Architecture of a generic instructional planner

Architecture of a generic instructional planner

Journal of Network and Computer Applications (1996) 19, 265–274 Architecture of a generic instructional planner Sun-Teck Tan Department of Informatio...

88KB Sizes 0 Downloads 71 Views

Journal of Network and Computer Applications (1996) 19, 265–274

Architecture of a generic instructional planner Sun-Teck Tan Department of Information Systems & Computer Science, National University of Singapore, Kent Ridge, Singapore 0511 This paper describes the architecture of an instructional planner in an intelligence tutoring system. A three-layer blackboard model is presented. In this model, each knowledge source can be designed independently. New knowledge can be added to the system without affecting the design of other knowledge sources. Therefore, the architecture provides the possibility of capturing more intelligence in future. Included in this architecture are three knowledge sources namely, the domain knowledge, pedagogical knowledge and the knowledge of individual student. The architecture also reflects multiple flexible tutoring strategies in providing instruction for individual student.  1996 Academic Press Limited

1. Introduction The days when sequential presentation of pre-set tutorial material to the computeraided instruction (CAI) user have long gone. For intelligence tutoring systems (ITS) of the current age, the system must either provide a very flexible user interface for the user to control the tutorial session, such as those suggested by Schank [1] or a good instructional planner must be developed and incorporated into the tutoring system to generate instructional plan to suit each individual user of the system [2,3]. An instructional planner is responsible for creating instructional plans, requesting execution of these plans and reacting to student’s responses by replanning [4]. Instructional planning requires to synthesize the knowledge of the domain, the knowledge of the student and the knowledge of pedagogy. All the knowledge are independent knowledge sources. It is necessary to develop a conceptual model which allows for the incorporation of the above knowledge sources. Furthermore, as the attempt to include more intelligence, the ITS system should allow additional knowledge sources to be embedded easily. Blackboard model provides a generic knowledge-based problem solving mechanism that can manipulate multiple knowledge sources effectively [5]. Therefore, the main objective of the instructional planning component in our research is to ensure that our ITS system is able to help students learn the domain knowledge in an instructionally sound manner by providing effective instruction. Our instructional planner is modeled by a blackboard model [6]. The planner will be used in an ITS system to teach the PASCAL programming language to the first year students in the National University of Singapore. In the next section, we provide a brief introduction to the blackboard control mechanism. This will be followed by the description of how we could extend the blackboard model to form the architecture of our planner. We will then describe the relationship between the instructional planner and the other components of a typical ITS system. A brief summary will conclude this paper. 265 1084–8045/96/030265+10 $18.00/0

 1996 Academic Press Limited

266 S.-T. Tan

Knowledge source

Blackboard

Knowledge source

Knowledge source

Figure 1. A blackboard model.

2. A blackboard model A blackboard model is a problem-solving model which is a scheme for organizing reasoning steps and domain knowledge to construct a solution to a problem [7]. It consists of two basic components: knowledge sources and a blackboard, as shown in Fig. 1. The knowledge sources are particular kinds of knowledge needed to solve the problem in a domain. These knowledge sources are kept separately and logically independent. The blackboard is a global database which provides the only communication and interaction media for the knowledge sources. The knowledge sources, activated one at a time, produce changes to the blackboard which lead incrementally to a solution to the problem. However, the blackboard model described in Fig. 1 is a conceptual entity, not a computational specification. Given a problem to be solved, the blackboard model provides enough guidelines for sketching a solution, but a sketch is a long way from a practical system. Therefore, in a practical blackboard system, there must be a controller added as in the following figure. 2.1 The knowledge sources The objective of each knowledge source is to contribute information. It can be a knowledge-based system, an algorithm or a human. During problem solving, the knowledge sources are activated by the controller to read from and write to the blackboard. It develops the solution incrementally. The knowledge sources themselves cannot communicate with each other. They communicate by modifying the data on the blackboard. All modifications to the solution state are explicit and visible.

Architecture of a generic instructional planner 267

Knowledge source Blackboard

Panel Knowledge source

Knowledge source

Controller

Figure 2. A blackboard framework.

2.2 The blackboard The purpose of the blackboard is to hold computational and solution-state data needed by and produced by the knowledge sources. It contains data to be used by the knowledge sources in the performance of their tasks as well as information relating to the current session. To represent and organize different kinds of information, a blackboard can be partitioned into several panels as shown in Fig. 2. Furthermore, the information in the panels is carried by objects which are units for carrying information between the knowledge sources and the blackboard.

2.3 The controller The responsibility of the controller is to monitor the changes on the blackboard and to decide what kind of actions to take next. It activates/suspends a knowledge source by sending a message to it. A message is a request for a knowledge source to carry out one of its actions. On the basis of the current information on the blackboard, the controller sends a message to the appropriate knowledge source and generally decides to take the following three kinds of actions: (1) to activate the next knowledge source; (2) to determine which objects the knowledge source should apply; and (3) to pursue on next solution.

