Handing Pedagogical Scenarios Back over to Domain Experts: A

Scenario Authoring Model for VR with Pedagogical Objectives

Mathieu Risy

, Val

erie Gouranton

and Bruno Arnaldi

Univ. Rennes, INSA Rennes, Inria, CNRS, IRISA, France

Keywords:

Virtual Reality, Scenario, Pedagogy, Authoring, Welding Use Case.

Abstract:

Teachers and trainers make pedagogical decisions for their training courses, so why not do the same for Virtual

Reality (VR) training courses? Virtual Environments for Training (VETs) are becoming prominent educational

tools. However, VET models have yet to propose scenario authoring aligned with pedagogical objectives that

can account for the diversity of approaches available to teachers. This paper proposes a scenario authoring

model for VET that directly involves domain experts and validates their pedagogical objectives. In addition,

it proposes the coexistence of multiple pedagogical scenarios within the same VET, using three types of

scenarios. The validity of the model is then discussed using a VR welding application as a use case.

1 INTRODUCTION

Virtual Reality (VR) has shown promising educa-

tional beneﬁts (Dalgarno and Lee, 2010) manifested

by the growing number of Virtual Environments for

Training (VET). However, current VR authoring tools

remain difﬁcult for trainers (Ashtari et al., 2020).

While hardly standardized, VET design generally re-

quires a Domain Expert at its initiative, and an ex-

pert in VR development. In this paper, for the sake of

simplicity, the term Domain Expert refers to a person

knowledgeable in the learning content, and includes

the roles of both teachers and pedagogical experts.

A scenario, in the context of a Virtual Environ-

ment (VE), organizes the sequences of events that un-

fold at runtime. It characterizes the user’s actions and

interactions, and the VE entities’ behaviors. The ex-

ecution of a VR application always results in a sce-

nario unfolding, whether expressed explicitly or im-

plied. The explicit expression of the scenario serves

a monitoring purpose and allows it to manage the

VE. Scenario authoring is the writing of a machine-

interpretable scenario for the VE.

Current development practices leave scenario au-

thoring to the VR expert, with no standards taking

into account the Domain Expert’s Pedagogical Speci-

ﬁcations. We argue that it would be more efﬁcient for

https://orcid.org/0009-0004-6248-9988

https://orcid.org/0000-0002-9351-2747

https://orcid.org/0000-0002-2868-8826

the Domain Expert to write the VET scenario as the

person responsible for the learning content and ped-

agogical decisions. The term VET here includes all

types of higher education learning in VEs and does

not only apply to professional training. In this paper,

we propose a scenario authoring model (Section 3)

that meets the following criteria:

• Domain Experts are able to author Pedagogical

Scenarios according to their Pedagogical Objec-

tives.

• The model uses pedagogical principles that are fa-

miliar to Domain Experts and understandable by

the VET.

• The model allows for the deﬁnition of multiple

teaching methods under the same VET.

We then illustrate the model implementation using a

use case (Section 4): a welding training application

(Figure 1). We conclude by discussing the capabilities

of the model with Domain Experts (Section 5).

2 RELATED WORK

This section highlights related work on pedagogy

integration in VETs and VET models. It outlines

the main pedagogical approaches used in VETs. It

then presents the integration of their speciﬁcations in

machine-interpretable models. Finally, observation of

related VET models provides insight into the current

state of pedagogical scenario authoring.

Risy, M., Gouranton, V. and Arnaldi, B.

Handing Pedagogical Scenarios Back over to Domain Experts: A Scenario Authoring Model for VR with Pedagogical Objectives.

DOI: 10.5220/0012397800003660

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2024) - Volume 1: GRAPP, HUCAPP

and IVAPP, pages 103-114

ISBN: 978-989-758-679-8; ISSN: 2184-4321

103

(a) Error-spotting environment. (b) Welding station environment.

Figure 1: Environments of the VR welding training application use case.

2.1 Pedagogical Approaches in VET

The educational use of VR has extended beyond

the ﬁeld of research and is mature enough to allow

for a closer examination of its practices. Some au-

thors (Garz

on et al., 2020; Mikropoulos and Natsis,

2011; Radianti et al., 2020) state that a minority of

VETs declare a pedagogical approach. Among these

approaches, almost all take reference from the learn-

ing theory of constructivism (Piaget, 1950) or more

speciﬁc theories derived from its principles. Accord-

ing to constructivism learners actively construct their

knowledge based on their previous experiences. This

emphasizes the need for an active learning process

with contextualized learning material. Constructivist

principles lend well to VR learning affordances (Dal-

garno and Lee, 2010), notably allowing for engage-

ment in contextualized experiential learning.

While constructivism seems to be the most com-

monly used learning theory, some VETs rely on other

pedagogical approaches. The “VR nugget” based ap-

proach (Horst et al., 2022) combines generic stan-

dalone micro-learning patterns in VR to construct a

course that provides on-demand interactions. Con-

structivism is also less compatible with lecture-based

learning. Bowman et al. decided not to use a con-

structivist approach to illustrate abstract design prin-

ciples to students (Bowman et al., 1999). Finally, Ra-

dianti et al. reported on a VET using the learning the-

ory of behaviorism. This theory states that knowledge

is external and acquired through reinforcement of re-

warding or punishing consequences (Schunk, 2012).

2.2 Pedagogical Speciﬁcations in VET

Machine-interpretable Pedagogical Speciﬁcations in-

troduce pedagogical logic in virtual environments.

Using generic models, they can represent multiple

types of pedagogical decisions in the VET.

The structure of a course is a classic speciﬁcation

for planning learning sessions. However, VET mod-

els often represent only low-level activities, such as

tasks and actions (Johnson and Rickel, 1997; Ger-

baud et al., 2008; Buche et al., 2010). While this may

be sufﬁcient for short procedures and intervention, it

fails to meet the organizational needs for longer and

more modular learning interventions. Higher levels of

structure can be achieved by interconnecting multiple

learning activities (Udeozor et al., 2023) or using ex-

isting learning modeling speciﬁcations. IMS Learn-

ing Design (IMS LD) (Koper et al., 2003) is a well-

known speciﬁcation that has been integrated in VET

models (Marion et al., 2009). It uses a generic “Play-

Act-Learning Activity” structure to model learning

interventions of arbitrary length, while being generic

enough not to enforce a procedural type of learning.

