Skills and Competencies as Representable Metaknowledge for Tele-learning Design
by
Gilbert Paquette
CIRTA-LICEF Research Center
Télé-université, Montréal, Canada
gpaquett@teluq.uquebec.ca
http://www.licef.teluq.uquebec.ca/gp
Abstract
Problems, skills and competency are important interrelated concepts we need to investigate
anew, especially in the context of tele-learning systems’ design. Because they involve generic
skills, competency statements can only be described at the metaknowledge level, more
precisely as generic processes related to domain-specific knowledge. To describe both levels
of knowledge and their interrelation, we have developed a graphical knowledge representation
language and a tool, MOT, central to our instructional engineering methodology (MISA). We
will present here an integrated taxonomy of skills that draws upon three fields of knowledge:
education, software engineering and artificial intelligence. This taxonomy will help us support
some of the main instructional design products in MISA: standardized representable
competency statements, multi-level knowledge models, process-based learning scenarios and,
finally, actors, roles and resources in a telelearning system.
INTRODUCTION
A search on the Internet is sufficient to show the renewed importance given to competency-based learning
and training. Ministries of education, school boards and teacher training institutes present competency
profiles to define school programs or required qualities from the teachers, especially in the use of
technologies in education. Consulting companies present their expertise by enumerating competencies,
marketing their services in this way. Other companies offer services or computerized tools to help their
prospective customers define or manage the competence of their staff, identified as the main asset of an
organization in a knowledge management perspective. Governmental bodies or professional associations
use competency-based approaches to define conditions to the exercise of a profession and to orient their
vocational training programs.
Problems, skills and competency are important interrelated concepts we need to investigate anew,
especially in the context of tele-learning systems’ design. We seek an integrated view of these concepts
for competency-based instructional design, especially in the context of tele-learning. We can achieve this
integrated view by modeling generic skills and specialized knowledge in different byt interrelated
knowledge domains. Theses domains will be labeled respectively as meta-knowledge domain and
application domain.
Meta-knowledge representation
To describe both levels of knowledge and their interrelation, we have developed a graphical knowledge
representation language and a tool, MOT, central to our instructional engineering methodology (MISA).
We have described in Paquette 2000 the main features of the MISA method. Some aspects of the MOT
system have been presented in Paquette 1994, 1999.
Drawing upon three fields of knowledge and corresponding viewpoints, education, software engineering
and artificial intelligence, we will present here an integrated taxonomy of skills that is instrumental for
instructional design. This taxonomy will serve to support some of the main instructional design products
in our instructional design methodology: standardized competency statements, multi-level knowledge
models, process-based learning scenarios and, finally, actors, roles and resources at delivery time in a
telelearning system.
1. PROBLEMS, SKILLS AND METAKNOWLEDGE
Competencies can be defined by associating an actor with the general skills he/she can apply to an
application knowledge model to solve a corresponding class of problems. To solve classification,
diagnosis or construction problems for example, it is necessary to achieve some corresponding task and,
for this, to mobilize corresponding classification, diagnosis or construction skills. A competency
statement linking skills to knowledge in an application domain is defining, at the same time, learning and
teaching goals as well as intellectual processes for the resolution of a class of problems related to the
skills. In this section, we will define skills as active metaknowledge enabling us to represent them in a
knowledge model linked to an application model. Both skills and application knowledge, and their
relationships, thus become objects to be learned.
1.1
Competencies: skills related to knowledge
From the literature on competency profiles, we have hypothesized the following principles that help
situate this concept into the theory of action framework.1 .

The persons whose competencies are described are not simple operators or factors to be evaluated;
they are actors endowed with intentions, situated in a cognitive and social context.

The heart of a competency lies in the association between skills, seen as generic cognitive processes,
and specific knowledge; we seek here to avoid the atomization of competence into the traditional
categories of knowledge, skills and attitudes2.

Competencies are components of a person’s mental model resulting from active meta-knowledge
enacted on specialized knowledge in a application domain, allowing to act on the later in various
ways; this approach integrates the cognitive and meta-cognitive aspects that must be both present for
thoughtful human action and competencies.

Competencies also describe outcomes qualified by the level of excellence of the observed
performance and confirmed by social reward, as in current models for educational objectives.

