A Pedagogical Scenario Language for Virtual Learning Environment

based on UML Meta-model

Application to Blood Analysis Instrument

eric Le Corre, Charlotte Hoareau, Franck Ganier, C

edric Buche and Ronan Querrec

UEB/ENIB/CERV, 25 rue Claude Chappe, F-29490 Plouzan

e, France

Keywords:

Intelligent Tutoring System, Virtual Environment, Generic, PedagoGical Scenario, Knowledge Base, Learn-

ing, Training.

Abstract:

Training to learn the use and maintenance of biomedical devices have various constraints. In order to complete

these trainings, we proposed to use virtual reality based on pedagogical scenarios and Intelligent Tutoring Sys-

tems (ITS). In this paper, we ﬁrst established the existing pedagogical scenario models and ITS. Subsequently

we presented our proposal of a formal model based on the concept of learning organization by extension of

UML in order to describre some pedagogical scenario and ITS. The use of this model is illustrated by an appli-

cation of a virtual biomedical analyzer with the aim of learning the technical procedures of the device. Finally,

we performed two experiments in order to verify the efﬁciency of virtual reality training.

1 INTRODUCTION

In the biomedical domain, the traditional training

method consists of several learners in a classroom

with a live instructor manipulate the device alter-

nately: method one-to-many which is known to be

irrelevant (Bloom, 1956; Bloom, 1984) This is the

case for the STA-R



instrument (Figure 1) produced

by STAGO who sponsored this work. Unfortunately

there are several constraints inherent in this training

method.

Figure 1: Picture of the biomedical analyzer STA-R



First, there is always a potential biohazard while

dealing with biological ﬂuids. Second, learners have

to use reagents to prepare the blood tests, which in-

crease the cost and variability of the learning sessions.

In some cases, the biomedical instrument is trans-

ported to the site of the learning session, which could

damage it and again can impact the cost. Sometimes,

the learning session takes place in the customer’s site;

therefore the instrument cannot be available for train-

ing. Finally the use of diagnostic device is time-

consuming for the learner because of inherent time

needed during the analytical.

Considering the above, the use of informatics

tools for biomedical training sounds interesting. The

objective of this training is to manipulate the device

according to well-deﬁned procedures. However the

successful completion of these procedures requires

biology knowledge. A learning strategy like “learning

while doing” is relevant : the learner has to manipu-

late the device. Therefore, we consider that virtual re-

ality is a major contribution in this context. Virtual re-

ality has proven to be efﬁcient to solve those kinds of

issues (Mikropoulos and Natsis, 2010; Okutsu et al.,

2012). Pedagogical situation using virtual reality are

called Virtual Learning Environment (VLE).

However in a lot of professional environment and

particularly in biomedical sector, there is a lot of staff

turnover. It became then pretty common for employee

to not go through the classical learning sessions. The

objective of this paper is to propose a solution to im-

plement the pedagogical strategy of the trainer inside

the VLE so that the learners could train without hav-

301

Le Corre F., Hoareau C., Ganier F., Buche C. and Querrec R..

A Pedagogical Scenario Language for Virtual Learning Environment based on UML Meta-model - Application to Blood Analysis Instrument.

DOI: 10.5220/0004845803010308

In Proceedings of the 6th International Conference on Computer Supported Education (CSEDU-2014), pages 301-308

ISBN: 978-989-758-020-8

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

ing the initial training. Moreover, each individual

learner can have its own basal level. So, our VLE

have to be able to adapt itself to the learner in order to

be really efﬁcient. In medical ﬁeld, some work were

conducted like an application for medical problem-

based learning (Suebnukarn and Haddawy, 2007).

This paper is organized in ﬁve sections. In sec-

tion 2 we determined the existing solutions in the con-

text of pedagogical scenarios and ITS. In section 3

we proposed our model based on a deﬁnition of a

language for a pedagogical scenario (Koper, 2001).

The ultimate goal for STAGO was to make an applica-

tion based on the biomedical devices: VIRTUALANA-

LYZER, presented in section 4 and we also conducted

two experiments to verify the efﬁciency of virtual re-

ality training. Finally, we concluded, in section 5, by

listing the characteristics of our model that needed a

further improvement and detailed the terms of an in-

coming experiment.