Role speciﬁcations can be effective for modeling

multiple behavior types for pedagogical agents. They

allow for the assignment of different objectives and

interactions to multiple actors (Claude et al., 2014).

Roles can represent asymmetric interactions such as a

trainer-trainee collaboration or hierarchical positions.

This idea can be expanded to teams and organiza-

tions (Buche et al., 2004; Claude et al., 2015). Such

teams and organizations facilitate modiﬁcations to the

behavior of a group of actors with different roles but

similar objectives.

Recurring pedagogical patterns are also a form

of Pedagogical Speciﬁcations. “VR nuggets” (Horst

et al., 2022) use generic VR patterns such as manipu-

lating, decomposing, or tagging objects.

The expression of learning theory speciﬁcations in

VET models raises the question of their integration in

automated environments. The Game-Based Assess-

ment Framework (GBAF) (Udeozor et al., 2023) uses

Constructive Alignment (Biggs, 1999) as a design ap-

proach for VET. This constructivist theory proposes

to design learning activities from reliable and observ-

able descriptions of the Pedagogical Objectives. The

learner’s progress toward the objectives is measured

using Assessments derived from these criteria. Thus,

it provides easily implementable automatic assess-

ment that is relevant to the Domain Expert.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

104

2.3 Pedagogical Scenario Authoring

Pedagogical scenario authoring is the writing of ped-

agogical decisions in the VET scenario. The simplest

approach is to manually enable or disable pedagogi-

cal support such as instructions, guidance, and feed-

back. However, this approach lacks adaptability and

can result in the pedagogy blending with the code. As

a result, pedagogical scenarios typically use explicit

representations to facilitate their editing, analysis, and

reuse (Gerbaud et al., 2008).

Procedural scenarios provide a clear baseline of

the task sequences and goals to be achieved in or-

der to complete the procedure (Claude et al., 2014).

For example, pedagogical components can hint at

future steps (Gerbaud et al., 2008; Richard et al.,

2021). Conversely, deviations from the baseline fa-

vor the detection of errors. Procedural scenarios also

provide precise control over authorized actions. In

GVT (Gerbaud et al., 2008), the pedagogy engine

supports strategies to allow or block the actions avail-

able to users in real-time. The Domain Expert can

write adaptive strategies, but still requires a VR ex-

pert to implement them. Another approach is the

transmission of pedagogy through agents. In MAS-

CARET (Buche et al., 2004), agents form pedagogi-

cal strategies based on their role, knowledge, and the

learner’s progress. A domain model, in the form of

an ontology written by the Domain Expert, contains

semantic information about the environment and the

learning domain. While this provides a more ecolog-

ical environment, pedagogical authoring is scarcely

constrained by the Domain Expert. The types of reac-

tions of the agents are pre-planned, leaving only the

selection of the agents and their roles to the authoring.

The authoring of error responses is a powerful

pedagogical tool. HERA’s (Amokrane et al., 2008)

risk model and errors notify a rule-based pedagogical

module that allows or blocks scripted responses and

the unfolding of risk consequences. The PEGASE

model (Buche et al., 2010) introduces the authoring

of multiple pedagogical approaches in the form of

rules for agents’ attitudes and reactions to error de-

scriptors. However, deﬁning these rules still requires

coding knowledge. Procedural approaches have po-

tent capabilities, but action sequence descriptions lack

generality. They quickly become unsuitable for sce-

narios with low interest in representing intermediate

steps.

“Emergent approaches, by a clever modelling

of small behaviours of the world, allow new situa-

tions to arise” (Lanquepin et al., 2013). These ap-

proaches focus on a rich description of interactions

and constraints rather than modeling users’ progres-

sion (Claude et al., 2014). The HUMANS suite (Lan-

quepin et al., 2013) relies on reasoning based on se-

mantic information, causality, and domain knowledge

without using scripted events. It provides a reactive

ecological environment, but pedagogical authoring is

limited to deﬁning the initial situation and to non-

trivial authoring of agents’ reasoning.

Goal-based authoring approach uses environmen-

tal constraints and VET states as key scenario points

to be achieved (Porteous et al., 2010). It provides a

more familiar authoring format while retaining peda-

gogical control by using a higher-level description of

the VE. It relies on modeling shared properties (Bou-

ville et al., 2015) and causal information (Buche et al.,

2010). Goal-based approaches remain rare, although

Steve (Johnson and Rickel, 1997) is one of the best-

known VET examples. Steve is an agent that provides

guidance on a maintenance task using a set of goals

connected by causal links and preconditions.

Assessment-based approaches (Udeozor et al.,

2023) are more ﬂexible than goal-based approaches.

Progress within the scenario is based on complete

or partial validation of assessment. Such assess-

ments can concern, for example, following each step

of a procedure, setting the environment in a speciﬁc

state, attaining an objective, or providing a correct

answer. Assessment-based approaches propose a new

scenario layer that can observe both the progression of

goals and procedures. To the best of our knowledge,

there have been very few assessment-based proposals.

Udeozor et al. proposed the GBAF as a framework for

Domain Experts to author VETs aligned with Ped-

agogical Objectives. Assessment-based approaches

present interesting properties drawing on procedural

and goal-based approaches while being closer to the

pedagogical practices of Domain Experts.

2.4 Toward a New VET Framework

While great efforts have been made toward producing

credible agents and ecological environments, we have

found no satisfying solution to provide a complete

VET scenario authoring process for Domain Experts,

aligned with their objectives and supporting multi-

ple learning approaches. Only one model has pre-

cisely described support for multiple pedagogical ap-

proaches (Buche et al., 2010). While it offers precise

authoring control, it is inaccessible without a strong

background in coding and does not guarantee peda-

gogical coherence.

