Adaptive Sequencing and Navigation:

Generic Meta-Models for Learning Technology Standards

Vladimir Goodkovsky, PhD

University of Virginia, USA

vladimir@virginia.edu

Abstract

The paper presents a generic hierarchy of basic tutoring meta-models that are necessary and sufficient for adaptive sequencing learning activities/assets/resources within a unit of computer-based instruction. Proposed meta-models cover existing Computer-Based Training (CBT), Web-Based Training (WBT), Job Support Systems, Intelligent Tutoring Systems (ITS) and indicate the path for smooth transition from simple to intelligent sequencing in Sharable Content Object Reference Model (SCORM) development.

Keywords: Adaptive, Intelligent, Tutoring, Learning, Sequencing, Navigation, Technology, Standard, SCORM.

Introduction

Adaptive and/or intelligent sequencing of learning activities/resources is one of the most challenging topics in SCORM development. Many research groups from the Intelligent Tutoring Systems (ITS) community develop their solutions for adaptive sequencing. But it appears that available solutions are still too specific, heuristic, complex and not mutually compatible. There is no one generic model able to cover all known approaches yet.

Widely known classical definition of ITS specifies only their main components: Expert System, Learner Model, Instructional Module and Interface [ Wenger E. (1987), Burns H. L. & Capps C. G. (1988), Mandl H. & Lesgold A. (1988)]. This definition is not very explanatory and constructive for practical use. Moreover it focuses on procedural knowledge/skills learning only.

In this paper, a more explanatory and constructive conceptual architecture of ITS (an Intelligent Instructional Unit or Adaptive/Intelligent Learning Object) is proposed based on a nested hierarchy of the following models:

1. Domain Situation, which is a snapshot of the domain under study;
2. Domain under the learner's study;
3. Task(s) to solve/perform in the Domain by the learner;
4. Expert, who is able to solve the task(s) in the Domain;
5. Tutoring Situation representing a snapshot of the learner defined as an extension of the expert with possible deviations;
6. Tutoring Domain specifying all possible Tutoring Situations (learner's snapshots);
7. Tutoring Task(s) to solve, which in general is a sequencing of domain Situations /Tasks and/or control over learner's own navigation among them;
8. Tutor-Expert, which is able to solve the tutoring task(s) and thus representing a Goal of instructional research/design, and
9. Administrator, who manages access of learners to specific course units.

Each next model includes all previous ones (in simplified and more abstract form). See fig.1.

Fig.1. Hierarchy of Meta-Models

This hierarchy of meta-models is generic enough to put together known ITS conceptual pieces, to cover CBT, WBT and ITSs, and to map the road for smooth transition of SCORM from simple sequencing to adaptive/intelligent one.

What is already available?

Let us consider how proposed hierarchy explains already available Computer-Based Training systems (CBT), Simulator/Game-based Intelligent Tutoring Systems (ITS), Expert System-based ITS, Job Support Systems, and Learning Management Systems (LMS).

CBT, which are most widely used and only covered by existing Sharable Content Object Reference Model (SCORM), are able to:

1. recreate specific (multimedia) learning situations and by this way
2. represent the Domain under study;
3. present specific tasks/questions as well as
4. pre-store correct answers (as a trivial representation of an expert's solution of the task(s)/question(s));
5. pre-store incorrect answers (as a trivial extension of the expert model into a learner model) representing the tutoring situations;
6. represent the entire tutoring domain (including all possible learner's responses on tasks and situations);
7. represent the tutoring task (what is the next situation/task?) and
8. represent a script/flowchart pre-sequencing the next situations and tasks (as a trivial tutor-expert model).

Known simulators/games are often able to model a continuum of learning situations in domain under study. Corresponding simulator/game-based ITSs can:

1. present specific domain situations to the learner and by this way
2. represent the entire domain under study;
3. present specific domain task(s) to the learner;
4. pre-store a correct solution/performance (as a simple expert model) as well as
5. may pre-store incorrect solution / performance (as an extension of the expert model into a learner model) representing the tutoring situation;
6. represent the entire tutoring domain (including all possible learner's responses on situations and solutions/performances of tasks);
7. represent tutoring task (at least the task to compare learner's solution with pre-stored expert's one), and
8. represent the tutor-expert (at least realizing performance comparison, shallow feedback, and score evaluation).

