WO2000067223A2

WO2000067223A2 - Case-based, agent-assisted learning system and method

Info

Publication number: WO2000067223A2
Application number: PCT/US2000/011740
Authority: WO
Inventors: W. Lewis Johnson; Erin Shaw; Rajaran Ganeshan
Original assignee: University Of Southern California
Priority date: 1999-04-30
Filing date: 2000-04-28
Publication date: 2000-11-09
Also published as: WO2000067223A3; AU4686400A

Abstract

A case-based, agent-assisted learning tool is described that operates from a remote terminal. The case-based, agent-assisted learning tool permits delivery of animated pedagogical agent technology to remote terminals. The case-based agent-assisted learning approach supports collaborative instruction of multiple students working together. Each problem is presented as a simulated 'case' which the students must solve. The agent at a remote terminal is provided with just enough knowledge of the subject matter to be able to handle the case appropriately. This simplified approach makes it possible to download to the remote computer on demand just the information needed to enable the pedagogical agent to support the student's learning activities. This in turn makes it possible to integrate case-based agent-assisted learning with Web-based instructional materials. The case-based agent-assisted learning approach is capable of a wide range of pedagogical interactions, including question answering, hinting, opportunistic instruction, and referrals to background literature.

Description

CASE-BASED, AGEN -ASSISTED LEARNING SYSTEM AND METHOD

RELATED APPLICATION

This application claims the benefit of priority under 5 U.S.C. 119(e) of U.S. Provisional Application No. 60/131,746, filed April 30, 1999.

TECHNICAL FIELD

This invention relates to computer assisted learning, and more particularly to case-based agent-assisted learning.

BACKGROUND In agent-assisted learning, students solve problems on line. The students are assisted by a pedagogical agent (i.e., a virtual tutor) that monitors each student's actions and provides advice, feedback, and guidance as necessary. The pedagogical agent appears as an animated character that responds naturally to the student's actions, using speech and gestures, much like a real human tutor. The pedagogical agent incorporates intelligent tutoring capabilities, allowing the agent to provide students with guidance and assistance. One drawback of agent-assisted learning is the ability to access the program remotely. The amount of information required to run an agent-assisted learning program

- l - renders the program infeasible to run via a remote connection such as the Internet. Agent-assisted learning programs typically require a large database of information. In previous agent-assisted learning programs, this database of information needed to be available to the remote terminal. The inability to easily access the entire database in realtime from a remote terminal made operation of the program from a remote terminal unworkable. What is needed is an agent- assisted program that reduces the amount of information needed at a remote terminal, therefore allowing operation of the program from the remote terminal.

SUMMARY

In the present invention, each problem is presented as a simulated "case" which the students must solve. The agent at a remote terminal is provided with just enough knowledge of the subject matter to be able to handle the case appropriately. This simplified approach makes it possible to download to the remote computer on demand just the information needed to enable the pedagogical agent to support the student's learning activities. This in turn makes it possible to integrate case-based agent-assisted learning with Web-based instructional materials. Case-based agent-assisted learning improves on other animated pedagogical agent by permitting delivery of animated pedagogical agent technology to remote terminals. The case- based agent-assisted learning approach is capable of a wide range of pedagogical interactions, including question answering, hinting, opportunistic instruction, and referrals to background literature. The case-based agent-assisted learning approach supports collaborative instruction of multiple students working together. One aspect of the invention is a method of providing an agent-assisted learning tool comprising storing an agent and a simulation engine on a client and requesting case information from a server. The case information only includes data related to a specific session. The method further comprises transferring the case information onto a client and running the session on the client using the agent and simulation engine.

Another aspect of the invention is a case-based, agent-assisted learning tool comprising a server having a database. A plurality of cases are stored within the database. Each of the cases includes information directed only to a specific session. The learning tool further includes a client having an agent and a simulation engine. A communications path transfers one of the cases from the database to the client.

DESCRIPTION OF DRAWINGS

These and other features and advantages of the invention will become more apparent upon reading the following detailed description and upon reference to the accompanying drawings .

Figure 1 is a block diagram of the system architecture of the case-based, agent-assisted learning program according to an embodiment of the present invention.

Figure 2 is a flowchart illustrating the process of initializing a case according to an embodiment of the present invention.

Figure 3 is a block diagram of the system architecture of the case-based, agent-assisted learning program incorporating multiple clients according to an embodiment of the present invention.

Figure 4 is a diagram of a sample instructional window showing interaction with an agent according to an embodiment of the present invention.