Competency definitions can thus be used for the assessment of learners as well as developmental
objectives.
To concretize operationaly these ideas, we will now considered three domains offering different
viewpoints on the notion of skills: educational objectives, software engineering and artificial intelligence.
Bélisle et Linard 1996, group under the term « action theory » work in cognitive science by authors such as Vygotsky,
Leontiev, Piaget, Searle and Bruner.
1
2
Following Romiszowski 1981, attitudes are here considered as affective or social skills.
1
Meta-knowledge representation
1.2 Competencies and skills as educational objectives
We can reinterpret the taxonomies of educational objectives in the light of knowledge representation 3.
The taxonomies of objectives in the cognitive domain Bloom 1956 and in the affective domain
Krahtwohl et al 1964 have had a large influence on educational research and practice. From our
viewpoint, they identify intellectual skills such as memorization, understanding, application, analysis,
synthesis, and evaluation and also attitudes and values related to learning.
These authors’ intentions were to define operational training objectives to help to monitor their
acquisition and assessment. Romiszowski [1981] has proposed a definition of skills more in line with
action theory and cognitive science. Skills are « intellectual or physical actions, or even reactions, that a
person produces in a competent manner to reach a goal. To do so, knowledge stored in memory is used
(...). Any skill may be composed of four activities: perception, planning, prerequisite knowledge recall,
and finally, execution of the action (performance) ».4
Another interesting aspect of the classification of skills proposed by Romiszowski is the integration of
cognitive, affective, social and psychomotor skills. Rather than categorizing skills according to the type
of individual response to a stimulus (new knowledge, affective attitudes, social behavior or motor
actions), Romiszowski characterizes them according to their functions in the information processing
cycle. He presents a four-phase cycle of skills with twelve sub-skills. As we can see on table 1, this
approach can be related to the work of Polya 1957 that describes the problem solving process through
phases like “understand the problem”, “generate a solution plan”, “execute the plan”, “assess and
generalize the solution”.
Phase
Skill
Description
Attention
Capability to concentrate on a task
(External stimuli
reception)
Perceptual acuteness
Capability to recognize the stimulus
Discrimination
Capability to recognize the stimulus among other similar
ones
Recall from
memory
(Internal memory
operations)
Interpretation
Knowledge of the stimulus language
Procedure recall
Presence of an adequate algorithm in memory
Schema recall
Presence of relevant concepts and principles in memory
Planning
(Internal
processing
operations)
Analysis
Capability to restructure the problem
Synthesis
Capability to generate alternative solutions
Evaluation
Capability to assess alternate implications
Performance
(External
expression
operations)
Initiation
Capability to make decisions and act accordingly
Continuation
Capability to carry through with action
Control
Capability to self-adapt and self-correct
Perception
Table 1 – An skill’s taxonomy in the field of education Romiszowski 1981
1.3 Skills as generic problem solving processes
The notion of generic problems or tasks was already present in one of the first reference books about
expert systems [Hayes-Roth et al, 1983]; in this work, we find a first classification of generic problems
3
This applies also to the notion of “Learning types” introduced in Reigeluth 1999
4
Romiszowski 1981, page 253
2
Meta-knowledge representation
into ten categories. In other pioneering studies on generic tasks, [Chandrasekaran 1987] describes these
through a problem description and a resolution method, a specific algorithm. It introduces the idea of
combining a small number of generic methods to solve large classes of more complex problems. Other
work on generic problems [McDermott 1988], and the « components of expertise » approach [Steels
1990] must also be mentioned.
The KADS method [Scheiber et al, 1993; Breuker and Van de Velde, 1994] is a synthesis of these studies.
It actually constitutes one of the most complete methodologies encompassing knowledge acquisition for
expert systems, but also project management, organizational analysis and software engineering. In
KADS, an engineering software project materializes by building seven models. Four of them are of
interest here: the « domain model », the « inference model », the « task model » and the « strategic
model ».
In the « inference model », we find a decomposition of the generic task in a task tree. Inference diagrams
are associated to the leaves of the task tree. The « task model » provides algorithmic control principles,
rules to manage the tasks tree. The « strategic model », hardly developed in KADS, corresponds to
heuristic principles that guide tasks execution. Together, these three models correspond to the notion of
generic process, applicable to various application « domain model », another of the seven KADS models.
A generic problem is characterized by one or several goals or products to achieve; initial data and a
number of operations leading to the transformation of data into results or goals. The KADS method
defines eight classes of generic problems, presented in table 2.
Generic task
Generic problem
Input (data)
Products (goal)
Classify
Determine an object’s
category
Classes hierarchy; Object’s Classes containing the object
attributes
Diagnose
Determine the cause of
the problem
Symptoms, system’s
component model
Defective components
Predict
Determine the future
state of the system
System’s components;
attributes that will vary
States: classes of the system’s
possible instances
Supervise
Determine a deviation
class between a system’s
instance and another
which is said to be
normal
System’s components and Classes grouping instances
attributes; normal instances according to the difference with
the norm
Repair
Modify a system’s
component so it is in
working order
System’s model;
maintenance standards
Modified model
Plan
Break down the task into
interrelated steps
Deliverables, sub-tasks,
time constraints
Process: sequence of tasks,
input and output
Design
Build an object (artifact)
Artifact properties,
constraints to be met
Object model
Model
Build a behavioral
system model
Goals, constraints,
components, viewpoints.
Model of a system’s processes
and evolution strategies
Table 2 – Generic problems and tasks in software and cognitive engineering
To each generic problem corresponds a generic task, which is a generic procedure with input and results
as indicated in table 2. Breaking down this generic procedure into sub-procedures results in the KADS
method tasks tree. After a number of levels, terminal-level tasks (the « leaves ») are reached, to which the
KADS method associates an inference schema that completes the « task model ».
3
Meta-knowledge representation
Similarly, a skill can be broken down into sub-tasks, which are other skills. Each skill also has its input
and products, which is a type of knowledge resulting from applying the skill. For instance, «diagnose a
component system » has an input, a component system (a “part-of” hierarchy), and generates a list of
faulty components. Such a process is generic in the sense that it applies to types of knowledge in many
domains, rather than to knowledge in a particular domain. Also, these skills may be compared with one
another through specialization links: « diagnose a car breakdown » is a kind of diagnosis skill. In sum,
skills can be represented as generic or meta-processes.
1.4 Skills as active metaknowledge
Skills, generic problems and tasks can only be interpreted in the meta-knowledge domain. Although
many studies involve meta-knowledge, the term is not always explicitly used. These studies can be
found in various domains such as mathematical logic [Thayse 1988], scientific methodology [Popper
1961], problem sovling and its teaching [Polya 1967], education [Romisowski 1987, Merrill 1994],
software and cognitive engineering [Chandrasekaran 1987, Schreiber et al 1993], artificial intelligence
[Pitrat 1991].
Pitrat, a pioneer of Artificial Intelligence in France, has produced an important synthesis in which he
distinguishes several meta-knowledge categories and proposes the following definition: « metaknowledge is knowledge about knowledge, rather than knowledge from a specific domain such as
mathematics, medicine or geology. » [Pitrat 1993]. According to this definition, meta-knowledge is at the
heart of the learning process, which consists in transforming information into knowledge:

by associating values to application knowledge such as : truth, usefulness, importance, knowledge
priority, competence of an individual towards a knowledge object, etc.

by describing « intellectual acts » or cognitive processes facilitating knowledge processing in
application domains: memorization, application, analysis, synthesis, evaluation, etc.

by representing strategies to acquire, process and use knowledge: memorization techniques,
heuristic principles for problem solving, project management strategies, etc.
[Romisowki 1981] expresses very well the simultaneous phenomenon of knowledge acquisition in a
particular domain, and the building of meta-knowledge and skills : « The learner follows two kinds of
objectives at the same time - learning specific new knowledge and learning to better analyze what he
already knows, to restructure knowledge, to validate new ideas and formulate new knowledge », an idea
expressed in another way by Pitrat : « meta-knowledge is being created at the same time as knowledge ».
When learning new knowledge, a person uses meta-knowledge (at least minimally) without necessarily
being aware of it. However, using meta-knowledge should really be a learner’s conscious act. These
objectives justify the inclusion of meta-knowledge within a knowledge model that represents a learning
system’s contents, i.e. that provides a structured representation of « learning objects».
Jacques Pitrat define the notion of active meta-knowledge, as opposed to passive meta-knowledge
(knowledge properties, knowledge about an individual’s knowledge). Active meta-knowledge is
knowledge that “physically” handles other knowledge. Pitrat 1990] defines six types of active metaknowledge:

Knowledge acquisition consists in examining and diagnosing available information and knowledge,
completing it, if incomplete or inconsistent with other knowledge previously acquired, and
reformulating it, as needed, so it may be stored in memory.

Knowledge discovery regroups a set of operations like instanciation, specialization or analogy, which
allow transformation of acquired knowledge into new knowledge.
4
Meta-knowledge representation

Knowledge storage consists in deciding where and how to register knowledge in a structured way in
memory so it can become quickly available when needed, following the shortest association chains,
without having to systematically scan memory.

Knowledge search is essentially a set of knowledge reconstruction operations to extract from memory
the knowledge needed to solve a problem or accomplish a task.

Knowledge use regroups a set of operations required to apply knowledge that has been extracted or
reconstructed from memory, in order to build a solution for a problem, designing and managing
solution plans and results explanation.

Knowledge expression is the inverse of acquisition, to communicate acquired knowledge to another
information processing system, generally a human being; these operations enable a person to choose
what to say and how to say it, according to a model of the intended receiver.
In form ation s
I/P
I/P
Acqui re
I/P
C onstructe d
kn owle dge
Discove r
I/P
I/P
I/P
Evaluate d
kn owle dge
Store
I/P
I/P
I/P
Me ta-knowle dge
- Active : me ta-proce sse s
Acce ss proce ss
- Passive : knowle dge prope rtie s
and kn owl edge on the kn owl edge
I/P
of se lf an d othe rs
I/P
Knowl e dge from the
applicati on domain
I/P
I/P
Se arch
Extracte d
kn owle dge
I/P
I/P
I/P
Use
Expre ss
I/P
I/P
C ommu nicate d
kn owle dge
Applie d
kn owle dge
Figure 1 – Relations between active meta-knowledge objects
Figure 1 shows relations between different kinds of active meta-knowledge. Discovery is the only metaknowledge that produces new knowledge from raw data or structured information. Acquisition is the
meta-knowledge that allows integration of self-discovered or externally provided knowledge. The
resulting knowledge is structured, reorganized and assessed, assigning meta-concepts values such as
validity and interest. Storage meta-knowledge integrates new knowledge from the particular application
domain together with the associated meta-values, as well as the active meta-knowledge facilitating
subsequent knowledge search. These operations increase the knowledge base available to a cognitive
system. Reverse operations may then be used for memory search, and, later on, to express and use
knowledge.
5
Meta-knowledge representation
2. REPRESENTING A SKILL AS METAKNOWLEDGE
The goal of this article is to show that skills and competencies need to be described precisely at the metaknowledge level. We will summarize here the MOT graphical representation language that we will use as
a tool to describe skills and competencies in relation to an application domain.
2.1 The MOT representation system
A basic MOT model is composed of six types of knowledge and seven types of links between them.
Knowledge is represented by geometric figures that identify its type. We distinguish abstract knowledge
(concepts, procedures, principles) from their corresponding sets of facts (examples, traces, statements).
Relations between these entities are represented by oriented links with symbol (C, S, P, I/P, R, I and A)
representing the type of relation.