Known Expert Systems are able to model an expert solution of the domain tasks. Corresponding Expert System-based ITSs are able to:

1. present the learning domain situations,
2. the domain itself and
3. task(s) in the domain under study;
4. perform an expert solution of the task(s). In contrast to previous CBT and simulator-based ITS, the expert solution of the task is generated automatically;
5. represent the tutoring situation as a snapshot of learner's solution of the task;
6. represent tutoring domain as a variety of all tutoring situations;
7. represent tutoring tasks (at least the task to compare a learner solution with an automatically generated solution in real time, and provide shallow tutoring feedback and score evaluation);
8. realize the tutor-expert's solution of these tasks.

In known Job Support systems,

1. situations,
2. the domain and
3. task(s) are defined by the job itself. But in order to detect and react on occasional deviations/faults from normative/expert performance of the job/task(s), the advanced Job Support systems should include at least
4. the expert model. To be able to react on some specific (for example, critical) deviations/faults, the Job Support systems should also include
5. a worker model specifying such deviations/faults and defining by this way tutoring situations, which all together form
6. the tutoring domain;
7. Tasks of job support. The best Job Support should perform not only simple tasks such as to alert the worker about deviations from normal routine, but also a diagnostic task (to diagnose deep causes of those deviations) and a correction task (to correct the causes to prevent more deviations/faults). Actually such tasks are attractive not only for Job Support systems, but for all educational and training systems.
8. The tutor-expert should be able to solve all those tasks.

Known Learning Management Systems (LMS) can control over all above mentioned kind of systems (plus over some others) and provide access of each particular learner to each specific unit of instruction (to the tutor-expert model). But in common practice, all LMSs do is just supplying learners with learning content (learning situations, domain, and tasks) for self-study without any control.

What is missing?

Known models of learning situation, domain, task, and expert are pretty well developed by now. In contrast, the tutoring situation (or learner model), tutoring domain, tutoring task and tutor-expert models are rather under-developed in theory and trivial in practice. They are just much more complex in comparison with lower level models they include. Actually the formal models of tutoring situation, tutoring domain, tutoring task and expert-tutor of generative type are yet to be developed. See fig.2.

The good news is that the well-developed lower level domain/expert models can serve as a solid foundation for building under-developed high level tutoring models.

Fig.2. Tutoring Gap in available Meta-Models

State of the art

The existing variety of approaches to ITS research, design and development can be categorized as a Bottom-Up Approach or Generalization . Indeed, the Case-Based techniques are based on machine learning/generalization of demonstrated tutoring behavior, Rule-Based techniques - on human heuristic generalization of tutoring behavior; and Model-Based techniques - on theoretical models of tutoring behavior.

As a rule, researchers working within the Bottom-Up Approach focus on specific aspects/parts, not on the entire tutoring process/picture. They rely mostly on personal experience, best practices, specific pedagogical paradigms and available technologies. No doubt this is a very strong foundation. But by definition, the Bottom-Up generalizations are always limited and error-prone.

Advocated Approach

In order to minimize inevitable generalization heuristics in development of tutoring situations/domain/tasks/expert models and to meet, justify and align known results of Bottom-Up generalizations, the Top-Down Approach has been proposed. In essence, this is a Systematic Specification of approved and recognized theoretical meta-models of modern exact sciences.

Such a Top-Down Approach is very challenging because it is targeted to creation of the most generic and powerful meta-models of tutoring. It is also a very conservative approach because it is based on already approved and recognized theoretical frameworks. Simply put, it is just a right way of modeling, which is native to technology. For more info on this subject see [Goodkovsky (1997)].

Unfortunately by now, exact sciences are not distilled and organized into simple generic models to use. Actually known tendency of scientific knowledge integration is not so strong as the tendency of producing new specific knowledge. That is why one needs to study a lot of specific sciences to reveal such generic models. This is a serious challenge for all researchers, especially for those who work in interdisciplinary fields such as instructional design.