2 RELATED WORK

We focus on the acquisition of declarative and pro-

cedural knowledge on complex systems (many types

of heterogeneous elements) in virtual reality. In

this ﬁeld, the most popular and representative tu-

tor is STEVE (Rickel and Johnson, 1998), an an-

imated pedagogical agent. STEVE belongs to the

ITS ﬁeld. Classically, ITS are structured on four

models : domain, pedagogical, learner and interface

(Wenger, 1987). One of the classic problems in ITS

is to provide a generic language to describe domain

knowledge. In our case, this knowledge is complex

and from a high level of expertise. We have al-

ready proposed a meta-model MASCARET (based on

SYSML, an UML extension) in order to acquire the

system speciﬁcations to learn, directly from the con-

ception(Querrec et al., 2013). This method allows us

to generate the domain model and execute it in the vir-

tual environment. It is not conceivable to rewrite this

knowledge in the domain model, so it will be directly

imported. Similarly, these speciﬁcations will directly

lead to the generation of the virtual environment. A

second problem in ITS is the link between the rules

governing the pedagogical behavior of the tutor and

the pedagogical course of training. Under the train-

ing procedures on complex systems, work has already

proposed learning scenarios templates (explanation of

the system and subsystem, explanation and organiza-

tion of the procedures, repetition more or less guided).

In our more speciﬁc context, learning scenarios have

already been deﬁned, it is therefore necessary to im-

port it. In this paper, we propose a model to explain

this pedagogical scenario in the pedagogical model

related to the domain knowledge.

2.1 ITS

Our work belongs to ITS ﬁeld. Classically ITS are

structured around four modules:

• Domain : knowledge on the job to learn

• Pedagogical : knowledge on the pedagogical

strategies

• Learner : knowledge representation of the learner.

Very often, this knowledge is a subset of the do-

main knowledge where the learner had access.

• Interface : knowledge representation that the tutor

may have on the actions that the learner achieves

and the actions that the tutor can make in return.

This paper focuses exclusively on the pedagogical

and domain models. Work on these models aim to

make them generic, adaptative and individual (hence

the link with the learner model). In the ITS domain,

many research works were made during the past few

years: some projects aimed at developing the generic

part of ITS (Sanchez and Imbert, 2007; Shi and Lu,

2006; Sorensen and Ramachandran, 2007). Some

projects aim at individualizing the simulation for each

learner. This could be done for example through emo-

tional agents (Ailiya et al., 2010) or by Hollnagel clas-

siﬁcation (El-Kechai and Despr

es, 2007). At lower

scale, some have highlighted the adaptability charac-

teristic of ITS (Dos Santos and Osorio, 2004).

Typically, these models are deﬁned from cognitive

architectures like STEVE which is based on SOAR

or CTAT (Koedinger and Heffernan, 2003) which

is based on ACT-R (Adaptive Control of Thought-

Rational). Knowledge expressed in this system is de-

ﬁned as a set of rules like in ANDES (Vanlehn et al.,

2005). In this context we have already proposed PE-

GASE which is based on the four classical models (do-

main, learner, pedagogical, inteface) (Woolf, 1992).

PEGASE also contains an error model and a deﬁni-

tion of the pedagogical model (usable independently

of the exercise to do) in order to generalize it. One

last important part of PEGASE is its adaptability, al-

lowed by the auto-modiﬁcation of the pedagogical

model. PEGASE proposes a pedagogical model based

on a hierarchical classiﬁer system. This system orga-

nizes knowledge while taking the abstraction of the

data involved into account. It structures knowledge

according to three levels, from rules based on abstract

knowledge of educational methods (the pedagogical

approach), to the rules based on concrete knowledge

http://sitemaker.umich.edu/soar/home

CSEDU2014-6thInternationalConferenceonComputerSupportedEducation

302

of virtual reality (pedagogical techniques), via an in-

termediary level (pedagogical attributes).

A major difﬁculty in the design of these models is

writing these rules. For this there are three technical :

• Authoring tools like in CTAT (Koedinger et al.,

2004), but can not be as generic as desired due to

the proposed interfaces.

• Model-tracing which observe the actions of ex-

perts to build the domain model.

• Meta-model deﬁnition like the KBT-MM model

of Murray (Murray et al., 2003) which has how-

ever a too high level of abstraction.

