A Virtual Agent Toolkit for Serious Games Developers

Full text

A Virtual Agent Toolkit for Serious Games
Developers
Samuel Mascarenhas, Manuel Guimar˜
aes,
Rui Prada, Jo˜
ao Dias and Pedro A. Santos
INESC-ID and Instituto Superior T´
ecnico
Universidade de Lisboa
2744-016 Porto Salvo, Portugal
samuel.mascarenhas@gaips.inesc-id.pt,
manuel.m.guimaraes@ist.utl.pt
rui.prada@tecnico.ulisboa.pt,
pedro.santos@tecnico.ulisboa.pt,
joao.dias@tecnico.ulisboa.pt
Kam Star, Ben Hirsh
and Ellis Spice
PlayGen
8-9 Talbot Court, London, UK
kam@playgen.com,
ben@playgen.com,
ellis@playgen.com
Rob Kommeren
Stichting Praktijkleren
3821 AR Amersfoort, NL
r.kommeren@stichtingpraktijkleren.nl
Abstract—The design of serious games requires developers to
tackle pedagogical challenges calling for advanced solutions that
the entertainment industry might deem too risky to pursue. One
such challenge is the creation of autonomous socially intelligent
characters with whom players can practice different social skills.
Although there are several architectures in the field of virtual
agents that are designed specifically to enable more human-like
interactions, they are still not widely adopted by game studios
that develop serious games, in particular for learning. In this
paper, we present a virtual agent toolkit that was specifically
developed with the intent of making agent-based solutions more
accessible and reliable to game developers. To this end, a
collaborative effort was established with a game studio that has
used the toolkit to develop two different serious games. Among
other advantages, the toolkit facilitated the inclusion of a dynamic
model of emotions that affects not just how the character looks
and acts but also how the player’s performance is determined.
Index Terms—serious games, virtual agents, authoring tools,
interactive storytelling, affective computing
I. INTRODUCTION
The industry of video games has seen tremendous growth
to the point that the budget for highly anticipated games
can surpass the cost of big Hollywood films [7]. This led
to extensive development times and quite large development
teams and corresponding high expectations from the players
[13]. On one hand, this state of affairs has enabled the creation
of very detailed game worlds with stories and characters that
players find very engaging to interact with. But, on the other
hand, the huge risk that is now associated with failing to meet
the expectations of players has led the industry to primarily
focus on what has been known to work in the past. This is also
then reflected in the available development tools, with popular
game engines like Unity1being primarily designed to support
the typical requirements and methods used in entertainment
games that were previously successful. As a result, game
developers that are interested in developing games with more
unique characteristics or requirements, which is often the case
1https://unity3d.com/
for pedagogical games, usually find themselves having to
spend a significant amount of time in developing their own
tools and methods.
The serious games industry is growing as well, supported
by the continuous research on the potential in using games for
other purposes than just entertainment [10], [12], [21]. Serious
games can be used to train and teach players on various
subjects (e.g. math fractions [14], logic operators [16]) or raise
awareness on social issues (e.g. sustainability [17], cultural
diversity [4], bullying [20]). In fact, one of the more interesting
aspects in developing games that are designed to teach is that
their design is centered around pedagogical challenges. As
such, even if the game is very engaging for players it will still
fail to achieve its purpose if it does not have a pedagogical
outcome. But, in turn, the game might have great pedagogical
content but fail to deliver it in an engaging manner. One of
the important aspects that make players engaged in a game
world is the appeal of its characters. Particularly, non-player
characters provide the opportunity for the player to engage
in social interactions in a safe environment and within the
confines of the game rules and structures. From a training
perspective, players are free to experiment and observe the
effects their actions have on simulated others in order to obtain
and practice certain social skills. However, the range of social
interactions that are typically offered to players is still quite
limited when compared to real human interaction.
With the goal of expanding the range and complexity of
social interactions between characters and humans, there has
been a substantial amount of research dedicated to the creation
and study of virtual agents. These are embodied characters that
are designed to be able to interact with humans in a natural
manner [8]. The architectures that have been developed for
these characters can be rather complex, having to deal with
the challenges of interpreting and synthesizing both verbal and
non-verbal actions as well as modeling cognitive and affective
processes related to decision making.
Although researchers have been able to successfully apply