Procedural approaches lend themselves well to

pedagogical authoring, but lack generality. While

ontologies seem to provide powerful insight into the

learning domain, they do not ensure pedagogical con-

Handing Pedagogical Scenarios Back over to Domain Experts: A Scenario Authoring Model for VR with Pedagogical Objectives

105

SCENARIO AUTHORING

Write

MONITORING

Error

Scenarios

Trigger

Reference

Scenario

Pedagogical

Scenarios

Control Progression

VIRTUAL

ENVIRONMENT

PEDAGOGICAL SPECIFICATIONS

DOMAIN EXPERT

LEARNER

ValidateInform

Control

Observe

Interact

Assess

Observe

Figure 2: Overview of the components that make up the scenario authoring model.

trol. In addition, they are difﬁcult for Domain Experts

to ﬁnd and write. Emergent scenarios offer rich envi-

ronments but give little control to the Domain Expert.

Finally, goal- and assessment-based approaches have

rarely been implemented, but show promising capa-

bilities for a generic pedagogical authoring model.

Additionally, these approaches highlight interesting

authoring possibilities that may conform with the Do-

main Expert’s objectives and assessments.

3 SCENARIO AUTHORING

MODEL WITH PEDAGOGICAL

OBJECTIVES

We argue that the scenario of a VET should be writ-

ten by the Domain Expert. Much like Domain Ex-

perts, scenarios drive the learning session, monitor

progress, and can adapt both content and Pedagogi-

cal Guidance. We propose a scenario authoring model

that aligns the VET with the Pedagogical Objectives

of the Domain Expert. Furthermore, we introduce the

authoring of multiple pedagogical approaches within

the same VET in the form of a new pedagogical sce-

nario layer.

This section provides a general overview of the

model. It then details each type of scenario that com-

poses the model and their use of the Pedagogical

Speciﬁcations. Finally, it explains the monitoring as-

pect of the model.

3.1 Model Overview

Our scenario authoring model uses an assessment-

based approach (Udeozor et al., 2023). It allows for

an intuitive and ﬂexible authoring process based on

the Pedagogical Speciﬁcations of the Domain Expert.

In essence, Assessments represent monitoring speci-

ﬁcations that validate the Pedagogical Objectives and

inﬂuence progression within the scenario.

The scenario authoring model presents the learner

with a global learning scenario that combines three

types of scenarios: a Reference Scenario, Error Sce-

narios, and the active Pedagogical Scenario (Fig-

ure 2). The Reference Scenario (Section 3.2) de-

ﬁnes what the learner can do as long as the action

is not classiﬁed as an error. Most importantly, the

Assessments and objectives it contains deﬁne what

the learner is expected to do. Error Scenarios (Sec-

tion 3.3) represent typical deviations from the Refer-

ence Scenario that should not occur. Finally, the Ped-

agogical Scenario (Section 3.4) implements the peda-

gogical decisions. It provides Pedagogical Guidance

and adapt the learning experience. This distinction fa-

cilitates the writing and coexistence of multiple ped-

agogical approaches by separating the measurement

of the learner’s progress from the pedagogical deci-

sions. These decisions are added on top of the Refer-

ence Scenario by Pedagogical Scenarios.

The scenarios belongs to one of the four key inter-

acting components that compose the model. Figure 2

illustrates the communication between the Pedagogi-

cal Speciﬁcations, Scenario Component, Virtual En-

vironment (VE), and Monitoring Component.

• Pedagogical Speciﬁcations contain the deﬁnition

of the Pedagogical Objectives, the Assessments

required to achieve them, and the Learning Activi-

ties that support their achievement (Section 3.2.1).

• The Scenario Component is informed by the

Pedagogical Speciﬁcations and manages their im-

plementation in the VE. Its task is to orchestrate

events and interactions to help achieve the Peda-

gogical Objectives (Section 3.4).

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

106

• The learner interacts with a Virtual Environment

informed by the Pedagogical Speciﬁcations and

controlled by the Scenarios’ progression.

• The Monitoring Component observes the state

of the VET and provides the means to interact

with it outside of scenario execution (Section 3.5).

3.2 Reference Scenario

The Reference Scenario sets the expected standard of

the learner’s proﬁciency. It represents the achieve-

ment of the Pedagogical Objectives.

First, the Reference Scenario provides the means

of monitoring the learner’s progress toward the ob-

jectives. It considers a theoretical user who validates

the highest level of each Assessment without the need

for support or instructions. Consequently, it does not

force the learner to pass the Assessments. These de-

cisions are left up to the Pedagogical Scenarios. Us-

ing a welding training application as an example (see

Section 4), one Assessment of its Reference Scenario

evaluates weld quality. The reference does not change

whether a learner succeeds or fails the Assessment.

However, Pedagogical Scenarios can react to the As-

sessment in order to change the outcomes.

Second, the Reference Scenario deﬁnes the scope

of the actions that can be performed without being

classiﬁed as an error. In the case of a procedural ac-

tivity, the deﬁnition of possible actions might take the

form of branching paths. However, this representa-

tion is not suitable for every activity. In such cases,

a goal-based approach might be more appropriate. It

implicitly contains every action possible, unless they

trigger a state of error. We propose a hybrid approach

that uses the most appropriate scenario representation

for each speciﬁcation on a case-by-case basis.

The Reference Scenario uses automated Assess-

ments to monitor the learner’s actions and mea-

sure progression toward the Pedagogical Objectives.

Therefore, the Pedagogical Speciﬁcations need to

be machine-interpretable. Consequently, we de-

cided to introduce the Constructive Alignment prin-

ciple (Biggs, 1999) in our model. It allows for the

alignment of the VET’s behavior with the Domain

Expert’s needs (Udeozor et al., 2023). It has the addi-

tional beneﬁt of facilitating discussions between Do-

main Experts and VR experts.

3.2.1 Constructive Alignment

Constructive Alignment (Biggs, 1999) is an educa-

tional principle for the design of learning interven-

tions and programs. Its core proposal is that the

learner’s progression toward Pedagogical Objectives

Prepare

Guide

design

Allow

achievement

Measure

validation

levels

1. ILOS

(OBJECTIVES)

3. LEARNING

ACTIVITIES

2. ASSESSMENTS

Guide

design

Validate objectives using

Figure 3: Constructive Alignment, adapted from the works

of John Biggs (Biggs, 1999) and Elie Milgrom.