From this meta-model, Murray proposed a specializa-

tion called EON. Knowledge is then deﬁned using

ontologies. Many researches focused on some lan-

guages to describe the ontologies, like Resource De-

scription Framework (RDF) or Web Ontology Lan-

guage (OWL). The major problem of theses lan-

guages is that they do not allow to describe the dy-

namics of the situation when we precisely needed it to

teach how works the instrument. We are interested in

this approach but at an intermediate level compared to

KBT-MM. Indeed, we propose a model in the context

of learning procedures (handling, maintenance and di-

agnostic) on complex systems (industrial). Moreover,

these systems are so complex that it is not possible to

rewrite the system speciﬁcations (for the knowledge

base or for the conception of the virtual environment).

MASCARET is an extension of the meta-model

SYSML. It enables to describe the system structure

by blocks, attributes and compositions from SYSML.

The reactive behaviors of the system elements are de-

scribed by state machines and the domain procedure

are described by activities. MASCARET is a kind of

SYSML interpreter for virtual reality and provides an

operational semantic for every element of the meta-

model. This enables to make the knowledge explicit

for the agents during the simulation and to automat-

ically execute the system entities behaviors. As we

said in section 1, on this kind of system, work on the

training structuring has already been done and it is

therefore appropriate to incorporate the pedagogical

scenario knowledge in the pedagogical model.

2.2 Pedagogical Scenario

Prior to 2000, teaching scenarios were based primar-

ily on documentary approaches like Learning Ob-

ject Metadata

(LOM) or Sharable Content Object

Reference Model

(SCORM). From the 2000s Koper

http://www.lom-fr.fr

http://www.scorm.com

(Koper, 2001) initiated the use of educational mod-

eling languages like Educational Modeling Language

(EML). The main feature of EML is, that unlike doc-

umentary approaches, EML focus on the description

of activities and not on educational resources.

Afterwards, Koper (Koper et al., 2003) con-

tributed to the IMS-Learning Design (IMS-LD) stan-

dard, based on EML. IMS-LD focused on the concept

of learning unit as a base element of the description of

the learning process. Indeed, in IMS-LD a scenario is

considered as a series of educational activities. Each

of these activities is described by a text or a set of doc-

uments explaining the purpose of the activity, the task

to achieve, the instructions to be followed, etc.

However, IMS-LD has its own limitation, like this

review of Ferraris (Ferraris et al., 2005) based on

the lack of expressivity regarding the description of

the interactions between users in collaborative tasks.

These models from VLE do not fully meets our re-

quirements, so we had to look more speciﬁcally the

existing models in virtual reality for training.

In the literature we could identify several virtual

reality models for training. The FORMID (FOR-

Mation Interactive Distance) (Gu

eraud et al., 2004)

project focus on pedagogical scenario activities in

which learners interact with an interactive pedagog-

ical object. Another proposal is called GVT (Generic

Virtual Training) and is intended for procedure learn-

ing. The disadvantages of FORMID and GVT are

that they are not generic, they are not reusable and

there is no distinction between the activity scenario

for the procedures of the environment and the peda-

gogical scenario.

Marion (Marion et al., 2009) proposed a model of

a pedagogical scenario called POSEIDON, which aims

to be directly reusable in different environments. PO-

SEIDON covers various points like: educational ob-

jectives, prerequisites, education activities, education

organizations and environments. We have decided to

base our own pedagogical scenario model on POSEI-

DON because of this generic side. However the nega-

tive point of POSEIDON is its lack of link with an ITS.

Our goal is to create this link so that the ITS can use

the pedagogical scenario knowledge to reason.

3 MODEL

Classically, a pedagogical scenario is composed of

pedagogical objectives and prerequisites, pedagogical

organization (set of roles), pedagogical activities and

an environment (Koper, 2001). Following that deﬁ-

nition, we considered that the trainer activity is a do-

main activity just like any other. Thereby we could

APedagogicalScenarioLanguageforVirtualLearningEnvironmentbasedonUMLMeta-model-ApplicationtoBlood

AnalysisInstrument

303

Figure 2: Workﬂow for application conception.

propose a formal model of the concept of a pedagog-

ical scenario by extended UML. A pedagogical orga-

nization is considered like collaboration composed of

Roles. A role is the UML concept of Interface. This

means that a role lists a set of services without actu-

ally providing an implementation. The agent (artiﬁ-

cial or human) who will play this role will propose its

own implementation. Within this pedagogical organi-

zation, the role takes part to the pedagogical scenario

as an activity arranging the pedagogical actions which

could modify the virtual environment (composed of

the system or the pedagogical resources). These con-

cepts enable to create a pedagogical scenario and its

development is now presented.

