Designing interactive learning environments: an approach from first principles

The Authors

Bernard Scott, Defence Academy, Cranfield University, Shrivenham, UK

Chunyu Cong, Defence Academy, Cranfield University, Shrivenham, UK

Abstract

Purpose – Today's technology supports the design of more and more sophisticated interactive learning environments. This paper aims to argue that such design should develop from first principles.

Design/methodology/approach – In the paper by first principles is meant: learning theory and principles of course design. These principles are briefly outlined, key features of which include: the use of knowledge and task analysis methodology; the use of topic maps; and learning design that supports adaptive teaching. The paper goes on to describe how this approach to course design has been applied at the UK Defence Academy.

Findings – The paper shows how conversation theory serves as a source of first principles for the design of interactive learning environments, as exemplified in the Course Assembly System and Tutorial Environment (CASTE).

Originality/value – A principled approach to the design of interactive learning environments is of value in bringing order to a conceptually and methodologically confused domain of practice.

Article Type:

Research paper

Keyword(s):

Learning methods; Conversation; United Kingdom.

Journal:

Campus-Wide Information Systems

Volume:

24

Number:

3

Year:

2007

pp:

174-186

Copyright ©

Emerald Group Publishing Limited

ISSN:

1065-0741

1. Introduction

In this paper we present an approach from first principles to the design of interactive learning environments. In section 2, we present conversation theory (CT), a theory of learning and teaching with principled foundations in cybernetics. Some variants of CT are described in section 3. In section 4, we describe how the pedagogic principles derived from CT were embodied by Pask and Scott (1973) in the Course Assembly System and Tutorial Environment (CASTE). In section 5, we outline a framework for course design based on CT. In section 6, we describe how we have used a theory grounded, principled approach to develop courses for the UK Ministry of Defence at the Defence Academy. In section 7, we briefly describe our current work developing topic map based interactive learning environments that support adaptive teaching.

2. CT

CT, as developed by Pask and Scott (1973), originated from a cybernetics framework and attempts to explain learning in both living organisms and machines. CT takes as it starting point the concept of a self-organising system that adapts, habituates and learns. It goes on to describe human-machine and human-human interactive as the synergistic composition of self-organising systems into larger wholes (Pask, 1996). CT can be interpreted as a theory of learning and teaching, in which one participant (the teacher) wishes to expound a body of knowledge to a second participant (the learner). Parts of the body of knowledge are referred to as “topics”; the term “concept” is reserved for the mental procedures that indicate understanding of a topic. CT has become fairly widely known in the UK through the writings of Laurillard (1993) and Harri-Augstein and Thomas (1991). However, their accounts are fairly superficial, failing to capture the full depth and richness of CT and its applications (see Burt (2004) for a critique along these lines; see Scott (1993, 2001) for accessible overviews of CT).

The basic CT model is shown in Figure 1. Pask (1996) refers to this model as the “skeleton of a conversation”. It shows a “snapshot” view of two participants in conversation about a topic. The model distinguishes verbal, “provocative” interaction (questions and answers) from behavioural interaction via a shared modelling facility or “micro-world”.

The horizontal connections in Figure 1 represent the verbal exchanges. Pask argues that all such exchanges have, as a minimum, two logical levels. In the figure these are shown as the two levels: “how” and “why”. The “how” level is concerned with descriptions of how to “do” a topic: how to recognise it, construct it, maintain it and so on; the “why” level is concerned with explaining or justifying what a topic means in terms of other topics. These exchanges are “provocative” in that they serve to provoke participants to construct understandings of each other's conceptions and (possibly) misconceptions of topics and the relations between them. This is the essential aspect that makes CT constructivist and dialogical in approach and clearly distinguishes it from other approaches that see teaching as the transmission of knowledge from teacher to learner.