STRUCTURE

Prerequisites,

Levels

Prerequisites

Sub-objectives

ILO

Training

Course

Module

Pedagogical

Objectives

Learning

Taxonomy

Assessment

Learning

Activity

Figure 4: Pedagogical Speciﬁcations of the scenario author-

ing model.

should be assessed using reliable and observable cri-

teria, known as Intended Learning Outcomes (ILOs).

ILOs are deﬁned as “statements, written from the stu-

dents’ perspective, indicating the level of understand-

ing and performance they are expected to achieve as a

result of engaging in the teaching and learning expe-

rience” (Biggs and Tang, 2011). They derive directly

from the objectives and are used as core elements to

create Learning Activities. Constructive Alignment

requires three steps (see Figure 3):

1. Deﬁnition of the ILOs.

2. Deﬁnition of the Assessments based on the de-

scriptions of the ILOs.

3. Designing of the Learning Activities, using As-

sessments to validate the ILOs.

Our model directly integrates these three key elements

into the Pedagogical Speciﬁcations Component (Fig-

ure 4). They provide information relevant to both Do-

main Experts and VET scenarios. The Constructive

Alignment process is illustrated in Section 4.3 using

a welding training application as a use case.

3.2.2 ILOs

An ILO describes one or multiple levels of proﬁ-

ciency. It is written using Observable Action Verbs

(OAVs) to deﬁne precisely how its statements are to

be evaluated. Each ILO level is related to a learning

Handing Pedagogical Scenarios Back over to Domain Experts: A Scenario Authoring Model for VR with Pedagogical Objectives

107

taxonomy level to express the type of cognitive task

involved. Bloom’s revised taxonomy (Anderson and

Krathwohl, 2001) and SOLO taxonomy (Biggs and

Collis, 1982) are two well-known examples. Using

the same example of a welding training application,

Table 1 illustrates an ILO associated with the Peda-

gogical Objective of learning welding practice. The

Reference Scenario ensures the implementation of ev-

ery ILO under the Pedagogical Speciﬁcations.

Table 1: Example of an ILO description. Basic welding

practice for a welding training application.

ILO - Basic welding practice

Time: After preparation of the welding station

and application of safety measures.

Level 1

Taxonomy (Bloom): Remember

Taxonomy (SOLO): Unistructural

Description: The learner is able to:

- Move the welding torch in a straight line.

- Maintain constant and appropriate speed,

height, and angle.

Level 2

Taxonomy (Bloom): Analyze

Taxonomy (SOLO): Multistructural

Description: The learner is able to:

- Correct the welding parameters during practice

if they deviate from the standard.

Level 3

Taxonomy (Bloom): Evaluate

Taxonomy (SOLO): Relational

Description: The learner is able to:

- Evaluate the weld quality visually after practice.

- Correct their next practice using the identiﬁed

errors.

3.2.3 Assessments

Assessments naturally derive from the observable

statements of the ILOs. While they can be used for

grading purposes, their primary objective is to moni-

tor the learner’s progression. They are not required to

be visible to the learner and are often implicitly inte-

grated within the VE. An Assessment derives from a

single ILO and contributes to its validation. For ex-

ample, two Assessments can be derived from the ILO

in Table 1. The ﬁrst compares weld and gesture pa-

rameter values to the standard. The second evaluates

the evolution of weld quality over time. The valida-

tion of an Assessment is set by the Domain Expert

as a percentage of the progression of the ILO. Some

Assessments may even allow for partial validation.

3.2.4 Learning Activities

Learning Activities are deﬁned in the ﬁnal step of the

Constructive Alignment. They describe the learning

context for the achievement of the ILOs and how the

Assessments are applied. Let us take the example of

a Learning Activity relating to the ILO described in

Table 1 and other practice-oriented ILOs. This ac-

tivity describes the means to achieve ILOs and their

Assessments in the form of a contextualized welding

environment that contains several metal plates. It de-

scribes the interactions and behaviors for the weld-

ing of metal plates according to the conﬁgurations de-

scribed by the ILOs.

Learning Activities represent the basic units of a

scenario that a learner can complete. Pedagogical

Speciﬁcations include a Structure speciﬁcation (Fig-

ure 4) to help group and ﬁlter these activities. It uses

three hierarchical levels “Training Course-Modules-

Learning Activities”, where the Learning Activity is

the lowest. The Structure does not provide the order

of execution of its sub-parts. This pedagogical deci-

sion is left up to each Pedagogical Scenario.

3.3 Error Scenarios

Error Scenarios represent typical deviations from the

Reference Scenario. For the purpose of comple-

menting the Reference Scenario, the deﬁnition of

an error includes any behavior that demonstrates the

non-mastery of an ILO. This may include hesita-

tions, or taking too long without committing any mis-

takes. The Domain Expert is responsible for deﬁning

whether part of a scenario part belongs to the Refer-

ence Scenario, or is represented as an Error Scenario.

Error Scenarios have two roles. First, they notify

the VET when a state of error is triggered. This allows

the active Pedagogical Scenario and the Monitoring

Component to act where appropriate. Second, they

can represent either or both error consequences and

corrective measures that allow the learner to return to

the Reference Scenario. For example, in the case of

a welding training application, welding without open-

ing the gas bottle will trigger an Error Scenario. This

notiﬁes the VET of the error and represents the conse-

quences, namely more sparks and poor weld quality.

As the error is not fatal, it allows the learner to stop

welding, open the gas bottle, and return to the Refer-

ence Scenario.

Triggering an Error Scenario does not necessarily

result in its execution. Pedagogical Scenarios handle

the VET’s reaction to the triggering of an Error Sce-

nario, based on the pedagogical decisions they repre-

sent.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

108

STRUCTURE

SCENARIOS

Interact

Prerequisites, Levels

Prerequisites

Sub-objectives

Error Scenario

ILO

Training Course

Learning

Activity

Module

Pedagogical

Objectives

Learning

Taxonomy

VIRTUAL

ENVIRONMENT

Reference

Scenario

Assessments

Scenario Structure

Pedagogical

Guidance

Interact

Observe

Interact

DOMAIN EXPERT

Observe

Learner

Control

MONITORING

Scenario ILOs

Assessments

Pedagogical

Scenario

Assessments

Scenario Structure

Inform

Scenario ILOs

Assessment

Class

Triggers

Instances of Class

Figure 5: Detailed view of the scenario authoring model.