268 S.-T. Tan

CT R

Blackboard

KS

KS

CT

R

Blackboard

KS

KS

Figure 3. Extension architecture of a blackboard framework

2.4 Extensions to the blackboard model A knowledge source can be a blackboard framework itself with its own set of subknowledge sources, a blackboard and a controller, as shown in Fig. 3. This representation of a knowledge source as a blackboard model allows for a top-down decomposition of the knowledge sources. For the multi-layer blackboard architecture, the information communication between two layers is carried out by the controllers. The controllers have the capability to move information objects from one blackboard to the other.

3. Architecture of the instructional planner PASP Instruction planner is one of the components of an ITS, it should be designed in an ITS environment. This section introduces an ITS model in which an instructional planner PASP (PAScal Planner) is designed and followed by the architecture of PASP. 3.1 An ITS model overview Figure 4 is an ITS model which provides the research context for the instructional planning component (PASP). It consists of five knowledge sources: a Domain Expert, an Instructional Planner, a Student Module, a Communication Manager and a Student and a blackboard which records the information of the domain, of the student and of the curriculum. Although different ITS models have been proposed, most systems are capable of performing functions identified in the ITS models proposed in Fig. 4. The Domain Expert contains the instructional materials. It posts the domain knowledge structure to the blackboard to be used by the Instructional Planner. The Student

Architecture of a generic instructional planner 269

Student module

Instructional planner

Student model

Curricullum

S T U D E N T BB-0

Domain knowledge

Dialogue manager

Domain expert ITS Figure 4. A blackboard model for an ITS.

Module updates the student information recorded in the student model on the blackboard while the instructional session proceeds. The Instructional Planner provides curriculum and dynamically adapts it based on the information in the student model and the domain knowledge structure. The curriculum provided by the Instructional Planner contains the sequences of instructional lessons to be presented to the student, not including information about actual presentation style which is the responsibility of the Communication Manager. A possible session with this kind of architecture would begin with the Instructional Planner presenting the student with a problem through the Communication Manager. The student response with solution steps also through the Communication Manager. Using the solution of the Domain Expert, the Student Module critiques the solution, and updates the student model. The Instructional Planner selects tutorial material using the information from the student model, the teaching strategies, the curriculum and other knowledge base to decide what to present next and the process cycles. 3.2 The PASP architecture The main task of PASP is to dynamically create instructional plans, monitor plan execution process and replan at the appropriate time. Thus, three knowledge sources

270 S.-T. Tan BB-0 Domain panel

Planner

Student panel

Monitor

Control panel

Executor

C T R L 1

BB-1 Plan panel

Control panel

PASP Figure 5. The architecture of PASP.

which carry out the above three tasks are designed. Three knowledge sources include the Planner for generating instructional plans, the Monitor for monitoring the plan execution and recognizing the learning blockage and the Executor for plan execution. Figure 5 shows the architecture of PASP. Besides three knowledge sources, PASP includes its own blackboard and a controller as well. In addition, the domain knowledge and the student information provided by the Domain Expert and the Student Module on the highest-layer blackboard (BB-0) are also related to the PASP architecture. PASP is activated when an instructional objective is determined. Then, PASP synthesizes the knowledge of the domain and of the student to generate an initial instructional plan. Since instructional planning is an incremental process, configuring instructional sequences is interleaved with plan execution. Therefore, after the initial instructional plan is generated, the Executor is activated to execute the plan and, at the same time, the Monitor is activated to monitor the plan execution process at the background. Once the Monitor recognizes that the current instruction does not fit the student’s current learning situation, the Planner will be activated again to revise/ replan the previous plan and the cycle continues. The above plan generation/execution