3.1 Workﬂow

In order to achieve a complete virtual reality training

application (application, ITS, scenario), several peo-

ple of various ﬁelds must participate. The picture 2

summarizes all the contributors described in this sec-

tion.

First, an educationalist instructor assisted by a vir-

tual reality expert deﬁnes the ITS model. Indeed, his

instructor qualities enable him to deﬁne the best ped-

agogical strategies for the ITS, and the virtual reality

expert knowledge allows to implement the pedagog-

ical actions speciﬁc to virtual reality (set an entity in

transparence, switch the point of view,...) The knowl-

Figure 3: UML model of pedagogical scenario for system

knowledge acquisition.

edge is and must be generic. They are described and

implemented only one time and are reused for all ped-

agogical scenarios in all kind of ﬁelds. Second, an-

other contributor must participate to the application

development: the application domain expert who de-

scribes the entire domain model. The expert knowl-

edge is centered on the ﬁeld on which the application

is based on. Finally, the domain teacher imports the

ITS model, described by the educationalist instructor,

and the domain model, described by the domain ex-

pert, in order to produce some pedagogical scenarios.

This three step development enables to separate the

pedagogical scenario conception and the system con-

ception. Thereby, our pedagogical scenario model is

generic and can be applied to other instruments and in

other ﬁelds.

3.2 Pedagogical Scenario and ITS

Thus, the trainer could describe some pedagogical

scenarios to learn from the system. This can be of

two types: for knowledge acquisition or for proce-

dure acquisition (use, maintenance and diagnostic).

Based on an example of a plug-in timer application,

we can present a pedagogical scenario for knowledge

acquisition (Figure 3) including a trainer role who is

responsible for presenting the plug-in timer and each

of its buttons (Figure 4).

The “present” operation could be deﬁned by the

educationalist instructor in this manner:

• set all the entities in transparence except for the

one to present

• change the point of view on the entity to present

CSEDU2014-6thInternationalConferenceonComputerSupportedEducation

304

Figure 4: Picture of a plug-in timer in the virtual environ-

ment with the “present“ operation

• display the annotation from the domain model on

the entity to present

So, the trainer should rely on this pedagogy and the

domain model in order to create his pedagogical sce-

nario. We could also describe some pedagogical sce-

narios for procedure acquisition: for example the

teacher role presents the procedure and the learner

role has to carry out the domain procedure.

Moreover, we could describe ITS behaviors in the

same manner as the pedagogical scenarios described

previously. This kind of modelisation ensures a high

modularity of the system, which enables to create ITS

behaviors more or less complex. For example, in pic-

ture 5, the trainer described the learner role which had

to carry out a procedure of setting time of the plug-

in timer. The teacher role monitors the environment,

highlight the learner errors and perform an “undo“ on

the current action (back to the previous state). The

advantage is that the trainer can easily remove or add

some actions, making our system very modular.

In the same way, we could create another appli-

cation: VIRTUALANALYZER which is applied to our

biomedical training problematic.

4 APPLICATION

The proposed model has been applied to the problem

of training presented in section 1 in order to realize a

virtual reality application for STAGO.

Figure 5: UML model of a ITS behavior

4.1 Virtual Environment

While looking for replicating the real STA-R



in a

virtual environment, we made a virtual reality appli-

cation with the instrument in a 3D environment and

the control software modeled on ANDROID (Figure

6).

Figure 6: Virtual STA-R



application.

The environment includes a pre-operative blood

work procedure. This procedure is composed of 125

basic actions and is coupled with a simple ITS behav-

APedagogicalScenarioLanguageforVirtualLearningEnvironmentbasedonUMLMeta-model-ApplicationtoBlood

AnalysisInstrument

305

1 2 3 4

5 6

7 8 9 10

Number of trials

Figure 7: Learning curve: total time.

1 2 3 4

5 6

7 8 9 10

Number of trials

Figure 8: Learning curve: instructions.

ior: vocal assistance and blocking wrong actions. In

order to verify the efﬁciency of virtual reality training,

we performed two experiments. The ﬁrst one aimed

at verifying the procedure learning with 12 computer

science students during two days. The second exper-

iments goal was to verify the knowledge transfer on

a real STA-R



with 58 biology students during ten

days.

4.2 Experiments

4.2.1 Learning

First, the learners had to perform seven times the pro-

cedure described previously. After a week time, the

learners came back to complete the second session of

the experiment: they had to do three times the same

procedure. In order to evaluate the learning proce-

dure we collected behavioral data: total time of the

procedure achievement, number of audio instructions

consultation, number of wrong actions (Figure 7 and

8).