3.4 Pedagogical Scenarios

Pedagogical Scenarios represent the pedagogical de-

cisions made to allow the learner to achieve all or part

of the Pedagogical Objectives. Because of this status,

they complement and have control over the Reference

Scenario and Error Scenarios.

Each Pedagogical Scenario contains a subset of

the Reference Scenario’s ILOs, Assessments, and

Structure (Figure 5). This subset represents the focus

of the Pedagogical Scenario. The Structure can auto-

matically ﬁlter out its sub-parts that are not associated

with the subset of ILOs. It narrows down the Learning

Activities and Assessments that the learner will inter-

act with, either by blocking or not monitoring parts of

the Reference Scenario. For example, in the case of a

welding training application (Section 4), the Domain

Expert may wish to focus solely on safety as part of an

introductory course. The associated Pedagogical Sce-

nario thus contains only activities relevant to safety

and does not evaluate welding practice. In addition,

it provides relevant guidance to emphasize the role of

each welding safety practice and piece of equipment,

and to prevent dangerous behavior.

3.4.1 Structure Ordering

Each Pedagogical Scenario provides the ordering of

the Learning Activities contained in its Structure. By

default, the Reference Scenario does not provide an

order in which to execute the Learning Activities.

The ordering of activities is not necessarily lin-

ear. For example, one may establish a pedagogical

decision to provide the learner with a choice every

time, or only in some cases. In non-linear scenar-

ios, the deﬁnition of pedagogical prerequisites is an

important speciﬁcation that allows the learner to go

from one Learning Activity to the next in a logical

order. Such prerequisites may include the validation

of other Learning Activities, or the validation of an

Assessment above certain a threshold.

Functional prerequisites are indicative prerequi-

sites regarding the state of the VE. For example,

the “welding practice” Learning Activity requires the

“preparation of the welding station” Learning Activ-

ity to be performed beforehand. However, the Do-

main Expert can always decide that the latter Learn-

ing Activity is not to be performed by the learner, but

is assumed to be performed correctly. In such case,

the application must handle the prerequisites directly

instead of the learner, in order to begin the “welding

practice” Learning Activity in the correct conditions.

3.4.2 Error Handling

Pedagogical Scenarios handle the triggering of Er-

ror Scenarios according to pedagogical decisions.

Through this process, a Pedagogical Scenario can

modify or block the execution of an Error Scenario.

Taking the example of welding training detailed in

Section 4, “welding without a welding helmet” is an

error that deviates from an ILO on compliance with

safety measures during practice. Indeed, this error

represents a serious risk for the learner’s eyesight in

real-life situations. This error can be handled in mul-

tiple ways. The scenario can prevent the error from

happening by blocking the interactions susceptible to

causing it. It can also display an informative message

Handing Pedagogical Scenarios Back over to Domain Experts: A Scenario Authoring Model for VR with Pedagogical Objectives

109

when the Error Scenario is triggered, or, it can illus-

trate the consequences that would unfold. In the latter

case, the Pedagogical Scenario manages the error by

deciding not to intervene in its resolution.

3.4.3 Pedagogical Guidance

We deﬁne Pedagogical Guidance as elements that ori-

ent the learning process and are superﬂuous to an ex-

pert in the learning domain. They encompass ele-

ments such as instructions, feedback, or interaction

management. Consequently, they rely on the Peda-

gogical Speciﬁcations to inform their content. Peda-

gogical Guidance is used as an interface between Ped-

agogical Scenarios and the VE.

The activation of Pedagogical Guidance depends

on the scenario. For example, it may be active at all

times, triggered by an Assessment, or react to the trig-

gering of an Error Scenario. Conversely, interacting

with Pedagogical Guidance may trigger progression

of the scenario. It should be noted that the actions

of Pedagogical Guidance are not only additive. Pre-

venting actions and events is also an important role of

Pedagogical Guidance. Most notably, this might take

the form of blocking actions in order to prevent errors.

Pedagogical Guidance plays an intermediate ped-

agogical role, as illustrated in the detailed view of the

model (Figure 5). It is a key part of Pedagogical Sce-

narios, but can also be triggered by the Domain Ex-

pert through the Monitoring Component. In addition,

fostering learner autonomy is a recurring Pedagogical

Objective; therefore Pedagogical Guidance can also

be requested directly by the learner to provide control

over the learning process.

3.5 Monitoring

The Monitoring Component allows an external user to

monitor the VET and act upon it. The characteristics

of VETs make it difﬁcult for the Domain Expert to

monitor the progress of a learner using a VR device.

Providing a video feed might not be sufﬁcient for cor-

rect monitoring and does not scale well for multiple

learners. Thus, the monitoring process needs to pro-

vide other information. The Monitoring Component

observes the state of the VE, progression within the

scenarios, and the validation of ILOs, as illustrated

in Figure 5. It provides the Domain Expert with in-

formation that directly derives from the measurement

of the learner’s progress. In addition, the Monitor-

ing Component can also provide help and guidance to

the learner at runtime, without leaving the VE. The

Monitoring Component can interact directly with the

VE, but also provides the means of triggering Assess-

ments and Pedagogical Guidance if needed. It bridges

the gap between the immersive virtual environment

and the “outside world”. It introduces possibilities

for asymmetric modes of collaboration due to its abil-

ity to monitor learners’ performance and act upon the

VE. Such collaborations could take place between the

learner and the Domain Expert or between multiple

learners.

4 USE CASE

To assess the viability of our model, we applied it

to a real use case of a VR training course. The use

case was designed in collaboration with Domain Ex-

perts who teach introductory courses on welding the-

ory and practice to undergraduate students. These

courses constitute an introduction to industrial pro-

cesses and risk management as part of an engineering

curriculum.

The Domain Experts are involved in a develop-

ment project to create a VR training application that

helps teach safety and welding basics before actual

welding practice. In parallel with its development,

we had the opportunity to implement our model’s ap-

proach within this application and gather feedback

from the Domain Experts.