What is the most important is that only synergy of Bottom-Up generalization and Top-Down specification is able to provide the best unified basis for learning technology standards, SCORM in particular.

Meta-Models Developed

Tutoring Domain Model Composition:

• State Space:
  • Learning Objectives:
    • Achievement States:
      • Ground State;
      • Supplied State;
      • Achieved State;
  • Prerequisite Relations/Beliefs among learning objectives to specify tutoring strategy;
  • Learner Personal Data;
• Behavioral Space:
  • Tutoring Assignments:
    • Domain Situations/Tasks:
  • Expected Responses on each Tutoring Assignment;
• State-Behavior Relations:
  • State-Behavior Cross-Beliefs.

Tutoring Situation (extended Learner Model) Composition:

• State Part:
  • Learning Objectives:
    • Objective:
      • Ground State:
        • Belief;
      • Supplied State:
        • Belief;
      • Achieved State:
        • Belief;
• Behavioral Part:
  • Tutoring Assignments presented:
    • Domain Situation(s)/Task(s);
  • Expected Response(s) received;
• Learner Personal Data:
  • Strict Requirements of the learner;
  • Soft Preferences of the learner;
  • Adapting Parameters:
    • Difficulty Limit;
    • Testing Delay;
    • Granularity et cetera.

Tutoring Task Model Composition:

• What is given to the tutor:
  • Tutoring Domain Model including:
    • Learning Objectives in
      • Achieved State;
  • Tutoring Situation Model;
• What is required from the tutor:
  • To sequence Tutoring Assignments in order to provide specific domain situations/tasks to achieve achieved states of learning objectives by the most effective way (saving on resources and/or time).

Expert-Tutor Model Composition:

• Logic Aspect (Intelligent Sequencer):
• Media Aspect (Content Delivery):
  • Learning Domain Multimedia:
    • Domain Situations (Presentations, Simulations, ...);
    • Domain Tasks (questions, test items, exercises, missions, ...)
    • Expert Responses (correct answers / actions / solutions / performances, ...)
    • Learner Responses (correct + incorrect answers / actions / solutions / performances,...)
  • Tutoring Persona Media:
    • talking head,
    • feedback window,
    • ...;
• Logic-Media Converter:
  • Controller: Logic --> Media;
  • Monitor: Media --> Logic.

Intelligent Sequencer Composition:

• Tutoring Knowledge Framework:
  • Tutoring Domain Model:
    • State Space;
    • Behavioral Space;
    • State-Behavior Relations;
  • Tutoring Situation Model;
• Tutoring Run-Time Engine;
  • Decision Maker;
  • Knowledge/Data Processor.

Tutoring Run-Time Engine Composition:

• Tutoring Decision Maker:
  • Strategic Decision Maker;
  • Tactic Decision Maker;
  • Operative Decision Maker;
• Tutoring Knowledge/Data Processor:
  • Tutoring Situation Model Updater;
  • Tutoring Situation Model Reviser.

Strategic Decision Maker Function:

• Data to take into account:
  • Tutoring Situation Model;
  • Predefined Thresholds:
    • Supplying,
    • Testing,
    • Diagnosing;
• Decisions to make:
  • All objectives are supplied;
  • All objectives are achieved;
  • One objective is diagnosed.

Tactic Decision Maker (Mode Sequencer) Function:

• Data to take into account:       
  • Available Modes:
    • Supplying,
    • Testing and
    • Diagnosing;
  • Current Tutoring Mode:
    • Supplying,
    • Testing or
    • Diagnosing;
  • Tutoring Style Parameters;
    • Fault Tolerance;
    • Testing Delay;
    • ...
  • Tutoring Situation Model;
• Decision to make:
  • The next Tutoring Mode.

Operative Decision Maker (Assignments Sequencer) Composition:

• Sharp filter, which pre-selects a subset of assignments (assigning certain activities/resources) for a learner own navigational choice;
• Soft filter, which provides rating of pre-selected assignments (assigning certain activities/resources) for a learner's informed navigation/choice;
• Final selector provides one single assignment of specific activity/resource for the learner (if the learner has chosen nothing so far).

All these models can be used in SCORM development in order to check, justify, and upgrade already available meta-models for simple sequencing.