The vertical connections in Figure 1 represent causal connections with feedback, a hierarchy of processes that control or produce other processes. At the lowest level in the control hierarchy there is a canonical world, a “universe of discourse” or “modelling facility” where the teacher (or computer-based surrogate, as in CASTE, below) may instantiate or exemplify the topic by providing non-verbal demonstrations. Typically, such demonstrations are accompanied by expository narrative about “how” and “why”, the provocative interactions of questions and answers referred to above. Note that the form of what constitutes a canonical “world” for construction and demonstration may itself be a topic for negotiation and agreement.

In turn, the learner uses the modelling facility to solve problems and carry out tasks set. He or she may also provide narrative commentary about “how” and “why”. In a computer-based environment these may be elicited using computer aided assessment tools with a variety of different question styles.

Pask (1996) refers to learning about “why” as comprehension learning and learning about “how” as operation learning and conceives them both as being complementary aspects of effective learning. The distinction between “how” and “why” allows for a formal definition of what it means to understand a topic. In CT, understanding a topic means that the learner can “teachback” the topic by providing both non-verbal demonstrations and verbal explanations of “how” and “why”.

3. Some variants of CT

Laurillard (1993), drawing on CT, provides an elaborated account of the exchanges that make up the skeleton of a conversation, interpreted for the kinds of learning conversation that take place in higher education. She distinguishes a domain of exchanges of descriptions, conceptions and misconceptions about both “how” and “why” from a general domain of “tasks”. “Tasks” are interpreted liberally as any learning activity the learner is asked to engage in which generates some product or outcome. In exchange for generating the product, the learner receives feedback about the quality of the product (formative assessment). In turn, the teacher adapts tasks to fit a learner's current level of competence. Throughout, conversation continues with exchanges of conceptions and misconceptions, tasks completed and feedback received engendering topics for further discussion. The Laurillard “conversational framework” has become widely used in the UK, particularly for informing discussions about how best to deploy learning technologies and multimedia.

In some learning contexts, there is a major emphasis on the higher level, generic processes of becoming an effective learner, “critical thinker” or “reflective practitioner” rather than on particular processes of learning in a subject domain (Harri-Augstein and Thomas, 1991; Schon, 1983; Novak and Gowin, 1984; Brockbank and McGill, 1998). Apropos of this, Pask notes that conversations may, by recursive “laddering”, have many logical levels above the basic “how” and “why” levels: levels at which conceptual justifications are themselves justified and where there is “commentary about commentary”. Indeed, reflexively the conversation may itself become a topic of conversation. Harri-Augstein and Thomas (1991) make these notions central in their work on “self-organised learning”. Their emphasis is on helping students “learn-how-to-learn” rather than just on mastery of a particular subject domain.

In brief, Harri-Augstein and Thomas (1991) propose that a “full learning conversation” has three main components:

  1. conversation about the “how” and “why” of a topic, as in the basic CT model;
  2. conversation about the “how” of learning (for example, discussing study skills and reflecting on experiences as a learner); and
  3. conversation about purposes, the “why” of learning, where the emphasis is on encouraging personal autonomy and accepting responsibility for one's own learning.

The model in Figure 2 shows the relationships between the components.

In this brief summary, we have presented CT as a theory of learning and teaching. It helps understand what is supportive of effective learning with respect to a given body of subject matter. It also helps understand what is supportive of empowering learners to become self-organised. CT is particularly useful for understanding how effectively to use learning technologies to support student-centred resource-based learning. This latter theme is now taken forward by reviewing CASTE as an embodiment of CT, where we shall see principles of course design and effective tutorial strategies exemplified.

4. Embodiment of CT in CASTE

CASTE was developed by Pask and Scott (1973) in response to the need to provide learners with a description of a body of subject matter so that there could be “conversation” between a computer-based tutorial system and the learner about learning strategies. Whalley (1995, p. 190), with approval, refers to CASTE as a system that “provided both a ‘virtual’ environment for the student and a system to facilitate learning conversations about it” and “clearly worked as an integrated whole”.