The instance (I) link relates an abstract knowledge to a group of facts obtained by giving values to all
the attributes (variables); its origin is a concept, a procedure or a principle, and its destination are,
respectively, examples, traces or statements.

The composition link (C) connects a knowledge unit to one of its components or parts. Any object’s
attributes may be specified as component of the object.

The specialization link (S) connects one abstract knowledge object to a second one that is more
general than the first one.

The precedence link (P) connects two procedures or principles, where the first must be terminated or
evaluated before the second one can begin or be applied.

The input-product link (I/P) connects a concept to a procedure, the concept being the input of the
procedure, or a procedure to a concept that is the product of the procedure.

The regulation link (R) is directed from a principle towards a concept, a procedure or another
principle. In the first case, the principle defines the concept by specifying definition or integrity
constraints or it establishes a law or relation between two or more concepts. On the other hand, a
regulation link, from a principle to a procedure or another principle means that the principle exerts
external control on the execution of a procedure or the selection of other principles.

The application link (A) is used to associate an object defined as meta-knowledge in another domain,
to knowledge in an application domain. We will show that this link is central to the representation of
competencies and their relation to skills and application knowledge.
2.2 Representation of meta-knowledge
Every domain such as physics, sociology or law is made of knowledge and facts. The domain that studies
knowledge per se is particularly important for learning. Knowledge from this domain will be called metaknowledge or generic knowledge. Since meta-knowledge is also knowledge (about knowledge), we can
also distinguish between three categories of abstract meta-knowledge and their corresponding meta-facts,
as we do for knowledge.

Meta-concepts represent knowledge attributes. They are concepts that define value systems to apply
to knowledge from various domains.
For instance, when one claims that some knowledge in physics or economics is a priority, is valid or
is useful, he uses a concept of priority, validity or usefulness that does not belong to physics nor
economics, but to the domain that studies knowledge. Instanciating such a meta-concept to
knowledge in an application domain results in the assignment of a meta-value to the knowledge we
want to talk about. As a result, we get meta-examples in various domains such as: « the concept of
atom is essential », « the break-even point calculation procedure is useful in micro-economics » .
6
Meta-knowledge representation

Meta-procedures are operations on knowledge. They are actions on knowledge or facts in various
domains where they are applied.
Classification, defined as a set of operations to determine the smallest class of a taxonomy to which a
particular object belongs, is an example of a meta-procedure. It is composed of operations intended to
determine of what class the object is an example; we first consider the first-level in the taxonomy,
then we examine the second-level sub-classes and so on, up to the terminal classes of the taxonomy.
To instanciate such a meta-procedure consists in choosing the taxonomy and the object we want to
classify within the taxonomy, for example, a taxonomy of vertebrates and a bat, or a taxonomy of
professions and a given individual. The result is a meta-trace of specific operations in the application
domain: « the bat satisfies the definition of vertebrate, of mammal, of cheiropter » or « the individual
satisfies the definition of the members of a liberal profession, of law professionals, of lawers ».

Meta-principles are generic statements that apply to various domains aiming to control the use of
other meta-knowledge objects or to establish relations between them. Depending if they are action or
relational principles, they can be described as « knowledge control principles » or « knowledge
association principles ».
One example is the following principle: “ to solve a complex problem, first solve a particular case of
it”. Instanciating such a meta-principle is in fact choosing knowledge, here a problem, to which the
meta-principle will be applied. Results are meta-statements like « to build a general procedure for
compound interest calculation, first solve the problem using an interest rate of 10% », or « To
diagnose a car breakdown, first diagnose the state of the electrical system ».
Figure 2 shows how we assign a generic process (a skill) such as simulation to a search procedure in the
Internet domain, using an A link from the generic process, instanciated in the Internet application domain,
to a procedure in this domain. We also use another A link to assign a meta-value to a knowledge, tagging
it as very important knowledge.
Figure 2 – A simulation meta-process applied (link A) to the Internet domain.
In the application domain of figure 2, the main knowledge unit is a procedure. It defines the main purpose
of a possible learning unit “Search for information on the Internet”. This procedure is decomposed into
sub-procedures using C links. One of them is “Execute the request”: it has a “Request” concept as input
and produces (through an I/P link) a list of “Interesting Web sites”. These are in turn used as inputs to
another sub-procedure, “Identify interesting information”, which precedes the final procedure: “Transfer
7
Meta-knowledge representation
information in a text editor”. The “Refine the request” sub-procedure is regulated (R link) by principles
helping a user to refine a request.
The application (A) link on the main procedure shows that the learner will have to apply a generic skill:
«Simulate a process» to the Internet domain. Figure 3 describes a graphic model for that meta-knowledge.
The MOT graph provides a precise definition of the “Simulate a process” skill, adding to it more details
such as inputs and outputs of sub-processes. If needed, we could add more details to the model by adding
some of the control principles that heuristically could help achieve the generic task. The principles would
have to be stated in sub-models of the process. Also, procedures and concepts in the model could be
described with more detail.
Notice that the instance “Simulate a process in the Internet domain” appears on both figures, linking the
two models by a co-reference link (a meta-link embedded in the MOT system). On figure 2 it is a metafact from the a viewpoint in the Internet domain because it is defined in a meta-knowledge domain. On
figure 3, it is shown as an instance (a trace) of the meta-process “Simulate a process” which gives a
precise and inspectable definition of the skill.
Figure 3 – Graph of a meta-process: “Simulate a process”5
3. AN INTEGRATED TAXONOMY OF SKILLS
We will now use the knowledge representation technique just outlined to define an integrated taxonomy
of skills. We first present a comparison between the taxonomies presented in section one and then discuss
the properties of this taxonomy.
3.1 A general to specific taxonomy
We have built a taxonomy of meta-processes representing skills, as well as problems and tasks, for our
learning system engineering method MISA [Paquette 1999]. Table 3 presents the first three layers of this
5
The numbers in the figure refer to the skills taxonomy presented in the next section .
8
Meta-knowledge representation
taxonomy and compares it to the taxonomies presented in section 1. Although these taxonomies have
different purposes and terminologies, they roughly correspond.
A first layer of skills (or meta-processes) in our taxonomy corresponds to four general processes
generally agreed upon as representing basic information processing phases. The second layer includes ten
generic processes that can be ordered from simple to complex, as we will see later on. Third layer skills
correspond to more specialized skills that are widely used in instructional design.
We can of course extend this specialization hierarchy to more layers. From layer to layer, we get more
and mode specialized skills until every aspect of a skill is totally instanciated in a particular application
domain such as in the following chain, from general to specific: Reproduce – Analyze – Diagnose –
Diagnose a health problem – Diagnose a heart problem in a child.
Skills taxonomy layers
Creation
Reproduction
Reception
1
2
3
Generic
problems
(KADS)
Cognitive
objectives
(Bloom)
1- Pay Attention
2- Integrate
2.1 Identify
2.2 Memorize
3Instantiate/
Specify
3.1 Illustrate
3.2 Discriminate
3.3 Explicitate
Memorize
Skills cycle
(Romiszowski)
Attention,
Perceptual acuity,
Perceptual
discrimination
Knowledge search
and storage
Understand
Interpretation
Knowledge use,
Knowledge
expression
Apply
Recall procedures
Recall schemata
4- Transpose/ Translate
5- Apply
5.1 Use
5.2 Simulate
6- Analyze
6.1 Deduce
6.2 Classify
6.3 Predict
6.4 Diagnose
Knowledge
discovery
Prediction,
Supervision,
Classification,
Diagnosis
Analyze
Restructure
Generate
alternatives
Repair
7- Repair
8- Synthesize
Selfmanagement
Active metaknowledge
(Pitrat
Planning, Design,
Modeling
8.1 Induce
8.2 Plan
8.3 Model/
Construct
Knowledge
acquisition
9- Evaluate
10Self- manage
Synthesize
Evaluate
10.1 Influence
10.2 Self-control
Think of
implications, act
on a decision, see
through the
action, selfcorrect
Table 3 - Comparative multi-layered taxonomy of skills, problems and metaknowledge
According to this taxonomy, “5.2 - Simulate a process”, a skill used in the example on figure 2 and 3, is a
specialization of the “Apply” skill. Each of the skills in this taxonomy is described very precisely by its
inputs and its products and by a detailed generic process similar to figure 3 showing how the inputs are
transformed into specific products. The “Simulate a process” skill is compared below to the “8.3
Construct a process” skill which is a specialization of the second layer “Synthesize” skill.
9
Meta-knowledge representation
Skill
Input
Product
Simulate a
process
A process, its
procedures, inputs,
products and control
principles.
A trace of the procedure: set of
facts obtained through the
application of the procedures in
a particular case
Construct a
process
Generic process
-Choose input objects
-Select the first procedure to execute
-Execute it and produce a first result
-Select a next procedure and execute it
-Use the control principles to control the flow of
execution
Definition
A description of the process: its -Give a name to the procedure to be constructed
constraints to be
inputs, products, sub-procedures -Relate it to specified input and product
satisfied such as
with their input and output, and
-Decompose the procedure
certain inputs,
control principles.
-Continue to a point where well understood steps
products and/or steps
are attained.
Table 4 – Examples of two skills as meta-processes
From the description of the two generic skills on table 4, we can see that a pedagogical scenario on the
same subject of “Information search on the Internet” but with a different skill objective such as
“Construct a process” would be very different from the one based on the “Simulate a process” skill. In the
first case, a kind of walk-through of the process is sufficient, while in the second case, we could need a
project-based scenario where learners and engaged in a more complex problem-solving activity.
3.3 Ordering skills from simple to complex
The question whether skills are ordered from simple to complex is not a simple one. Between, The
definitions presented in table 5 supports this hypothesis for skills in the first layer: reception skills
involve only attention and memory operations, reproduction skills are essentially instantiation from more
general knowledge, creation skills produce new knowledge from more specialized ones and, finally, self
management skills involves explicit meta-cognitive operations.
Name of skill
Definition
Examples

