. Advantages and disadvantages of two learning modes during system exploration (specific vs. global) are discussed. In two empirical studies, their impact on later problem solving within an economics computer simulation is investigated. In the correlational Study 1 evidence is presented that specific learning mode may prove useful in complex, semantically rich domains. In the experimental Study 2 specific learning mode as exploration strategy was induced in the experimental group. This group yielded more success in later problem solving than the control group.
Hans Gruber, Alexander Renkl, Heinz Mandl, Wilfried Reiter

Determinants of Learning in Simulation Environments across Domains

A first objective of this study was to determine the nature of the relation between intellectual ability and metacognitive skill as predictors of novice learning. More specifically, the invariability of this relationship across domains was investigated. A second objective concerned the impact of domain knowledge on the relation between intellectual ability and metacognitive skill. Twenty-eight high or low intelligent psychology students passed through three different simulation environments: Heat lab; Stat lab, and a fictitious Deton (ation) lab. They were either novice or advanced in the domain of heat theory, whereas all of them were novices in the other domains. Thinking aloud protocols were analyzed on quality of working method (an operationalization of metacognitive skillfulness). Several measures of learning assessed the declarative and procedural knowledge for each domain. Results showed that the working method of novices had a reasonable amount of common variance across domains. Furthermore, it appeared that their working method, although related to intellectual ability, partly contributed to learning independent of intellectual ability. No such relationship was established for advanced students in the domain of heat theory. Their working method in Heat lab was unrelated to intellectual ability but it was related, however, to the level of prior knowledge about heat theory.
Marcel V. J. Veenman, Jan J. Elshout, John C. J. Hoeks

SEPIA: An Intelligent Training System For French Nuclear Power Plant Operators

SEPIA is an Intelligent Computer Aided Training System (ICAT) for training of incident management in the case of steam generator tube rupture in a nuclear power plant. SEPIA has been regularly used for 2 years by a few hundred operators of 20 French nuclear plants operated by EDF (French Electrical Utility).
the whole Computer Based Training facility consists of : 1) a Simulator on which the operator executes its training session during 1 or hours, and 2) SEPIA which produces intelligent debriefing, corrections and advices about the simulated session. The SEPIA specific architecture supports instructional model and measures, like detailed explanation functionalitites, a learner model allowing local quantitative and qualitative corrections as well as theoretical thermodynamics lessons, and a trainer model based on a comparison between theoretical and practical behaviours.
Written C and using the expert system shell S1 on UNIX, SEPIA has taken advantages of its object oriented modelization to extend successively from 900 MW to 1300 MW reactors domains and from training centers to auto-tutoring facilities. The multi-disciplinary approach of a team has led to a very high level of satisfaction recoreded at regular users groups meetings from learners, instructors and training managers.
Vincent Mercier, Daniel Delmas, Pascal Lonca, Jean-Jacques Moreau

Learning Impacts of the Alpin Expert Systems on its Users

The introduction of expert systems into the workplace has generally been studied from a technical point of view while the influence of such systems on human work environments remains poorly explored. In this paper we show that the introduction of an expert system implies new professional, organizational and cultural learning for operators. Employees attempt to restructure their previous professional skills according to their representation of the new tasks. This restructuring implies the rethinking of their decision making processes and how their work is organized. Expert systems contribute in this fashion to change in corporate culture.
Eric Brangier, Kent Hudson, Hélène Parmentier


Weitere Informationen