4.1 Welding Training Application in VR

The welding training application teaches safety prac-

tices and introductory notions for Metal Inert Gas

(MIG) welding – a type of Gas Metal Arc Weld-

ing (GMAW) that uses an electric arc to provoke the

fusion of metal. The application contains an error-

spotting environment (Figure 1a) and a welding prac-

tice environment (Figure 1b). The error-spotting en-

vironment is designed to help identify clothing that is

unsuited for safe welding practice, and risks posed by

loose hair. The second environment features a func-

tional MIG welding station, welding material, and

safety equipment. The learner can interact with Per-

sonal Protective Equipment (PPE), set up the welding

station, and perform a weld on a simple metal plate.

4.2 Implementation

The welding training application is being developed

with Unity 2021. This preexisting code basis fea-

tures ecological interactions with welding material

and two environments (Figure 1). Additional Peda-

gogical Guidance was developed and integrated into

the environment to answer the needs of the Pedagog-

ical Scenarios.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

110

STRUCTURE

ILO

Training

Course

Learning

Activity

Module

Pedagogical

Objectives

Assessment

Figure 6: Pedagogical Speciﬁcations writing protocol.

The model was implemented using Xareus

, a

Petri net-based scenario engine (Bouville et al., 2015;

Claude et al., 2014), and C# classes. The Pedagogical

Speciﬁcations (Figure 4) are represented in C# and

instantiated from the application’s Pedagogical Spec-

iﬁcations written in natural language. Their instanti-

ation allows the construction of objects representing

the focus of a Pedagogical Scenario (Figure 5). The

scenarios, including the Learning Activities content,

were created with Xareus’ visual authoring interface

for no-code approaches. In addition, Xareus’ sen-

sors/effectors logic was extended with the automatic

activation of the relevant scenario sections based on

the content of a given scenario focus parameter.

4.3 Model Application

Our scenario authoring model was applied to the

welding training application in accordance with the

Domain Expert’s authoring process (Figure 6). First,

this requires the deﬁnition of the Pedagogical Spec-

iﬁcations. Second, these speciﬁcations are used to

author the Reference Scenario and Error Scenarios.

Finally, two Pedagogical Scenarios were proposed to

show the possibilities of our model.

4.3.1 Reference Scenario

The Reference Scenario uses Pedagogical Speciﬁca-

tions to align the VET with the Pedagogical Objec-

tives. This allows for the automation of the relations

between the speciﬁcations and facilitates authoring.

For example, selecting an ILO subset disables the

Learning Activities that do not serve to achieve them.

The speciﬁcations were written using a six-step proto-

col (Figure 6) based on Constructive Alignment (Fig-

ure 4). Following the validation of the speciﬁcations

https://xareus.insa-rennes.fr/

by Domain Experts, minor corrections were made to

the welding practice ILOs.

Five ILOs were identiﬁed for this application.

• ILO

: Identiﬁcation of safe welding clothing

among a panel of virtual humans.

• ILO

: Choosing and wearing suitable PPE.

• ILO

: Preparation of the welding station.

• ILO

: Application of safety instructions during

welding practice.

• ILO

: Basic welding practice.

A detailed description of ILO

is presented in Table 1.

Each ILO contains a validation value made accessible

to the Assessments and the Monitoring Component.

The Reference Scenario then exposes each VET’s

aspect monitored by the Assessments of the ILOs. For

example, the ILO

Assessment records the interac-

tions of the learner with PPE. It expects the apron,

welding helmet, and gloves to be in an “equipped”

state and to have been equipped in this order. Thus,

the Reference Scenario provides a procedural descrip-

tion of the actions and access to the states of the PPE.

Finally, three Learning Activities (LA) were de-

vised using the ILOs and Assessments.

• LA

: Identiﬁcation of safe welding clothing.

• LA

: Preparation of the welding station in accor-

dance with the safety measures.

• LA

: Basic welding practice.

Each Learning Activity knows the Assessments it

implements and thus transitively knows the ILOs it

serves to achieve. The Reference Scenario segments

the Assessments’ validations into sections represent-

ing their Learning Activity. Each section can be easily

enabled or disabled without impacting the others.

4.3.2 Error Scenarios

In this use case, most Error Scenarios are used to no-

tify the VET. For example, in LA

touching a burning

metal plate is a fatal error. Representing its conse-

quences was deemed unnecessary, but it does not pre-

vent Pedagogical Scenarios from handling the Error

Scenario using Pedagogical Guidance.

Some cases are better suited to represent the con-

sequences of errors. If the welding action is triggered

while the gas bottle is in the “closed” state, the Error

Scenario increases the production of sparks and de-

grades the quality of the weld. In addition, if the state

of the gas bottle changes to “open”, the scenario stops

and allows the learner to return to the Reference Sce-

nario. This case supports the enabling or disabling of

consequences based on the needs of the Pedagogical

Scenarios.

Handing Pedagogical Scenarios Back over to Domain Experts: A Scenario Authoring Model for VR with Pedagogical Objectives

111

4.3.3 Pedagogical Scenarios

This use case contains two Pedagogical Scenarios

to illustrate the capabilities of the model – a safety-

focused scenario and a practice-focused scenario.

The safety-focused scenario introduces safety no-

tions to the learner using a signiﬁcant amount of Ped-

agogical Guidance. It makes use of the three Learn-

ing Activities. At the end of each activity, the Ped-

agogical Scenario informs the Reference Scenario of

the next activity to be performed. This scenario pro-

vides instruction and corrective feedback to validate

safety ILOs. For example, in LA

, it triggers instruc-

tional guidance in the form of a checklist, highlights

PPE entities, and provides dynamic textual feedback

if components of PPE are missing or equipped in the

wrong order. This Pedagogical Scenario does not

evaluate ILO

“Preparation of the welding station”.

Consequently, the scenario handles the preparation of

the welding station by changing the state of the rele-

vant elements. In LA

, the Pedagogical Scenario fea-

tures the handling of Error Scenarios. The learner can

test the welding torch while dangerous interactions

are blocked with explanatory feedback. This prevents

the activation of the torch while the welding helmet is

not lowered, for example.