Receive
Input = internal or external stimulus;
Reproduce
Product = facts or knowledge located or

stored in memory

Input = knowledge and models;

Produce/
Create
Self-manage
Products = facts obtained through

instancing or knowledge obtained

through reformulation
Pay attention to an event, to a movement, to an emotion, to a
social context;
Identify knowledge, associated impressions;
Memorize knowledge, impressions.
Use examples to explain or illustrate a concept, a procedure
or a principle;
Use a model to explain facts;
Simulate a process.


Products = new knowledge or models

resulting from analysis and synthesis

Input = knowledge, models, generic 
facts;

Product = knowledge, models, meta- 
knowledge linked to domain model
Classify objects according to a taxonomy;
Repair defective system components ;
Plan a project;
Model and build a system.
Assess knowledge validity or self competence;
Initiate a change process after assessing the situation;
Apply a generic strategy to improve learning and
performance.
Input = knowledge and models;
Table 5 – Comparison of the more general skills, from simple to complex
On the other hand, the skills on the third layer of the taxonomy are probably not ordered from simple to
complex. For example, the integrate sub-skills are simply inverse operations for simple retrieval and
storage, while the four analyze sub-skills or the three synthesize sub-skills on table 3 are much on the
same level of complexity.
So it seems that as we move from general to more specific layers, the skills in a layer are less likely to be
ordered form simple to complex. On table 3, we have assigned numbers to the second layer. We have now
to show evidence that this layer can be ordered from simple to complex.
10
Meta-knowledge representation
This assertion is not evident and was sometimes disputed in the case of taxonomies presented in section 1.
For example, the authors of the KADS method have preferred to put emphasis on the organization of
sequences of generic tasks than of a hierarchical order among them6.
On the other hand, Bloom has insisted on the hierarchical organization between educational outcomes:
“Our attempt to order the educational behavior from simple the complex is based on the idea that a given
simple behavior can become integrated with another simple behavior to form a more complex behavior.
Consequently, our classification can be perceived as that behavior of type A forms a class, behavior of
type AB another class and behavior of type ABC still another class”. One finds a similar preoccupation in
the elaboration of the taxonomy of the affective domain. “This organization of constituents seems to
describe a process according to which certain phenomenon or value progress from one level of simple
awareness to a level where it drives or controls the behavior of a person.”7
Experimental studies have tried to verify this hypothesis. Tests have been given in a large number of
students containing questions connected to various complexity levels in both taxonomies. With this
experimental setting, one should notice a bigger percentage of failure for questions related to the higher
taxonomy levels.
As far as the taxonomy of the cognitive domain, according to Martin and Briggs8, some studies support to
a certain extent the hypothesis of the organization of levels. The evidence is stronger in the first levels
than in the more advanced levels. One finds the same kind of conclusions in the case of the taxonomy of
the emotional domain, even though there are less studies to support this.
We suspect that the limited evidence coming from certain studies is due to the absence of a metaknowledge representation scheme for skills. For example, certain studies note similar results for analysis
on one hand, and for evaluation and the synthesis on the other hand. But if one distinguishes synthesis
from analysis by the ascent in abstraction, and if one distinguishes evaluation both from synthesis and
analysis by the use of meta-values that are properties of knowledge, it is likely that one can maintain the
hypothesis of an increasing order of complexity for Bloom’s taxonomy as well as for the second layer of
our taxonomy.
We will give here a clear definition of a skill’s complexity: A skill A is more complex than a skill B if the
generic process representing B appears as a sub-process in the model of the generic process A.
Figure 3 gives an example of this definition. The “Simulate a process” skill (level of complexity 5) is
decomposed into four sub-processes such as: produce examples of the input concept (instantiate: level 3),
identify the next applicable procedure (identify: level 2), assemble the simulation trace (transpose: level
4), and finally execute the procedure using its execution principles, a specialization of the application skill
(same level as simulation).
Figure 4 present another example of simple to complex ordering within the second layer of the taxonomy.
A generic process for self-control and adaptation (10.2) is here presented. It starts by obtaining a
description of a project or a process of change in a particular domain, as well as some criteria for success
appropriate for this domain. At the beginning, the actor plans (8.2) the activities in a project and
influences (10.1) the participants so that they coordinate themselves to achieve the activities of the project
and perform according to the success criteria.
Afterwards, progress is constantly estimated and re-evaluated. If unforeseen problems arise, it is
necessary for the actor to adapt himself to re-order the course of events, to redefine the roles of the
participants or to adapt the criteria of success. The principles of control and adaptation rule the transition
6
Breuker and Van de Velde 1994 op cit, pp.57-61
7
Bloom 1956, op. cit. p. 18 and Krathwohl et al 1964, op. cit. p. 27
8
Martin et Briggs 1986 op. cit. pp. 69-71 et pp. 79-81
11
Meta-knowledge representation
between sub-processes, for example by specifying when one has to estimate progress and success or when
one has to adjust the criteria or end the generic process.
Periodically, the actor evaluates (9) the distance from the goal with regard to the success criteria available
at a given moment. If the group is far from the goal, the actor begins to reorganize so as to increase the
chances of success. He can also modify the criteria of success.
This representation of a level 10 skills shows that it is more complex that the level 8 or 9 skills that are
invoked as sub-processes.
Control and
adaptation
principles
Success criteria
Project definition:
object, events,
participants, goals
I/P
C
I/P
A
(10.2)
Control and
adapt a project
S
R
C
C
(8.2) Pl an
the e ve nts
C
C
C
Goal
achi e ve d
C
Evaluation
principles
S
(10.1)
Infl ue nce the
parti cipants
R
C
P
C
C
C
A
A
(6.4) Vé rify i f
goal
achi e ve d
I/P
Ne w crite ri a
P
(9) Eval u ate
progre ss
Unfore se e n
e ve n ts
R
P
P
C
I/P
Adapt to
un fore se e n
e ve n ts
P
A
S
Criteria
adjustment
principles
I/P
Goal sti l l
far away
I/P
End the
proje ct
P
P
R
C
Productions
achieved
I/P
(7) Ajust
succe ss
cri te ria
(7) Pe rse ve re
and adapt
course of acti on
Figure 4 – A meta-process for control and adaptation
3.4 Integrating cognitive, affective, social and psycho-motor domains
The generic process on figure 4 shows a strong emphasis on affective and social skills. For example, “To
persevere and to adapt the course of action” is a sub-process where affective and social skills are deeply
involved. We can indicate this by labels C, P, A and S representing a cognitive, psycho-motor, affective
or social skills. More over this generic process can be applied in domains as different as a research project
(cognitive skills), a psychosomatic therapy (change of the emotional attitudes), organizational change
(social skill) or sports training (psychomotor skill).
This kind of integration is a fundamental feature of our skills taxonomy. Figure 5 shows another example
where a diagnosis meta-process model is represented together with two very different instanciations in a
cognitive and an affective application domain.
At the metaknowledge level, the diagnosis process takes a component model of a situation as input and
returns a list or faulty element to correct. First, a sub-model is selected to focus on a probably faulty submodel. Then this model is decomposed to generate hypothesis or faulty candidates. Each hypothesis is
12
Meta-knowledge representation
tested with attributes compared to some norms. If the corresponding component is faulty, it is added to
the list; if not, a new hypothesis must be generated and tested.
The only difference between the two applications is the nature of the input and output of the diagnosis
process.

