![](bg3.png)
their tasks and simulated meetings, access and
execute training scenarios, submit deliverables at
prescribed milestones, accessing documents and
answering tests.
The server side contains the GSD scenarios, with
the information required for their execution,
including cultural and language knowledge.
The cultural knowledge base is based on the
existing literature of Hall (Hall, 1976) and Hosfstede
(Hofstede and Hofstede, 2005) considering the use
of titles, presentations and greetings, starting and
finishing conversations, motivation and rewards,
requests, negotiations, conflicts resolution, etc.
The language knowledge base contains the rules
for all the language pairs that the VC will use to
correct learners’ mistakes, such as those that take
place when English is used as a lingua franca: the
incorrect use of “false friends”, incorrect plural
formations, avoidance of passive forms, the absence
of the third person, the use of redundant
prepositions, the overuse of certain verbs, etc.
3.3 Scenarios
A training scenario consists of a VC and one or
more VAs, a set of documents and tests associated to
the problem to solve and a Meeting Workflow that
will guide the interaction with the student.
When the learner starts a virtual meeting, the VC
will present him/her the scenario and the VAs
involved. Learners will take a proactive role in the
meeting and the VC will guide them and will detect
cultural problems and the inappropriate use of
language, so the learners will learn from their
mistakes and learn to solve them. The scenario
finalizes when the learner complete all the
documents associated to the scenario and fill in a
questionnaire that can be automatically or semi-
automatically evaluated according to a template.
A scenario is defined by a set of phases that
define a small part of the conversation. Each phase
has a concrete conversational knowledge and also a
context specific language and cultural knowledge
which are used in that context for that scenario.
These phases are arranged setting up the
Scenario Workflow. We also employ decision points
in which the student will influence the execution
path of the scenario based on their textual responses
or actions. The phases can also store information
about its priority, and this serves to evaluate the
learners’ actions and the correctness of their
decisions.
Using the scenario definition, the VAs and the
VC guide the virtual meeting following a logical
sequence according to the learners’ actions. This
design of the scenarios makes it possible to simulate
profound and insightful conversations, avoiding
speech repetitions and out of context interventions.
At the same time, this model helps to reduce the
loss of time on incorrect phases, prevents the learner
from returning to a previous conversation, removes
unnecessary complexity in the scenario design and
helps to structure the cultural and language problems
based on the context of the conversation.
The Meeting Workflow can be created and edited
from the instructors’ interface. For each phase, the
instructors can introduce the conversational
knowledge, documents associated and the cultural
and language knowledge.
Phases of the workflow can also be simple or
composed what means that can contain other
workflows with the aim of organizing the
conversation with a high granularity level.
4 TEACHING REQUIREMENTS
ELICITATION
We have focused our first scenario on the Global
Requirements Elicitation (GRE) stage, since it is a
highly communicative process, particularly affected
by poor communication, and cultural and time
differences.
In our proposed scenario, Spanish learners,
playing the role of analysts, will have to
communicate with a virtual customer from United
States using English in order to elicit a set of
functional and non-functional requirements and
develop a specification document of the software.
For this purpose, we have prepared our scenario’s
culture and language knowledge bases to include the
typical problems or mistakes for that cultures and
languages.
Figure 1 shows a part of a Meeting Workflow for
our design, in which the VC will guide the learner in
the context of the meeting and the Virtual Customer
will answer the questions according to the associated
conversational knowledge.
During the interview, the learner will determine
the next phase in the workflow by chatting to the
VAs. Since real GRE interviews usually require
more than one meeting, the scenario can also store
several meeting’s workflows dealing with the same
subject in a different phase of the project.
TRAINING GLOBAL SOFTWARE DEVELOPMENT SKILLS THROUGH A SIMULATED ENVIRONMENT
273