Figure 5 is a flowchart illustrating the agent's lesson program according to an embodiment of the present invention. DETAILED DESCRIPTION

The system architecture for a case-based, agent- assisted learning system 100 according to the present invention is shown in Figure 1. The case-based, agent- assisted learning system 100 includes a central server 110 and a client 105. The central server 110 includes a database of information 155 and an agent for distant education (ADE) server 150. The client includes a simulation engine 115, an agent component 120, a text-to-speech engine 1.35, a task planner, assessor 145, and an ADE client 140.

The four main components of the system architecture are the pedagogical agent 120, the simulation engine 115, the client-server interface 140, 150, and the server database 155. The pedagogical agent consists further of two sub-components, the animated persona 125 and the reasoning engine 130. The reasoning engine 130 monitors the student's actions and provides appropriate feedback to the student. The central server 110 maintains the database 155 that includes the student's progress and when appropriate, provides synchronization for collaborative exercises carried out by multiple students on multiple computers.

The reasoning engine 130 performs all monitoring and decision making. The decisions of the reasoning engine 130 are based on a student model, a case task plan, and an initial state, which are downloaded from the server 110 when a case is chosen, and on the agent's current mental state, which is updated as a student works through a case. Upon completion, a record of the student's actions is saved to the database 155 in the server 110 where it can be used to assess the level of the student's expertise and determine how the program will interact with the student in future cases.

The simulation can be authored using the language or authoring tool of one's choice. All simulations communicate with the agent via a common application programming interface (API) that supports event (e.g., the student orders a lab) and state change (e.g., the lab value is updated) notifications as defined by the simulation logic. The animated persona 125 may be a Java applet that can be used alone with a JavaScript interface or incorporated into larger applications, such as with simulation-based exercises. If spoken words are desired, the animated persona 125 includes a text-to-speech engine 135. The present invention uses a case-based approach to intelligent tutoring. The courseware author encodes for each case just the knowledge needed to tutor that case. The ADE client' s 140 formal knowledge representation is concerned mainly with the procedural knowledge necessary to work through the case and cope with the contingencies that might arise in the course of the case work-up. Other relevant knowledge, such as background information, is incorporated into textual hints and Web-based reference materials. This information requires little memory and can be presented to the learner as needed. Thus, the run-time knowledge base and the reasoning engine 130 are simplified by reducing the amount of information required by the ADE client 140.

Figure 2 is a flowchart illustrating the process 200 of initializing a case according to an embodiment of the present invention. The process 200 begins at a START state 205. Proceeding to state 210, the user at the client 105 logs into the server 110. The client 105 may be a terminal in a general network, an intranet, or may be a remote terminal accessing the server 110 via modem or the Internet.

Proceeding to state 215, the client 105 requests case information from the server 110. The case information may be defined by a specific assignment the student is to complete. For example, if the case-based, agent-assisted learning system 100 is being used to instruct in a medical environment, the case may involve treatment of a patient suffering from a stab wound. The client 105 would request the stab wound case from the server 110.

Proceeding to state 220, the ADE server 150 retrieves the case information from the database 155. The ADE

- i server 150 accesses the database 155 to retrieve only the specific case requested by the client 105.

Proceeding to state 225, the ADE server 150 transfers the case information to the ADE client 140. Because the client 105 already contains the pedagogical agent 120 and the simulation engine 115, the case information is all that is necessary to process and run the case on the client 105.

Proceeding to state 230, the client 105 executes the case using the ADE client 140, the pedagogical, agent 120 the simulation engine 115, and the task planner, assessor 145. Details on running the case are discussed below. After the case is executed, the process 200 terminates in END state 235.

Artificial intelligence applications in general suffer from the so-called "knowledge acquisition bottleneck:" the difficulty of entering new knowledge into the system. The case-based approach significantly ameliorates the knowledge acquisition bottleneck, since knowledge acquisition focuses on one case at a time. Not only does this reduce the amount of knowledge that is required in a given case, it also places the knowledge in the concrete context of a particular case. These factors make it possible for nonprogrammer subject matter experts to take an active role in the formalization of knowledge used in the cases. Figure 3 is a block diagram of the system architecture 300 of the case-based, agent-assisted learning program incorporating multiple clients according to an alternate embodiment of the present invention. The multiple client system architecture 300 includes a server 355 having a session manager 360, a database 365, and a session planner 370. Multiple clients 305, 310 communicate with the server 355. The clients 305, 310 each contain an ADE client 315, 335, a simulation interface 325, 345, an agent interface 330, 350, and a case manager 320, 340.