Pask and Scott (1972) had shown that in a “free learning” situation, with no imposed teaching strategy, students typically exhibited a preferred style or approach to their learning. Two main strategies for accessing and using learning materials were observed: a holist strategy, in which students accessed many “topics” (chunks of learning material) in order to build up an overview of the subject matter before attending to specific detail, and a serialist strategy, where students worked in a one-step–at-once manner, learning the details of a particular topic before accessing further topics. In follow-up studies, using teaching materials embodying either a holist or serialist teaching strategy, it was shown that mismatch between a teaching strategy and a student's preferred learning style could lead to little or no effective learning.

CASTE presented subject matter topics in a way that supported holist and serialist learning strategies. Using the conversational features of CASTE, system and learner agreed what was likely to be an effective learning strategy and established an associated “learning contract”. This latter typically included the agreement that progress was contingent on the student successfully “teaching back” what he or she had learned so far. Using these contractual constraints, effective learning to “mastery” level (Block, 1971) was regularly achieved.

The main features of CASTE are shown in Figure 3:.

Transactions were monitored and learners' behaviours modelled by using a suite of tutorial heuristics, written as computer programs. In the original version of CASTE, these were accessed using a time-shared terminal (not in view). The main role of the tutorial heuristics was to specify permitted learning routes, taking into account what was known about a student's current understanding of topics and his or her preferred learning style. When following a holist learning strategy, students were permitted to work on more than one topic at once. For serialist strategies, topics were worked on, to mastery, one at a time.

The basic rules of the system were that:

  1. the learner could only work on a topic, if she had demonstrated that she understood a set of prerequisite, subordinate topics from which the topic in question could be derived (there may be several such sets if there are analogies depicted in the entailment structure); and
  2. having received one or more demonstrations of a topic the learner was constrained, at some stage, to produce a different demonstration to show that she understood the topic (all the transactions at the modelling facility were mechanically detectable and scorable).

A typical transaction sequence would be:

  1. The learner “explores” the subject domain by accessing brief descriptions of topic content and examining the relations between topics shown on the entailment structure.
  2. The learner “aims for” a topic as one she wishes to come to understand. The topic must, of course, be one that she does not already understand.
  3. The learner chooses topics to “work on”. These may be one or several; they must be subordinate to the aimed for topic or be the aimed for topic itself. Furthermore, all topics “worked on” must have sufficient subordinate topics already understood (as in rule 1, above).
  4. The learner requests demonstrations of the topics worked on (this step is optional, the learner is free to omit demonstrations if she feels she can proceed without them): typically, the demonstrations take the form of written text and graphics showing how to set up a model on the modelling facility.
  5. The learner “explains” the topics she has worked on by constructing models for the topics but (as in rule 2) that are not just copies of the demonstrations she has received.
  6. If the (non-verbal) explanation is correct, the topic in question is marked understood. If not, the learner is directed to request further demonstrations.
  7. At some stage, by some route, the learner is led to demonstrate his understanding of the head topic(s) and the tutorial is over.

Overall understanding of “why” is ensured because the learner has accessed topics in a logically coherent sequence, supported by the expository narrative as revealed by the relations between topics shown in the entailment structure. In experimental studies of learning style, explanations of “why” were elicited by asking learners to justify by “teach back” why they chose a particular learning route. In fully automated tutorial mode, this requirement was approximated using multiple choice style questions, delivered on BOSS, asking the learner to select from amongst a set of possible sequences that which they had followed and assign confidence levels to their answers.

Overall understanding of “how” is ensured because the knowledge and task analysis guarantees that in explaining a particular topic by modelling activities, the learner is also demonstrating understanding of “how” for all subordinate topics.

The BOSS was also used to study learners' uncertainties about topic content and about their strategic plans. Clear differences between learners following serialist and holist strategies were observed. A learner following a serialist strategy would typically be confident about the content of a topic selected to be worked on but would exhibit uncertainty about the topic sequence she intended to follow. In contrast, a learner following a holist strategy would not be so confident about the content of topics (usually more than one) selected to be worked on but would be quite confident about the topic sequence she intended to follow.