In the first case, it is applied to the model of a hardware system with components being pieces of
equipment down to very small part that can be deficient; the output is a list of faulty parts.

In the second case, it is applied to an affective situation where the components are facts and opinions
that people have expressed on a certain event that has caused guilt to occur in a person; the results can
be acts that should not have been made, or opinions that are not supported by facts or are clearly
exaggerated.
Figure 5 – A diagnostic skill applied to a cognitive and an affective situation
This example shows us that skills in our taxonomy are described mainly by their functions in the process
by which a person perceives and transforms knowledge, acts, reacts or interact in a given situation, rather
than according to the type of stimulus or response: cognitive acts, motor actions, affective or social
attitudes.
Contrary to their traditional use in the definition of the educational objectives, skills are here learning
goals, (meta-) knowledge that one can represent, analyze or evaluate in itself or in relation to knowledge
of various domains. Skills as meta-knowledge must be present in domain knowledge (as meta-traces) if
we want to be able to propose them as targets of learning activities, in the same way as the knowledge in a
specific domain.
It may seem ambitious to propose a taxonomy integrating the cognitive, psychomotor, emotional and
social domains, while so many practitioners in education use taxonomies of skills separated for each of
13
Meta-knowledge representation
these meta-domains. We believe on the contrary that in instructional engineering, it is important to
integrate them. As underlined by Martin and Briggs: "This subdivision is relatively arbitrary because the
psychologists and the educators agree that, in the reality of educational practice, no real separation
between the cognitive, emotional and psychomotor states is possible "9. Martin and Briggs quote in
support to this assertion several other authors, notably some having produced important taxonomies such
as [Bloom 1975] and [Gagné 1970].
Although recent developments in neuro-physiology suggest that regions of the brain are specialized in
cognition, emotions or psychomotor commands, research in this domain shows evidence of an integration
between the various constituents of the brain in each of our activities. As an example, Daniel Goleman
underlines that "our emotional faculties drive us constantly in our choices; they work of concert with the
rational spirit and allow - or forbid - the exercise of the very thought processes. Also, the cognitive brain
plays an executive role in our feelings." 10
In Paquette et al 2000 we have built a complete table showing examples in the cognitive, affective,
social and psycho-motor meta-domains for each of the 10 major skills on the second layer of the
taxonomy. It shows that this taxonomy can be interpreted in each of the four meta-domains (cognitive,
psycho-motor, affective or social). For example, we can repair theories and movements, as well as
attitudes or social relations. What differentiate these four meta-domains is essentially the type of input to
a skill and its resulting production. If the stimuli or the result concerns rational thought, motor capacities,
affectivity or social interactions, we will label the skill to be cognitive, psychomotor, affective or social.
More generally, we could say that somebody is "intelligent" on the rational, physical, emotional or social
dimension if it he or she is capable of applying in most of cases, all types of skills for that dimension.
This is basically what the American psychologist Howard Gardner suggests by taking into account
multiple intelligences as the basic conceptual structure of the intellect.11
4. USE OF THE TAXONOMY FOR INSTRUCTIONAL ENGINEERING
We will now briefly outline some applications to instructional engineering of the skills taxonomy
presented above. We will first present a way to standardize the interpretation of competency profiles, our
starting point in this paper. Then, we will use a meta-knowledge approach to define learning needs and
help focus content definition. We will use a skills representation as a basis for a learning scenario and to
identify the roles and resources of different actors in a tele-learning system Paquette 2001.
4.1 Giving an operational meaning to competency profiles
We present here a process to analyze a group of existing of competency statements or to build them in a
standard way. This process contains following sub-processes:

determine one or several target actors for which the competency is defined;

identify the tasks of these actors, as well as the corresponding knowledge and represent them in a
knowledge model;