The practice-focused scenario is oriented toward

teaching welding practice and the preparation of a

MIG welding station. It uses only LA

and LA

and

disables LA

. In this scenario, LA

explicitly eval-

uates the preparation of the welding station and im-

plicitly expects compliance with safety measures. It

uses Pedagogical Guidance to prevent welding until

all preparation steps have been completed. Addition-

ally, it does not indicate nor enforce safety-related As-

sessments, although failure to apply safety measures

triggers an alert via the Monitoring Component. Fi-

nally, LA

uses dynamic Pedagogical Guidance in the

form of a “ghost” of the welding torch (Figure 7) to

demonstrate the expected speed, height, and angle pa-

rameters. It helps with the achievement ILO

(Ta-

ble 1). This activity also provides the learner with the

possibility to display informative graphics as optional

guidance.

4.4 Discussion with Domain Experts

This section reports feedback on the acceptance of the

authoring model from three Domain Experts. Two

of them are welding teachers with no VR experience,

and one is a pedagogical expert familiar with VR de-

velopment using no-code approaches. The discussion

followed a qualitative guided interview. Before and

after the model presentation, Domain Experts were

Figure 7: Pedagogical Guidance in the form of a welding

torch ghost and weld quality feedback.

asked to express the pedagogical objectives, speciﬁ-

cations, and scenarios they considered necessary to

create the use case. Then, they were asked about their

reported conﬁdence and motivation to use the compo-

nents of the model. Finally, they were given the op-

portunity to comment on the relevancy of the model.

Domain Experts reported feeling conﬁdent in

writing each type of Pedagogical Speciﬁcations (Fig-

ure 6). One reported a strong interest in Construc-

tive Alignment, as it allows for introspection on the

learning material and evaluations. It should be noted

that two experts already had intermediate knowledge

in terms of describing learning outcomes with OAVs.

When asked about directly authoring a scenario,

Domain Experts reported irresolute conﬁdence in

writing and high conﬁdence in editing. Welding

teachers did not feel conﬁdent writing Pedagogical

Scenarios without the support of a pedagogical ex-

pert. They reported having a “one scenario per stu-

dent” approach, and not being in a scenario authoring

mindset. Conversely, the pedagogical expert felt con-

ﬁdent in writing scenarios using our model and ex-

pressed interest in using it for professional projects.

Higher levels of conﬁdence were reported for edit-

ing pre-existing Pedagogical Scenarios. Domain Ex-

perts were interested in the possibility of selecting and

potentially editing a scenario just before a training

session, for example, by choosing only the relevant

Learning Activities and Assessments.

The capabilities of a model with multiple Peda-

gogical Scenarios were of great interest to the Domain

Experts, particularly to allow for adaptation to differ-

ent categories of learners. When presented with this

possibility, they suggested using Pedagogical Scenar-

ios to provide more guidance to foreign students or to

gradually increase the level of difﬁculty.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

112

5 DISCUSSION

This section discusses the capabilities of the model

based on the feedback from Domain Experts (Sec-

tion 4.4). They reported high motivation and feeling

conﬁdent in using the model in its entirety, showing a

promising acceptance of its principles.

Domain Experts’ conﬁdence in authoring scenar-

ios highlighted potential difﬁculties in understanding

VR scenarios without pedagogical expertise or exam-

ples. On the other hand, Pedagogical Speciﬁcations

authoring received a high level of conﬁdence. This

shows the importance of the model’s speciﬁcations to

provide a common language understandable by both

Domain Experts and VR experts. While complete au-

thoring might require additional guidance, Domain

Experts exhibited a keen interest and conﬁdence in

editing pre-existing scenarios. This interest extends

after the development phase, suggesting that the VET

can become a more time-resilient educational tool.

This work assumes the existence of a simpliﬁed

authoring interface for Domain Experts. Such an in-

terface is not the focus of this paper as its implemen-

tation would require its own pedagogical decisions.

Nonetheless, we consider it beneﬁcial for scenario

editing and generating code-based speciﬁcations.

Domain Experts identiﬁed the adaptation to differ-

ent learner categories as an advantage of using multi-

ple Pedagogical Scenarios. Extending the model with

the authoring of automatic adaptation would provide

Domain Experts with adaptive authoring tools that

can inspire new ways of thinking about teaching.

6 CONCLUSION

This paper proposed a scenario authoring model for

VET to respond to a lack of solutions providing Do-

main Experts with hands-on control over the peda-

gogical scenario authoring of VETs. It described how

the integration of Constructive Alignment principles

can shape VET design to guarantee the validation of

Pedagogical Objectives. It then proceeded to explain

the three types of scenarios used by the model. This

separation of scenario roles allows for the authoring

of multiple Pedagogical Scenarios by separating the

learner’s Assessment from pedagogical decisions.

Following the description of the model, its capa-

bilities were illustrated through a use case pertaining

to a welding training application. While an overall as-

sessment of the model is a difﬁcult task, discussions

with Domain Experts revealed high levels of interest

in its use. In particular, for its learner progress mon-

itoring, authoring of multiple pedagogical scenarios,

and editing capabilities.

Future work should aim to expand scenario adap-

tation to learners. Notably, the addition of dynamic

pedagogical scenario changes and scenario blending

should be investigated. Further integration of Ped-

agogical Speciﬁcations could lead to the automated

generation of a VET skeleton for the beneﬁt of VR de-

velopment. Finally, the high-level nature of the model

opens up its extension to Collaborative Virtual Envi-

ronments and to the eXtended Reality (XR) contin-

uum.

ACKNOWLEDGEMENTS

This work has beneﬁted from state aid managed by

the “Agence Nationale de la Recherche” under the

“Investissements d’Avenir” program under the refer-

ence ANR-21-DMES-0001.

We want to thank the AIR Project pedagogical en-

gineer and INSA Rennes’ teachers for their participa-

tion in the model discussion.

REFERENCES

Amokrane, K., Lourdeaux, D., and Burkhardt, J.-M. (2008).

Hera: Learner tracking in a virtual environment. Int.