5. A framework for course design

The framework for course design that we propose is not dissimilar to other approaches such as that proposed by Rowntree (1990). However, its basis in CT provides a coherent theoretical underpinning, three central features of which are:

  1. the assessment of understanding based on “teachback” procedures (first expounded in Pask and Scott (1972)), both for formative and summative assessment purposes;
  2. the use of a sophisticated knowledge and task structure elicitation and representation methodology (Pask et al., 1975; Scott, 1999); and
  3. a well-articulated theory of scaffolding and adaptive teaching (see Lewis and Pask, 1965; Patel et al., 2001.)

The framework has four major components:

  1. a description of course aims and desired learning outcomes;
  2. a specification of course content, describing knowledge and skills and desired learning experiences;
  3. a specification of the learning designs and tutorial strategies to be employed, including sequencing of learning experiences, choice of media and the role of dialogic interactive activities designed to encourage and reinforce effective learning; and
  4. the assessment strategy to be used (for both formative and summative assessment).

The essential principles of good quality course design can be summarised as follows:

These ideas are summarised in Figure 4.

6. Developing UK defence e-learning programmes

Cranfield University supports military colleagues at the UK Defence Academy in the delivery of a wide range of educational courses relevant to the needs of the defence sector. We have been engaged in developing quality web-delivered distance learning courses for the British Army as part of the Review of Officer Career Courses (ROCC) initiative. The ROCC initiative aims to provide ‘life-long learning’ for Army officers up to the point of retirement. Our particular concern has been with developing courses aimed at officers in the early and middle periods of their careers. These are known as Military Knowledge 1 (MK1) and Military Knowledge 2 (MK2) and provide in total some 200 hours of self-directed learning, covering, as the titles suggest, basic knowledge of military doctrine, service functions and organisation, together with coverage of relevant science, technology and project management topics. For logistical reasons, MK2 was the first of the courses, as of December 2004, to go live online. (See Mackain-Bremner and Scott (2006) for more information about these programmes.)

The MK learning materials are a multimedia and interactive activity rich resource. Input from instructional designers has ensured that content is supported by clear statements of learning outcomes, activities that support learning and summaries and formative assessment questions that support self-directed learning and revision. The materials are also designed to be an effective on the job “just-in-time” training and reference tool. The MK course is divided into parts, modules, sections and lessons. Each lesson is further divided into as many as five topics. Topic maps have been defined to aid non-linear navigation through the course content down to the level of the topics within individual 20-40-minute lessons. The maps provide learners with a conceptual view of content at course level (knowledge map) and lesson level (lesson map) and also serve as interactive navigation aids (see Figures 5 and 6). The approach used for knowledge mapping and task and process modelling is directly based on CT methodology as implemented in CASTE.

The ROCC team are interested in tracking progress, but not necessarily formative assessment results, of MK students at the topic level. Although our current implementation is based on the IMS Simple Sequencing Specification, we have designed the product to achieve the required degree of tracking by packaging each lesson at topic level to SCORM Content Aggregation Model 2004 specifications and capturing activity status tracking data as defined by SCORM Run-Time Environment 2004 and SCORM Sequencing and Navigation 2004 (SN). The Harvest Road Hive Learning Content Management System (LCMS) is used for content development; delivery is via the Blackboard virtual learning environment; assessments are delivered using Questionmark Perception.

Key features of the approach to course design include:

7. Concluding comments

This paper presented an approach to learning theory and course design based on first principles. The MK courses use topic maps to support learner navigation and tracking. Adaptive teaching with progress through lesson content contingent on assessment attainments has not yet been implemented. The next step for the Cranfield e-learning research team is to prototype such a system both as a demonstrator of pedagogic principles and also to serve as a CASTE-like sophisticated, flexible interactive learning environments for the study of student learning under different free learning and adaptive pedagogic regimes.