identify the skills required by the actors to be applied to the knowledge.
9
Martin et Briggs 1986, p. 10
10
Goleman 1997 op cit, p.53
11
Gardner 1993
14
Meta-knowledge representation
To illustrate this process, we will use as a case study a competency profile of a multimedia director
resulting from a general analysis of the domain of multimedia trades having lead to fourteen actor
definitions and their corresponding competency profiles12.
Figure 6 – A multimedia competency profile representation and raw data analysis results
Figure 6 shows part of a competency model we have built in MOT+, each individual competency
statement being defined by an its actor, skill and knowledge components. Here the statement would read
“A multimedia director must be able to elaborate a production method for a multimedia project”. There
are of course other competency statements like this in the model. Furthermore, the model is linked to
description of qualities required for the job and a task description that will help identify sub-skills for this
job profile.
A knowledge model for this domain is essentially procedural based on the tasks definition, but it can be
completed by concepts and principles required to achieve the tasks. In this model, three main processes
must be governed by the multimedia director: the elaboration of the method of production, the
management of the creation / production process and the quality control of the product. The first of this
process is presented on figure 7.
The central task of this model is to elaborate a method of production and a definition of the project from
expectations and requirements of the customer. Besides this knowledge, it requires from the multimedia
director the use of knowledge on the technical and graphic feasibility of the project, on the production
steps of a multimedia project, on the possible audio-visual support, on the tools for multimedia creation
on PC, MAC AND UNIX, on various approaches to develop, implement and deliver a Web site and
finally, on the use, potential and limits of the Internet and multimedia technologies .
To each kind of knowledge, we can associate a skill selected from our taxonomy. For example, for the
production method, a suited skill is " to construct", for the definition of the project, "to plan ", and so on.
12
These competency profiles can be consulted at http://www.technocompetences.qc.ca/html/frame_rech_etud.html
15
Meta-knowledge representation
Technical and
A
graphic feasability
Évaluate
T rac e
A
Client needs
and
expectations
Plan
T rac e
A
T rac e
Simulate
A
Production steps
IP
IP
Project
def inition
IP
T rac e
Discriminate
A
Audio-visual
support
IP
Élaboration
procedure
IP
IP
Production
method
A
A
IP
T rac e
Explicitate
A
Creation tools on
PC, MAC, UNIX
Approaches to
develop,
A
implement and
deliver a web site
T rac e
IP
IP
T rac e
A
Transpose
Construct
Uses, potential and
limits of multimedia
and Internet
technology
Figure 7 – Sub-skills identification for a main competency statement for a multimedia director
Table 6 presents the overall result of this competency analysis process. It shows the initial competency
statement we started from, the reformulation that results from our analysis, and the decomposition of each
statement in a standard interpretation. This interpretation now has a precise meaning given by the
definition and the meta-models in the skills taxonomy. These can be used for instructional design or other
purpose. Also, the process has revealed missing or ill-defined competency statements.
Competency statement
Initial formulation
Reformulation
Interpretation
Skill
Type
Parent
skill
Knowledge
component
Discovered through the
process
Model a production method
Model
C
8Synthesize
Production method
Discovered through the
process
Plan a project definition
Plan
C
8Synthesize
Project definition
Capacity to evaluate the
technical and graphical
feasibility of a project
Evaluate the technical and
graphical feasibility of a
project
Évaluate
C
9Évaluate
Technical and
graphical feasibility of
a project
Knowledge of each
production step
Simulate a production process
Simulate
C
5- Apply
Production steps
Knowledge of audio-visual
support
Discriminate between
properties of audio-visual
support
Discriminate
C
3Instanciate
Properties of audiovisual support
Knowledge of the use,
potential and limits of
Internet and multimedia
technologies
Transpose in the project the
use, potential and limits of
Internet and multimedia
technologies
Transpose
C
4Transpose
Use, potential and
limits of Internet and
MM technologies
Superficial knowledge of
multimedia creation tools
(PC/Mac/Unix)
Explicitate the main
properties of multimedia
creation tools (PC/Mac/Unix)
Explicitate
C
3Instanciate
Main properties of
multimedia creation
tools (PC/Mac/Unix)
Knowledge on approaches
to develop, implement and
deliver a web site
Transpose to the project a
useful way to develop,
implement and deliver a web
site
Transpose in
a useful way
C
4Transpose
Approaches to
develop, implement
and deliver a web site
Table 6 – Standard interpretation of competency statements from a meta-knowledge point of view
16
Meta-knowledge representation
4.2 Focusing knowledge modeling for content definition
We will now illustrate a method to guide the modeling of a knowledge domain..
For this, we need the concept of learning need. The identification of learning needs is an important
element of the initial phases of analysis and design of a learning system. Learning need is a meta-concept
applicable in all the domains of knowledge and it can be defined precisely in the meta-knowledge
domain. 13 It rests on an evaluation of the distance between the current competency of a learner or a group
of learners and a target competency to be achieved at the end of the learning process.
Target competencies are found in competency profiles as discussed above. So, if one wants to propose a
plan for training persons to become multimedia directors, it is necessary to estimate the distance between
its current competencies and the ones defined in a typical profile for a multimedia director.
To estimate this distance, we can use a scale to evaluate the progress in the acquisition of a competency.
In MISA, we use such a scale to associate to every competence of an actor (skill + knowledge) a value
between 0 and 10 subdivided into four levels: sensitization [from 0 to 2,5], familiarization (from 2,5 to 5],
mastery (from 5 to 7,5] and expertise (from 7,5 to 10].
Competency scale– For each target population
State :
Value : 0
Sensitization
2,5
Familiarization
Mastery
5
Expertise
7,5
10
A : Construct a production method
B : Discriminate between audio-visual supports
C : Simulate a production process
D : Plan a project definition
Figure 8 – Learning needs for an actor (multimedia director) on a competency scale
Figure 8 presents a progress scale for the competencies of an actor, for example a multimedia director.
Competence A, " To build a production method ", presents a rather important gap (around 4,2), because
the multimedia director has to master this competence (5,8), while the persons for whom training is
intended have reached on average a simple sensitization level (1,6). On the other hand, competency B, "
To discriminate between the properties of audio-visual aids ", presents a smaller learning need of about
1.6, representing a progress from one level of familiarization to the other.
MISA proposes to guide knowledge modeling in a domain by means of elaboration principles similar to
elaboration theory Reigeluth 1983. We develop a model by successive levels taking into account the
learning needs assigned to so-called "principal knowledge". A principal knowledge is one to which a skill
and a competency have been assigned in a model, thus enabling the evaluation of learning needs. We then
apply the following heuristic principles.
A knowledge for which the learning need is large (for at least one group of learners) will deserve to be
clarified by indicating some of its components, inputs and products, and its regulating principles. It will
be developed into a sub-model on possibly many levels until we reach a state where the learning needs are
small for each knowledge unit (no new knowledge can become principal). On the contrary, if the learning
need is very small for a principal knowledge, it can be removed from the model unless it serves a
clarification purpose towards other knowledge units around it.
13
Such a meta-concept has been implemented in different ways, first in the didactic engineering workbench (AGD) [Paquette et
al 1994], and then into the succeeding versions of the MISA method [Paquette et al 1999]
17
Meta-knowledge representation
To define learning needs and guide the modeling process, we need to assign skills to knowledge using the
A link. Here are some skill selection principles based on the second layer of the skills taxonomy.

If a knowledge is fundamental to actors in the target population and these actors have to reach an
advanced level of expertise to be able to advise other persons, the level of skill should be high: 7Repair , 8-Synthesize , 9-Evaluate or 10-Self-manage.

If a knowledge is important to actors in the target population, requiring from them a large level of
autonomy, the level of skill should be above average: 5-Apply/Use, 6-Analyze , 7-Repair, 8Synthesize.