J. Virtual Real., 7(3):23–30.

Anderson, L. W. and Krathwohl, D. R. (2001). A taxon-

omy for learning, teaching, and assessing: A revision

of Bloom’s taxonomy of educational objectives: com-

plete edition. Addison Wesley Longman, Inc., New

York, complete edition.

Ashtari, N., Bunt, A., McGrenere, J., Nebeling, M., and

Chilana, P. K. (2020). Creating augmented and virtual

reality applications: Current practices, challenges,

and opportunities. In Proceedings of the 2020 CHI

Conference on Human Factors in Computing Systems,

page 1–13, Honolulu HI USA. ACM.

Biggs, J. (1999). Teaching for Quality Learning at Univer-

sity: What the Student Does. SRHE and Open Univer-

sity Press imprint. Society for Research into Higher

Education.

Biggs, J. B. and Collis, K. F. (1982). Evaluating the qual-

ity of learning: the SOLO taxonomy (structure of the

observed learning outcome). Educational psychology.

Academic Press, New York.

Biggs, J. B. and Tang, C. S.-k. (2011). Teaching for quality

learning at university: what the student does. SRHE

and Open University Press imprint. McGraw-Hill, So-

ciety for Research into Higher Education & Open Uni-

versity Press, Maidenhead, England New York, NY,

4th edition.

Bouville, R., Gouranton, V., Boggini, T., Nouviale, F., and

Arnaldi, B. (2015). #ﬁve : High-level components for

Handing Pedagogical Scenarios Back over to Domain Experts: A Scenario Authoring Model for VR with Pedagogical Objectives

113

developing collaborative and interactive virtual envi-

ronments. In 2015 IEEE 8th Workshop on Software

Engineering and Architectures for Realtime Interac-

tive Systems (SEARIS), page 33–40, Arles, France.

IEEE.

Bowman, D. A., Hodges, L. F., Allison, D., and Wineman,

J. (1999). The educational value of an information-

rich virtual environment. Presence: Teleoperators and

Virtual Environments, 8(3):317–331.

Buche, C., Bossard, C., Querrec, R., and Chevaillier, P.

(2010). Pegase: A generic and adaptable intelligent

system for virtual reality learning environments. In-

ternational Journal of Virtual Reality, 9(22):73–85.

Buche, C., Querrec, R., de Loor, P., and Chevaillier, P.

(2004). Mascaret: A pedagogical multi-agent sys-

tem for virtual environment for training. International

Journal of Distance Education Technologies, 2:41–61.

Claude, G., Gouranton, V., and Arnaldi, B. (2015). Roles

in collaborative virtual environments for training. In

Imura, M., Figueroa, P., and Mohler, B., editors, Pro-

ceedings of International Conference on Artiﬁcial Re-

ality and Telexistence Eurographics Symposium on

Virtual Environments , pages 1–8, Kyoto, Japan.

Claude, G., Gouranton, V., Bouville Berthelot, R., and Ar-

naldi, B. (2014). Short paper: #seven, a sensor ef-

fector based scenarios model for driving collaborative

virtual environment. In Nojima, T., Reiners, D., and

Staadt, O., editors, ICAT-EGVE, International Con-

ference on Artiﬁcial Reality and Telexistence, Euro-

graphics Symposium on Virtual Environments, page

1–4, Bremen, Germany.

Dalgarno, B. and Lee, M. J. W. (2010). What are the learn-

ing affordances of 3-d virtual environments? British

Journal of Educational Technology, 41(1):10–32.

Garz

on, J., Kinshuk, Baldiris, S., Guti

errez, J., and Pav

on,

J. (2020). How do pedagogical approaches affect the

impact of augmented reality on education? a meta-

analysis and research synthesis. Educational Research

Review, 31:100334.

Gerbaud, S., Mollet, N., Ganier, F., Arnaldi, B., and Tis-

seau, J. (2008). Gvt: a platform to create virtual

environments for procedural training. In 2008 IEEE

Virtual Reality Conference, page 225–232, Reno, NV,

USA. IEEE.

Horst, R., Naraghi-Taghi-Off, R., Rau, L., and Do-

erner, R. (2022). Authoring with virtual reality

nuggets—lessons learned. Frontiers in Virtual Real-

ity, 3:840729.

Johnson, W. L. and Rickel, J. (1997). Steve: an animated

pedagogical agent for procedural training in virtual

environments. ACM SIGART Bulletin, 8(1–4):16–21.

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

learning design speciﬁcation (version 1.0).

Lanquepin, V., Carpentier, K., Lourdeaux, D., Lhommet,

M., Barot, C., and Amokrane, K. (2013). Humans: a

human models based artiﬁcial environments software

platform. In Proceedings of the Virtual Reality Inter-

national Conference: Laval Virtual, page 1–8, Laval

France. ACM.

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 Proceedings of the First International Con-

ference on Computer Supported Education, volume 1,

page 253–258, Lisboa, Portugal. SciTePress - Science

and Technology Publications.

Mikropoulos, T. A. and Natsis, A. (2011). Educational

virtual environments: A ten-year review of empiri-

cal research (1999–2009). Computers & Education,

56(3):769–780.

Piaget, J. (1950). The psychology of intelligence. Rout-

ledge, New York.

Porteous, J., Cavazza, M., and Charles, F. (2010). Applying

planning to interactive storytelling: Narrative control

using state constraints. ACM Transactions on Intelli-

gent Systems and Technology, 1(2):1–21.

Radianti, J., Majchrzak, T. A., Fromm, J., and Wohlge-

nannt, I. (2020). A systematic review of immersive

virtual reality applications for higher education: De-

sign elements, lessons learned, and research agenda.

Computers & Education, 147:103778.

Richard, K., Havard, V., His, J., and Baudry, D. (2021). In-

tervales: Interactive virtual and augmented framework

for industrial environment and scenarios. Advanced

Engineering Informatics, 50:101425.

Schunk, D. H. (2012). Learning theories: an educational

perspective. Pearson, Boston, 6th ed edition.

Udeozor, C., Chan, P., Russo Abeg

ao, F., and Glassey, J.

(2023). Game-based assessment framework for vir-

tual reality, augmented reality and digital game-based

learning. International Journal of Educational Tech-

nology in Higher Education, 20(1):36.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

114