Architecture of a generic instructional planner 271 interleaved process won’t terminate until the instructional objective is achieved by the student. 3.2.1 Blackboard-0. Blackboard-0 is the blackboard of the ITS blackboard model. It is the communication channel between PASP and the other knowledge sources in the ITS model. All the information PASP needs for generating instructional plans are posted at this blackboard. The instructional plan generated is posted on it as well. The information on the blackboard-0 is categorized into three panels: the domain knowledge panel for the domain knowledge, the student model panel for the student information, and the control panel for the control data. 3.2.2 Blackboard-1. Blackboard-1 is the local blackboard of PASP. It is used to represent instructional plans as well as control information to co-ordinate the activation of the knowledge sources. Three panels are contained on this blackboard, namely, instructional plan panel which records instructional plans generated by the Planner, control panel which records the control information for the activation of knowledge sources and common data panel which records temporary information during the plan generation/execution process. 3.2.3 Controller-1. Controller-1 is the control of PASP which is responsible for coordinating knowledge sources by sending ‘activate’ or ‘suspend’ message to different knowledge sources and for transmitting information between blackboards in different layers. In particular, it handles the control message sent by knowledge sources, sends control message to the corresponding knowledge sources and transmits instructional plans back to BB-0, etc. With the help of controller-1, the knowledge sources of PASP are capable of accessing information on the blackboard (BB-0) in the ITS model. 3.2.4 Planner. The Planner is responsible for generating instructional plans. It is activated when it receives an ‘activate’ message sent by Controller-1. When it is activated, it generates an instructional plan based on the information of the domain and of the student on BB-0 and presents the planning result to the instructional plan panel on BB-1. After that, the Planner sends a ‘PlanGenerated’ message to Controller-1. The Planner itself is also a blackboard model which has its own knowledge sources and blackboard. Figure 6 is the blackboard architecture of the Planner. It comprises six knowledge sources: the Goal Generator, the Content Generator, the Course Sequencer, the Lesson Selector, the Presenter and the Plan Reporter, with its own blackboard and controller. The Goal Generator is responsible for dividing the instructional objective into a sequence of subgoals so that the system can help the student achieve the instructional objective step by step and select an appropriate one for the student to achieve at each instructional point. Content Generator is responsible for generating content plans that determine which concepts in the concept knowledge base should be taught at each instructional point for the student to achieve the current goal. The Course Sequencer transforms a content plan into a course sequence. A content plan is sequenced based on a bottom-up strategy. If a node has subnode(s), the subnode(s) will be taught before their parent node. The Lesson Selector selects the

272 S.-T. Tan BB-1 Plan panel

Plan reporter

Control panel

Presenter

Lesson selector

C O N T R O L L E R 2

BB-2 Goal panel

Goal generator

Conecpt panel

Content generator

Control panel

Course sequencer PLANNER

Figure 6. The blackboard framework for the planner.

most appropriate lesson for the student to learn. The Controller is responsible for coordinating the activation of the above knowledge sources. 3.2.5 Monitor. The Monitor is responsible for monitoring the plan execution process and determining what to do next, e.g. keeps silence at the background; interrupt plan execution process. It requires that the Monitor has the ability to recognize whether a learning blockage is encounter based on the student information provided by the Student Module and to decide whether to abandon the current plan totally or to revise it. Once the Monitor finds that the plan execution needs to be interrupted, it will send an ‘interrupt’ message to Controller-1. 3.2.6 Executor. The Executor is responsible for executing instructional plans. It is activated when it receives an ‘activate’ message and suspended when it receives a ‘suspend’ message sent by Controller-1.

Architecture of a generic instructional planner 273 DE

SM

Blackboard

IP

STU

DM ITS

M

Blackboard

E

CTR1

P PASP

Blackboard

CTR2

PLANNER Figure 7. A three-layer ITS blackboard framework.

4. Summary A three-layer blackboard ITS model as shown in Fig. 7 is proposed in our research. This architecture provides a genetic model for ITS. It enables each knowledge source to be designed independently so that the new knowledge can be included to the system without affecting the design of other knowledge sources. Furthermore, this architecture provides a possibility for ITS to capture more intelligent in future.

References 1. R. Schank 1991. Button theory: a taxonomic framework for student-teacher interaction in computer-based learning environment, Northwestern University, Evanston, Illinois. 2. P. Mohan et al. 1992. Enhancing the instructional capabilities of intelligent tutoring systems. Proceedings of the Second Intelligent Conference, ITS’92, 531–538. 3. D. R. Peachey et al. 1986. Using planning techniques in intelligent tutoring systems. Int. J. Man-Machine Studies, 24, 77–98.

274 S.-T. Tan 4. S. A. Macmillan et al. 1987. An Architecture for a self-improving instructional planner for intelligent tutoring systems. Computational Intelligence J., 3, 17–27. 5. L. D. Erma et al. 1975. A multi-level organization for problem solving using many diverse, cooperating sources of knowledge. In Proceedings of Fourth International Joint Conference on Artificial Intelligence, 1, Tbilisi, USSR. 6. B. Hayes-Roth 1985. A Blackboard architecture for control. Artificial Intelligence, 26, 251–321. 7. B. Wasson 1990. Determining the focus of instruction: content planning for intelligent tutoring systems. Research Report of the University of Saskatchewan, Canada. S.-T. Tan obtained his BSc from the Nanyang University, Singapore in 1980 and he received his MSc and PhD from Essex University, England in 1984 and 1989 respectively. He is presently a lecturer in the Department of Information Systems and Computer Science at the National University of Singapore. His present research interests include Expert Systems in Planning, Intelligent Tutoring Systems, Multi-medium and Chinese Computing.