The main server 355 component is the session planner 370. The session planner 370 provides information to the session manager 360 regarding an active simulation. The session manager 360 may create session objects. Each of the session objects defines a particular role in the simulation. A client 305 may create or join a session in a particular role. The case information for this role is then transferred to the ADE client 315 for processing by the simulation interface 325 and the agent interface 330. For example, in the stab wound case, the first client 305 may assume the role of attending physician while the second client 310 assumes the role of surgical consultant. Thus, separate case information is transmitted to each client 305, 310. Although each client 305, 310 is running a separate case, the session manager 360 coordinates the actions of the clients 305, 310 to allow interaction between the clients 305, 310 and to synchronize the simulation for all clients 305, 310.

Figure 4 is a diagram of a sample instructional window 400 showing interaction with an agent 410 according to an embodiment of the present invention. The instructional window 400 and an agent window 405 are representative of what would appear on the client system during the case. The instructional window may include a video section 415, a status section 420, and a data section 425. The agent window 405 may include an agent 410, a communications window 430, and event buttons 435, 440.

During execution of a case, the instructional window 400 may present to the user video information in the video section 415. This information may be a film clip, an animation, diagrams, pictures, or other video information. For example, in the stab wound case, a request by the user to examine the stab wound may produce a picture of the stab wound in the video section 415. The status section 420 provides information to the user on the specific case being executed. The status section 420 may include information on the case name, the elapsed time, other clients participating in the case, or any other information that needs to be conveyed to the user by the server. Of course, the status section 420 may be deleted if not desired.

The data section 425 may be used to provide information to the user and allow user interaction. The data section may provide information regarding the specific case that is not presented in the video section 415. For example, in a medical case, the data section may include ongoing information regarding the simulated patient's vital signs. The data section 425 may also contain interactive buttons that allow the user to request specific actions be performed in the case. The function of the interactive buttons may be defined by the case and may vary as the case progresses. For example, in the stab wound case, the interactive buttons may allow the user to start a blood transfusion or order an x-ray. The agent window 405 includes a graphical representation of a agent 410. The agent 410 may be an animated character or may be a video character. The agent 410 is capable of interacting with the instructional window 400. For example, the agent 410 may point to a specific item in the instructional window. The agent 410 responds naturally to the user's actions, using speech and gestures, much like a real human tutor. The agent 410 incorporates intelligent tutoring capabilities, allowing the agent 410 to provide users with guidance and assistance. The agent 410 is controlled by the both the animated persona 125 and the reasoning engine 130.

The agent window 405 also may include a communications window 430. The communications window 430 provides for a communication path with the user.

Communications in the form of text from the agent 410 may appear in the communication window 430. This allows the agent 410 to provide ongoing guidance to the user. For example, if a user requests a blood transfusion, the agent 410 may point out to the user that before starting the blood transfusion, it may be helpful to order a chest x-ray. This allows immediate feedback to the user, thus enhancing the learning experience. If a user desires more information, the user may press the "Why" button 435 to have the agent provide rationale of the course of action. Of course, the text-to-speech engine 135 may also be used to provide audible feedback to the user. The audible feedback may be used in addition to or in place of the communication window 430.

The reasoning engine 130 can be run in three modes. In the most restrictive mode, the reasoning engine 130 blocks actions whose preconditions are unsatisfied. The agent 410 uses this opportunity to provide unsolicited feedback about what should be done to satisfy the desired step' s preconditions. The agent window 405 displays a "Hint" button 440 so that a user may also ask for hints directly, before guessing or taking an incorrect action. In practice mode, the reasoning engine 130 does not block - the student can make mistakes - and the agent 410 does not provide unsolicited feedback, but still allows a student to ask for hints. In exam mode, the agent is not available.

The agents 410 knowledge representation focuses on the steps that the user should take to solve the case, the dependencies between them, and their rationales. The task steps and their dependencies are represented using a standard hierarchical plan. A plan hierarchy is comprised of steps, each of which is either a primitive action (e.g. corresponds to a simulation event) or a complex action (e.g. is itself a plan) . A step can have preconditions and end conditions, as well as hints, a rationale, a context and a role.

Preconditions and end conditions are represented by Boolean expressions in conjunctive normal form. The plan hierarchy is evaluated at each step to account for the dynamic nature of a simulation and the unpredictability of a student's actions. Actions whose goals become "undone" are automatically re- executed while those whose goals are implicitly satisfied are skipped.

The agent 410 also uses the notion of a situation space as a means of structuring the space of states associated with a domain so that it can be used to guide planning activity in dynamic situations. A situation is defined by a name, world state, goal expression, priority, and set of transitions which describe possible situations that can result from this situation whenever the associated conditions become true in the world state. Typically, when a situation is entered, a situation-appropriate sub-plan is instantiated to achieve the goal expression. Because the tutoring domain allows us to know' all possible situations a priori, all situational plans are pre-authored.