Implementation

Introduced meta-models have been under extensive research, development, testing, and honing for decades. Their prototypes have been used in several generations of ITS applications and authoring tools implemented in education, industry, business, and defense. All developed ITSs demonstrated the most effective dynamic adaptive education and training.

Summary

The tutoring activity is very complex subject. Despite its complexity, there is a possibility:

•  to generalize existing (mostly heuristic and not mutually compatible) representations of tutoring activity simulated in modern learning technologies;

•  to find relevant theoretical frameworks in modern exact sciences;

•  to specify the theoretical frameworks down to theoretically sound meta-models of tutoring activity for learning technologies.

This possibility has been explored. As a result a hierarchy of tutoring meta-models has been developed. These meta-models fill in the existing gap between known Learning Management and Content Delivery Systems.

Proposed meta-models are generic enough to cover a variety of regular CBT, WBT, Job Support and Intelligent Tutoring Systems. Due to this wide coverage they indicate the path for smooth transition from simple (like in CBT) to adaptive (like in ITS) sequencing of learning activities / assets / resources within instructional units.

The logical aspect of the models is completely separated from its media (presentation) aspect. It allows reusing already available media resources, multimedia-oriented authoring tools and content delivery platforms.

Generic (domain/learner-independent) part of the models' logical aspect can be used as authoring frameworks for easy design of new instructional units.

Due to openness of specification of tutoring domain, learner's requirements, preferences and adaptable parameters and possibility to specify them down to very fine detail, the adaptive capabilities of the tutor-expert based upon proposed models are practically unlimited.

The proposed meta-models are developed to contribute to SCORM intelligent sequencing.

References

Wenger E. (1987). Artificial Intelligence and Tutoring Systems. Morgan Kaufmann. Los Altos, CA.

Burns H. L. & Capps C. G. (1988). Foundations of intelligent tutoring systems: an introduction. Foundations of intelligent tutoring systems (Eds. Polson M. C. & Richardson J. J.). Lawrence Erlbaum, London, pp1-19. 

Mandl H. & Lesgold A. (1988). Learning Issues for Intelligent Tutoring Systems. Springer-Verlag, London. 

Goodkovsky, V.A. & Kazennov, A.Y. (1992) Intelligent Tutoring Systems: Theory, Technology, and Practice. Proceedings of the East-West conference on emerging computer technologies in education. Moscow, RU: ICSTI, pp.123-127.

Goodkovsky, V.A. (1996). Intelligent Tutor: Shell, Toolkit & Technology. ITS'96 Workshop on Architecture and Methods for designing cost-effective and reusable ITSs, Montreal, Canada, June 1996.

Goodkovsky, V.A. (1997). Intelligent Tutor: Top-down Approach to Intelligent Tutoring System Design. Developing Technical Standards for Learning Technology. Learning Technology Standards Committee (WG P1484).
http://ltsc.ieee.org/archive/harvested-2003-10/miscellaneous/goodkov/goodkov.htm

Goodkovsky, V.A. (2004). Unified Generator of Intelligent Tutoring. No: 10/909,101. US Patent Application.

Rob Coper. (2001). Modeling of units of study from pedagogical perspective the pedagogical meta-model behind EML. Educational Technology Expertise Centre. Open University of the Netherlands. First draft, version 2. June 2001.
http://eml.ou.nl/introduction/articles.htm

SCORM 2004 Sequencing & Navigation. Learning Technology, Publication of IEEE Computer Society. Technical Committee on Learning Technology (LTTC). Guest Editors: Dr. Eric Roberts and Dr. Michael W. Freeman
http://lttf.ieee.org/learn_tech/issues/january2005/#_Toc98675008

Technical Evolution of SCORM. Learning Systems Architecture Lab. Carnegie Mellon University. http://www.lsal.cmu.edu/lsal/expertise/projects/scorm/scormevolution/reportv1p02/report-v1p02.html

Modritscher, F. and others. Enhancement of SCORM to support adaptive E-Learning within the Scope of the Research Project AdeLE. http://www2.iicm.edu/cguetl/papers/adaptiveelarningstandards/adaptivestandard.pdf