The total time of the procedure achievement and

Table 1: Results of the knowledge transfer experiment.

Traditional Virtual Control

Achievement 100% 100 72%

SOS 0.2 0.7 3.1

Interventions 1.4 2.3 3.9

Consultations 9.5 1.9 48.7

Total time 19’48” 30’02” 39’09”

the number of audio instructions consultation de-

creased over repetitions. After seven days, the learn-

ers partially consulted the instructions only during the

ﬁrst try, and then their performance became similar

to the latest ones of the ﬁrst session. This is a typi-

cal learning curve. Thereby, this study conﬁrmed the

usability of VIRTUALANALYZER for learning proce-

dure. However, a large number of try would be nec-

essary to obtain a stabilization of performance and a

lower number on instruction consultation at the 8th

try, certifying the perfect acquisition of the procedure

and its storage in long-term memory. Learning by us-

ing a virtual environment is favorable only if the skills

acquired through this device can be used in a real sit-

uation. We attempted to verify this assumption in the

following experiment.

4.2.2 Knowledge Transfer

This experiment aimed to verify the knowledge trans-

fer on a real STA-R



with 58 biology students. The

students were divided into three groups: control, tra-

ditional teaching and virtual teaching. Each of these

groups had a theoretical training on STAGO, hemosta-

sis, biological tests and the instrument during one

hour. The traditional group had a classical train-

ing provided by a trainer from STAGO (six learn-

ers for one instrument) and the virtual group had a

training with VIRTUALANALYZER (one computer for

each learner): each group trained for two hours while

the control group had no training. Thereafter, each

learner had to individually perform the entire proce-

dure on the real STA-R



In order to evaluate the transfer procedure, we col-

lected some data: procedure achievement, number of

paper document consultations, number of technician

interventions, number of technician call for help and

total time (Table 1).

This experiment showed no signiﬁcant differences

between the data of the traditional method and the vir-

tual method, except for the total time. This difference

can be explained by several ways: our android inter-

face is very different from the real interface of the

STA-R



and the learners have no notion of time of

the working instrument. We hope to reduce this time

to a value close to the traditional one, by modifying

CSEDU2014-6thInternationalConferenceonComputerSupportedEducation

306

how to provide some information in our application.

The knowledge acquired with our virtual environment

is transferable in a real STA-R



5 CONCLUSION AND FUTURE

WORKS

STAGO wanted to use virtual reality and virtual envi-

ronments for training for their biomedical diagnostic

devices. Thereby, we proposed a training application

for a STAGO instrument called VIRTUALANALYZER.

This enabled us to verify the quality of the learning

during a training based on this application and also to

check if the knowledge is transferable in a real envi-

ronment. Therefore, our work met the objectives de-

ﬁned by STAGO. However, the conducted experiment

did not enable to verify the contribution of our model

on the learning. Indeed, the experiments are based on

the application rather than a complex ITS behavior.

This is why we would like to evaluate, in the future,

the contribution of the ITS with a complex behavior

(like the PEGASE one) on the application VIRTUAL-

ANALYZER. We also pointed out that the language

ergonomic for describing the pedagogical scenarios

is not intuitive for the trainer. Although the interface

is graphical and formalized, the fact that it relies on

UML concepts does not help the ergonomic and does

not facilitate its use by domain trainers. Finally, the

long-term goal for STAGO would be to deploy this

work to their whole range of instruments.

REFERENCES

Ailiya, Shen, Z., and Miao, C. (2010). An emotional agent

in virtual learning environment. In Transactions on

Edutainment IV, LNCS 6250, pages 22–33.

Bloom, B. (1956). Taxonomy of educational objectives. In

Educational Researcher, volume 1.

Bloom, B. (1984). The 2 sigma problem: The search for

methods of group instruction as effective as one-to-

one tutoring. In Educational Researcher, volume 13,

pages 3–16.

Dos Santos, C. T. and Osorio, F. S. (2004). Integrating in-

telligent agents, user models, and automatic content

categorization in a virtual environment. In ITS 2004,

LNCS 3220, pages 128–139.

El-Kechai, N. and Despr

es, C. (2007). Proposing the un-

derlying causes that lead to the trainee’s erroneous

actions to the trainer. In EC-TEL : European Con-

ference on Technology Enhanced Learning, volume

4753, pages 41–55.