Figure 5 is a flowchart illustrating the agent's lesson process 500 according to an embodiment of the present invention. The process 500 begins at a START state 505. Proceeding to state 510, the agent initializes and sets the appropriate attributes. The initialization is based on the information provided by the server in the case information. As stated above, the agent is only provided with the information about the specific case to be run. By reducing the amount of information supplied to the agent, the latency of operation from a remote terminal is reduced.

Proceeding to state 515, the agent determines the appropriate event to execute. The event may be dictated by the case or by the user. The event may be that the case is completed, in which case an appropriate message would be displayed. The event may also be an input provided by the user to a question or circumstance in the case.

Proceeding to state 520, the agent determines if the event is the user clicking on the "Why" button 435. If the user has clicked on the "Why" button 435 requesting an explanation, the agent proceeds along the YES branch to state 525. In state 525, the agent determines if the question is valid in the specific context, and if so provides the appropriate rationale. After providing the rationale, the agent returns to state 515 to determine the next event.

Returning to state 520, if the user has not clicked on the "Why" button 435, the agent proceed along the NO branch to state 530. In state 530, the agent determines if the user has clicked on the "Hint" button 440 requesting help. If the user has requested a hint, the agent proceeds along the YES branch to state 535. In state 535, the agent provides a hint to the user based on the information in the case. The hint may be a specific hint authored into the case for the circumstance, or may simply be a suggestion to perform the next step in the case. After providing the hint, the agent returns to state 515 to determine the next event.

Returning to state 530, if a hint is not requested, the agent proceeds along the NO branch to state 540. In state 540, the agent evaluates the current event to determine if the event is complete. If the event is complete, the agent progresses to the next event.

Proceeding to state 545, the agent determines if all of the pre-conditions for the event were accomplished by the user. If the conditions are met, the agent proceeds along the YES branch to state 550. In state 550, the agent provides a response to the user. The response may be to indicate the user successfully completed the step. After providing a response, the event terminates in END state 56.0. Returning to state 545, if the conditions for the event were not met, the agent proceeds along the NO branch to state 555. In state 555, the agent may have determined that the action the user has taken is not appropriate for the event. The agent then provides feedback to the user suggesting a proper course of action. After providing feedback, the event terminates in END state 560

Numerous variations and modifications of the invention will become readily apparent to those skilled in the art. Accordingly, the invention may be embodied in other specific forms without departing from its spirit or essential characteristics .

Claims

WHAT IS CLAIMED IS:

- 1. A method of providing an agent-assisted learning z tool comprising:

3 storing an agent and a simulation engine on a client;

- requesting case information from a server, wherein the

5 case information only includes data related to a current

6 session ;

7 transferring the case information onto a client; and

8 running the session on the client using the agent and

9 simulation engine.

1 2. The method of Claim 1, further comprising storing

2 the case information in a database on the server.

1 3. The method of Claim 1, further comprising

2 communicating the case information over a network.

1 4. The method of Claim 3, wherein the network is the

2 Internet.

1 5. The method of Claim 3, wherein the network is an

2 intranet.

1 6. The method of Claim 1, further comprising managing

2 the session from the server.

7. The method of Claim 6, further comprising communicating the case information to multiple clients and coordinating the session between the clients.

8. The method of Claim 1, wherein the agent includes a reasoning engine and an animated persona.

9. A case-based, agent-assisted learning tool comprising:

a server having a database;

a plurality of cases stored within the database, each of the plurality of cases including information directed only to a current session;

at least one clients, wherein the at least one clients includes an agent and a simulation engine; and

a communications path which transfers one of the plurality of cases from the database to the at least one clients.

10. The case-based, agent-assisted learning tool of Claim 9, wherein the communications path is a network connection.

11. The case-based, agent-assisted learning tool of Claim 9, wherein the communications path is the Internet.

12. The case-based, agent-assisted learning tool of Claim 9, wherein the communications path is an intranet.

13. The case-based, agent-assisted learning tool of Claim 9, wherein the agent comprises a reasoning engine and an animated persona.

14. The case-based, agent-assisted learning tool of Claim 9, further comprising a session manager which coordinates the session among the at least one clients.

15. The case-based, agent-assisted learning tool of Claim 13, wherein the agent further comprises a text-to-speech engine .

16. The case-based, agent-assisted learning tool of Claim 9, wherein the agent provides feedback to a user.