If a knowledge is useful to actors in the target population, requiring its regular use, the level of skill
should be average: 3-Instanciate , 4-Transpose , 5- Apply/Use or 6-Analyse.

If a knowledge is sometimes useful to actors in the target population, asking them to retain only the
main elements, the level of skill can be weak: 1-Pay attention, 2-Integrate or 3-Instanciate.
4.3 Building process-based learning scenario
We will now use the simulation generic process, presented of figure 3, to build a learning scenario based
on the “Simulate a production process” competency for a multimedia director. To do this, we lay out a
graph corresponding to the generic simulation process, but taking a “learning activity” viewpoint. As
shown on figure 9, the graph is instantiated in a way that the vocabulary of the specific application
domain (multimedia production) is used. It is also formulated in an “assignment style” displaying six
activities. Globally, based on the generic simulation process input and product, the learning scenario starts
on a description of the process to simulate and ends on producing essentially a trace report.
Documents
on the
multimedia
production
processes
Activité 1:
Choose a
process to
simulate
I/P
Activity 4:
Execute a
production
task
I/P
Activity 2:
Choose a typical
multimedia
project
Production
tasks and
principles
I/P
I/P
I/P
C*
Tasks'
products
I/P
A task
description
Case to be
simulated
I/P
Activity 3:
Identif y a
production task
Project
report
I/P
I/P
Activity 6:
Produce a
project report
on the process
I/P
I/P
P
P
No
P
Activity 5:
Verify is the
process is
complete
P
Yes
Figure 9– A learning scenario: simulate the “Search the Internet” process
Of course this scenario is not yet complete. For example, we could add some collaboration assignments
and a description of the method for learner evaluation. But the important thing here is that the generic
process will form the backbone of the learner’s assignments. In that way, we make sure that he exercises
the right skill, simulating a process, while working on the specific knowledge domain, thus building
specific domain knowledge and meta-knowledge at the same time.
18
Meta-knowledge representation
4.4 Defining actors and resources in a telelearning system
We now use a learning scenario as a basis to build an assistance scenario describing the activities and
products of other actors in a telelearning system such as a trainer, a content expert, a designer or a
manager. To identify the assistance activities ruled by these actors, we go back to the generic process
which gave birth to the learning scenario and we determine the principles governing the execution of the
generic process. Figure 10 supplements the generic simulation process on figure 3 with generic principles
controling each of the sub-processes.
Simulation
execution
principles
Description
principles
R
Simulate a
proce ss
I/P
I/P
(3) Produce
examples of the
input concepts
C
(2) Identify the
next
applicable
procedure
I/P
Procedure
identification
principles
I/P
Assemble the
simulation
trace
I/P
I/P
Products of
the procedure
Execute the
procedure using
its execution
principles
I/P
C
P
Example
generation
principles
Simulation trace of
the procedure
Execution
principlesCof the
simulated
procedure
C
C
R
R
R
R
Description of
the process to be
simulated
I/P
Presentation
principles
There are
more
procedures
R
Completeness
principles
(when to stop)
C
P
R
There are
no more
Figure 10 – Meta-principles for the management of simulation processes
The statement of these principles will be the base of the assistance scenario presented on figure 11. On
table 7, we first state some of the principles on figure 10 and decide on a corresponding type of assistance
to the learners.
Meta-procedure
Description of
the simulated
process
Generation of
examples
(cases to be
simulated)
Identification
of procedures
to be applied








Execution of
procedures
Completeness
of the
simulation




Examples of meta-principles
Inputs and products of the simulated process have to be clearly identified.
Simulated process must be decomposed into its main procedures, if
necessary on more than 2 levels.
Principles governing the execution of the process must be identified.
Every example has to contain a value for each of the inputs of the process to
be simulated.
Examples have to cover all the possible cases of execution of the procedure.
Type of assistance
Case studies on a
method to describe
simulated procedure
For each of the examples, build a structured list with the products of already
executed procedures and add them to inputs.
Eliminate from the preceding list the products which are not inputs of a still
unexecuted procedure.
Always choose procedures giving the greatest number of new products for
still unexecuted procedures
Once a procedure to be executed has been chosen, use its execution
principles to obtain new products.
Execution depends especially on the domain of application.
If the simulated process is sequential, in parallel or a decision tree, the
simulation is completed when every possible branch was executed for at
least an example,
If the process is iterative, it contains execution principles telling when to
stop cycling; the simulation is completed when each of these stop principles
have been tested for at least an example.
Texts presenting in
detail these principles
as well as examples of
execution traces
Interactive advisor
giving help adapted to
the examples supplied
by the learner
Interaction by e-mail
with a content expert
Presentation of these
completeness
principles and
dialogue with a trainer
to assess completeness
of a simulation
19
Meta-knowledge representation


Presentation
of the
execution
traces
Contextual help on the
presentation of these
standards
accompanied with
examples.
Presentation must contain the description of the process to be simulated
Simulated examples must be regrouped in categories according to the
structure of the process.
For every example, present the succession of executed procedures and their
products.
Table 7 – Examples of meta-principles and corresponding assistance

Figure 11 presents, the scenario of assistance superposed to the learning scenario of figure 9, every form
of assistance correspond to a principle stated in table 7. It shows three actors giving different forms of
Content
ex pert
R
Case studies on a
method to select and
describe a
procedure
Designer
R
Interact by
email
Activity 1:
Choose a process
to simulate
I/P
I/P
I/P
Prepare a
learning
material
I/P
Interactive advise
on ex amples
processed by
learners
I/P
I/P
Tex t presenting
ex amples of
simulations and the
identification of
procedures to be
I/P
Activity 2:
Choose a typical
multimedia
project
I/P
Activity 6: Produce
a project report on
the process
Activity 4:
Execute a
production task
Activity 3:
Identify a
production task
Activity 5:
Verify is the
process is
complete
I/P
Presentation and
discussion of
completeness
principels
Learner/Content
ex pert interactions
I/P
I/P
Frequently asked
questions and
ex emples on
presentation norms
Trainer
R
Use a f orum
software
R
I/P
Maintain a
FAQ
assistance to the learners.
Figure 11 – Assistance scenario for the simulation of a multimedia production process
This assistance scenario puts in evidence different activities of assistance (in inverted text on the figure)
producing help resources for each of six learning activities (the other details of the learning on figure 10
were omitted here).

Three of these assistance resources are materials prepared by a designer, being use as inputs to the
first three learning activities.

Another type of assistance in activity 4, is an interaction by e-mail with a content expert in
multimedia production.