Ferraris, C., Lejeune, A., Vignollet, L., and David, J.-P.

(2005). Mod

elisation de sc

enarios p

edagogiques col-

laboratifs. In Conf

erence EIAH.

eraud, V., Adam, J.-M., Pernin, J.-P., Calvary, G.,

and David, J.-P. (2004). L’exploitation d’objets

edagogiques interactifs

a distance : le projet

FORMID. STICEF, 11:103–163.

Koedinger, K. R., Aleven, V., Heffernan, N., McLaren, B.,

and Hockenberry, M. (2004). Opening the door to

non-programmers: Authoring intelligent tutor behav-

ior by demonstration. INTELLIGENT TUTORING

SYSTEMS Lecture Notes in Computer Science, Vol-

ume 3220/2004:7–10.

Koedinger, K. R. and Heffernan, N. (2003). Toward a rapid

development environment for cognitive tutors. In in

Proceedigns of the International Conference on Arti-

ﬁcial Intelligence in Education, pages 455–457. IOS

Press.

Koper, R. (2001). Modeling units of study from a pedagog-

ical perspective. In Educational Technology Expertise

Centre, Open University of the Netherlands.

Koper, R., Olivier, B., and Anderson, T. (2003). Ims learn-

ing design information model. In IMS Global Learn-

ing Consortium.

Marion, N., Querrec, R., and Chevaillier, P. (2009). Inte-

grating knowledge from virtual reality environments

to learning scenario models. a meta-modeling ap-

proach. In International conference of computer sup-

ported education, pages 254–259.

Mikropoulos, T. and Natsis, A. (2010). Educational vir-

tual environments: A ten-year review of empirical re-

search (1999-2009). In Computer and Education, vol-

ume 56, pages 769–780. Elsevier.

Murray, T., Blessing, S., and Ainsworth, S. (2003). Author-

ing Tools for Advanced Technology Learning Envi-

ronments: Towards cost-effective adaptive, interactive

and intelligent educational software. Kluwer Aca-

demic Publishers.

Okutsu, M., DeLaurentis, D., Brophy, S., and Lambert, J.

(2012). Teaching an aerospace en- gineering design

course via virtual worlds: A comparative assessment

of learning outcomes. In Computer and Education,

volume 60, pages 288–298. Elsevier.

Querrec, R., Vallejo, P., and Buche, C. (2013). MAS-

CARET: creating virtual learning environments from

system modelling. In Engineering Reality of Virtual

Reality (ERVR’13), volume 8649, pages 8649–04, San

Francisco. SPIE.

Rickel, J. and Johnson, W. L. (1998). Steve: A pedagog-

ical agent for virtual reality. In Second International

Conference on Autonomous Agent, pages 332–333.

Sanchez, L. and Imbert, R. (2007). An agent-based adapt-

able and conﬁgurable tutoring module for intelligent

virtual environments for training. In Edutainment

2007, LNCS 4469, pages 499–510.

Shi, R. and Lu, P. (2006). A multi-criteria programming

model for intelligent tutoring planning. In KES 2006,

Part I, LNAI 4251, pages 780–787.

Sorensen, B. and Ramachandran, S. (2007). Simulation-

based automated intelligent tutoring. In Human Inter-

face, Part II, HCII 2007, LNCS 4558, pages 466–474.

Suebnukarn, S. and Haddawy, P. (2007). Comet: A col-

laborative tutoring system for medical problem-based

learning. IEEE Intelligent Systems, 22(4):70–77.

APedagogicalScenarioLanguageforVirtualLearningEnvironmentbasedonUMLMeta-model-ApplicationtoBlood

AnalysisInstrument

307

Vanlehn, K., Lynch, C., Schulze, K., Shapiro, J. A., Shelby,

R., Taylor, L., Treacy, D., Weinstein, A., and Winter-

sgill, M. (2005). The andes physics tutoring system:

Lessons learned. Int. J. Artif. Intell. Ed., 15(3):147–

204.

Wenger, E. (1987). Artiﬁcial Intelligence and Tutoring Sys-

tems. Morgan Kaufmann, Los Altos, California.

Woolf, B. (1992). Building knowledge based tutors. In

Computer Assisted Learning : the 4th International

Conferences (ICCAL’92), pages 46–60.

CSEDU2014-6thInternationalConferenceonComputerSupportedEducation

308