ImageThe “skeleton of a conversation”
Figure 1The “skeleton of a conversation”

ImageA full “learning conversation”
Figure 2A full “learning conversation”

ImageCASTE, main features
Figure 3CASTE, main features

ImageA framework for course design
Figure 4A framework for course design

ImageMK knowledge map
Figure 5MK knowledge map

ImageExample MK lesson map
Figure 6Example MK lesson map

References

(1971), in Block, J.H. (Eds),Mastery Learning: Theory and Practice, Holt, Rinehart & Winston, New York, NY, .

[Manual request] [Infotrieve]

Brockbank, A., McGill, I. (1998), Facilitating Reflective Learning in Higher Education, SRHE and Open University Press, Buckingham, .

[Manual request] [Infotrieve]

Burt, G. (2004), “The conceptual degeneration of sophisticated knowledge: how others have used Gordon Pask's work”, Principled Discourse 13, available at: http://iet.open.ac.uk/pp/g.j.burt/main.htm, .

[Manual request] [Infotrieve]

Harri-Augstein, S., Thomas, L.F. (1991), Learning Conversations, Routledge, London, .

[Manual request] [Infotrieve]

Laurillard, D. (1993), Rethinking Higher Education, Routledge, London, .

[Manual request] [Infotrieve]

Lewis, B.N., Pask, G. (1965), "The theory and practice of adaptive teaching systems", in Glaser, R. (Eds),Teaching Machines and Programmed Learning, Vol. II, Data and Directions, National Education Association, Washington, DC, pp.213-66.

[Manual request] [Infotrieve]

Mackain-Bremner, M., Scott, B. (2006), "E-learning projects at the Defence Academy, Shrivenham", Military Simulation and Training, No.1, pp.243-62.

[Manual request] [Infotrieve]

Novak, J.D., Gowin, D.B. (1984), Learning How to Learn, Cambridge University Press, Cambridge, .

[Manual request] [Infotrieve]

Pask, G. (1996), "Heinz von Foerster's self organisation: the progenitor of conversation and interaction theories", Systems Research, Vol. 13 No.3, pp.349-62.

[Manual request] [Infotrieve]

Pask, G., Scott, B. (1972), "Learning strategies and individual competence", International Journal of Man-Machine Studies, Vol. 4 No.3, pp.217-53.

[Manual request] [Infotrieve]

Pask, G., Scott, B. (1973), "CASTE: a system for exhibiting learning strategies and regulating uncertainty", International Journal of Man-Machine Studies, Vol. 5 pp.17-52.

[Manual request] [Infotrieve]

Pask, G., Kallikourdis, D., Scott, B. (1975), "The representation of knowables", International Journal of Man-Machine Studies, Vol. 7 No.1, pp.15-134.

[Manual request] [Infotrieve]

Patel, A., Scott, B., Kinshuk (2001), "Intelligent tutoring: from SAKI to Byzantium", Kybernetes, Vol. 30 No.5/6, pp.807-18.

[Manual request] [Infotrieve]

Rowntree, D. (1990), Teaching through Self-instruction: How to Develop Open Learning Materials, Kogan Page, London, .

[Manual request] [Infotrieve]

Schon, D. (1983), The Reflective Practitioner: How Professionals Think in Action, Temple Smith, London, .

[Manual request] [Infotrieve]

Scott, B. (1993), "Working with Gordon: developing and applying conversation theory (1968-1978)", Systems Research, No.10, pp.167-82.

[Manual request] [Infotrieve]

Scott, B. (1999), "Knowledge content and narrative structures", in Pemberton, L., Shurville, S. (Eds),Words on the Web: Language Aspects of Computer Mediated Communication, Intellect Books, Exeter, pp.13-24.

[Manual request] [Infotrieve]

Scott, B. (2001), "Conversation theory: a dialogic, constructivist approach to educational technology", Cybernetics and Human Knowing, Vol. 8 No.4, pp.25-46.

[Manual request] [Infotrieve]

Whalley, P. (1995), "Imagining with multimedia", British Journal of Educational Technology, Vol. 26 No.3, pp.190-204.

[Manual request] [Infotrieve]

Corresponding author

Bernard Scott can be contacted at: B.C.E.Scott@cranfield.ac.uk