Two other forms of assistance, in activities 5 and 6, involve a trainer animating a forum on the
completeness principles of a simulation and also managing a FAQ (frequently asked questions) on the
presentation norms for the final report.
The choice of these assistance activities, as well as the learning activities that they support, result from the
generic process representing a skill associated to a principal knowledge in the learning unit. More exactly,
each of these forms of assistance draws its content from the principles describing how this skill can be
applied to knowledge processed in the learning unit. A skill’s generic process and its execution principles
20
Meta-knowledge representation
thus define the content of the assistance supplied by a person playing the role of a facilitator, directly or
through different types of teaching equipments or tools.
DISCUSSION
It is our firm belief that knowledge, skills and competencies have to be represented in a standardized and
easily interpretable graphic language. We are not pretending that the MOT language outlined here is the
best way to do it in every circumstance, nor that it should be imposed on designers, trainers or learners.
This would considerably impoverish the knowledge building activity that is central to learning, training
and designing and needs multiple representations.
But the MOT language is general enough to represent application domain knowledge as well as generic
skill and other meta-knowledge, and also their interrelations. We have accumulated considerable evidence
of this in the last ten years through projects where this representation method has served as an alternative
to more specialized graphic methods Jonassen et al 1993 such as conceptual maps, procedural task
representation, causal and influence diagrams, decision trees and rule-based diagrams, etc. We have, for
example, used the language to model the instructional processes and principles in our instructional
engineering method (MISA) or to describe the ways actors rule processes and interact with resources in a
computerized school.
Our hope is that the graphic language and the few examples outlined here will succeed in open up new
research directions where precise and inspectable representations can be given to competencies, skills and
problem taxonomies, thus favoring a larger exchange of instructional design knowledge. Libraries of
skills and knowledge models are central to the way we build designs for learning and tele-learning
environments. Because we need to reuse and adapt models from different sources in a computerized
support system (ADISA) Paquette 2001, an interchange graphic language was needed, in a similar way
that XML provide an interchange format between ways to store and retrieve data and metadata.
The most stimulating aspect of a skills taxonomy built at the meta-knowledge level concerns the
opportunity given to research teams to create a relatively complete set of representable, reusable and
significant components in this huge and challenging puzzle of learning systems engineering.
REFERENCES
[Alexander, 1987] J.R. Alexander, M. Freyling, S. Shulman, S. Rehfuss, S. Messick. Ontological analysis: An
ongoing experiment. Int. J. of Man-Machine Studies, 26(4), pp. 473-485
[Bélisle et Linard 1996] Bélisle C. et Linard M.. Quelles nouvelles compétences des acteurs de la formation dans le
contexte des TIC?. Éducation permanente, no 127, 1996-2, pp. 19-47
[Bloom, 1975] Bloom, Benjamin S. Taxonomy of Educational Objectives: the Classification of Educational Goals.
New York: D. McKay.
[Breuker et Van de Velde, 1994] Breuker J. and Van de Velde W. CommonKads Library for Expertise Modelling.
IOS Press, Amsterdam, 360 pages, 1994.
[Chandrasekaran 1987] B. Chandrasekaran. Towards a Functional Architecture for Intelligence Based on Generic
Information Processing Tasks. Proceedings IJCAI-87, Milan Italie, pp 1183-1192
[Friedlander 1996] P. Friedlander. Competency-Driven, Component-Based Curriculum Architecture. Performance
Improvement, February 1996, pp. 355-362
21
Meta-knowledge representation
[Gardner 1993] Gardner, H. Multiple Intelligences : The Theory in Practice. Basic Books, New York, 1993
[Gagné 1970] Gagné, R. M. The conditions of learning (2nd ed.) New York, Holt, Rhinehart & Winston, 1970
Goël et Pirolli, 1989] Goël Goel V., Pirolli P. Design within Information-Processing Theory: The Design Problem
Space, AI Magazine, Spring 1989: 19-36.
[Goleman 1997] L’intelligence émotionnelle. Traduction française chez Robert Laffont, 505 pages, 1997
Hayes-Roth et al. 1983. Hayes-Roth F,. Waterman D.A., Lenat. D.B. Building Expert Systems. Addison-Wesley
1984, 444 pages.
Holley & Dansereau, 1984. Networking: The technique and the emprirical evidence. In C. D. Holley & D. F.
Dansereau (Eds.), Spatial learning strategies: Techniques, applications and related issues. New York: Academic
Press. 1884.
Jonassen et al 1993. Jonassen D.H., Beissner K., & Yacci, M. Structural Knowledge – Techniques for
Representing, Conveying and Acquiring Structural Knowledge. Laurence Earlbaum Associates, New Jersey, 265
pages, 1993
[Martin and Briggs, 1986 Martin B.L. & Briggs L. The Affective and Cognitive Domains: Integration for
Instruction and Research. Educational Technology Publications, New Jersey, 494 pages, 1986
[McDermott, 1988] J. McDermott. Preliminary steps towards a taxonomy of problem-solving methods. In Marcus,
S., editor, Automating Knowledge Acquisition for Expert Systems, pp. 225-255. Kluwer Academic Publishers,
Boston, Mass.
[Merrill 1994] D. Merrill. Principles of Instructionnal Design. Educational Technology Publications, Englewood
Cliffs, New Jersey, 465 pages.
[Paquette et al, 1994] G. Paquette, F. Crevier, C. Aubin. ID Knowledge in a Course Design Workbench. Educational
Technology, USA, volume 34, n. 9, pp. 50-57, November 1994
[Paquette et al, 1996] G. Paquette and J. Girard. AGD: a course engineering support system, ITS-96, Montréal, June
1996.
[Paquette, 1996] G. Paquette. La modélisation par objets typés: une méthode de représentation pour les systèmes
d’apprentissage et d’aide a la tâche. Sciences et techniques éducatives , France, avril 1996
[Paquette 1998] G. Paquette. Metaknowledge representation, application to learning systems engineering. TL-NCE
technical reports, Vancouver, Canada, 1998
Paquette et al. 1999
Paquette G., Aubin C. and Crevier, F. MISA, A Knowledge-based Method for the
Engineering of Learning Systems, Journal of Courseware Engineering, vol. 2, August 1999.
Paquette 2000 Paquette, G.. TeleLearning Systems Engineering – Towards a new ISD model, Journal of Structural
Learning, Accepted paper, 2000
Paquette 2001 Paquette, G. I. Rosca and M. Léonard. ADISA- Instructional Engineering Tools for Web-based
Learning. SALT conference, (paper accepted for publication), 2001
[Pitrat, 1991] Jacques Pitrat. Métaconnaissance, avenir de l’Intelligence Artificielle. Her mès, Paris, 1991.
[Pitrat 1993] Pitrat J. Penser l’informatique autrement. Hermès, Paris, 1993.
22
Meta-knowledge representation
[Popper 1967] Popper K. R. The Logic of Scientific Discovery, Harper Torchbooks, New York, 1967..
[Reigeluth, 1983] Reigeluth C. Instructional Design Theories and Models: An overview of their current status. and
Models. Hillsdale, NJ: Lawrence Earlbaum, 487pages, 1983
[Reigeluth, 1999] Reigeluth C. Instructional Design Theories and Models (Volume II): A New Paradigm of
Instructional Theory. Hillsdale, NJ: Lawrence Earlbaum, 715 pages, 1999.
[Romiszowski 1981] A. J. Romiszowski. Designing Instructional Systems. Kogan Page London/Nichols Publising,
New York, 415 pages.
[Schreiber et al, 1993] Schreiber G., Wielinga B., Breuker J. KADS – A Principled Approach to Knowledge-based
System Development. San Diego : Academic Press. 457 p.
[Spector et al. 1993] Spector J.M., Polson M.C., Muraida D.J. (Eds) Automating Instructional Design, Concepts and
Issues, Educational Technology Publications, Englewood Cliffs, New Jersey, 364 pages, 1993
[Steels 1990] L. Steels. Components of Expertise. AI Magazine, vol. 11, no 2, Summer 1990.
[Tennyson, 1990] Tennyson, Robert D. Cognitive Learning Theory Linked to Instructional Theory. In Journal of
Structured Learning, Vol. 10(3): 249-258
[Thayse 1988] Thayse, A. Approche logique de l’intelligence artificielle, Dunod, Paris, 1988
[West, 1991] C. K. West, J. A. Farmer, P. M. Wolff. Instructional Design, Implications from Cognitive Science.
Allyn and Bacon, Boston, 271 pages.
23