virtual agent architectures in the development of serious games
(e.g. [1], [9], [11]), such architectures have not yet been widely
adopted by game studios. While the accessibility of these
architectures can be improved through the creation of better
graphical user interfaces and more extensive documentation,
there are also technical and conceptual issues that must be
addressed [18]. A virtual agent architecture relies on a type of
authoring that is oriented towards cognitive concepts such as
goals and beliefs, which are quite familiar for AI researchers
but not necessarily so for game developers. Also, an agent
model will promote a type of storytelling experience that
is distributed or character-centric [2] whereas popular game
developer tools like Articy:draft2or Twine3are designed
towards a plot-centric approach with branching dialogues.
While these tools can be used to create complex narratives
they make a strong distinction between the player and the
other characters, by giving dialogue options to the former but
not the latter. In the proposed toolkit, while certainly possible,
it is not necessary to tie dialogue options to a specific character
or the player.
In this paper, we present a novel toolkit that aims to
promote the adoption by game developers of virtual agent
tools for creating game characters that are more socially and
emotionally intelligent (e.g. are able to adapt to the situation
and to the players). The toolkit is based on the existing
FAtiMA Modular architecture [5], which is an architecture
that was has been successfully used in the past in several
research applications [1], [3], [4]. These improvements were
derived from a close collaboration with game developers at
the company PlayGen4that used the toolkit to develop two
games for learning. The first one is named Space Modules
Inc and is being developed for an educational institute in
the Netherlands named Stichting Praktijkleren5. The game is
designed to teach its players how to provide better customer
service in technical support. The second game is named Sports
Team Manager and is being developed for OKKAM6, a spinoff
company of the University of Trento in Italy. It is a single
player game where players assume the role of a sailing team
manager. Players must hire, fire and communicate with their
team members in order to succeed and, therefore, learn some
personel managemnet skills.
This collaboration is part of the ongoing RAGE project7,
which is an EU-funded project with the goal of developing and
promoting new technologies for directly supporting applied
game developers at creating better applied games and in a
manner that is more cost-effective [19].
II. FATIMA TOOLKIT
FAtiMA Toolkit is an open-source project8that contains a
collection of tools and libraries with the aim of enabling the
2https://www.nevigo.com/en/articydraft
3http://twinery.org
4http://playgen.com/
5https://www.stichtingpraktijkleren.nl
6http://www.okkam.it/
7http://rageproject.eu
8https://github.com/GAIPS-INESC-ID/FAtiMA-Toolkit
creation of interactive storytelling scenarios with non-player
characters that can interact socially with human players in a
variety of contexts.
Storytelling can bring multiple benefits to serious games
[15]. Not only are people more likely to remember what
they learned if the content is integrated in the context of a
narrative, but also, an emotionally engaging story will greatly
motivate players to achieve the intended learning goals of
the game. This form of storytelling centers on the ability
of players to shape how the story unfolds according to their
actions, as participants rather than as observers. This feeling
of agency increases player engagement and encourages them
to reflect more deeply on the consequences of their choices.
However, the more freedom given to players, the more difficult
it becomes to use a traditional scripting approach to author
the scenarios. This is because the branching factor of possible
narrative paths quickly becomes intractable.
Our proposed storytelling framework deals with this issue
by following a character-centered approach rather than a
plot-centered one. The authoring is thus focused around the
different roles that the characters might play in the game and
the narrative emerges from how the characters behave in their
given roles. The challenge then becomes to author these roles
in a way that characters act in a believable manner but also
serve the intended learning goals of the scenario.
As previously mentioned, the toolkit is the result of sev-
eral improvements that were made to the FAtiMA Modular
architecture [5]. For example, the code was ported from the
Java language to C# in order to streamline the integration with
game engines, such as Unity3D. Also, each component within
the toolkit is able to fully load and save its internal state to
aJSON file. As such, it is possible for the game developer
to use his or her text editor of choice to do any kind of
authoring task. However, the toolkit contains some complex
data structures that refer to one another, such as emotions,
an autobiographical memory, appraisal rules, among others.
For this reason, each component has an authoring tool with
a graphical user interface that help users’ in the creation of
content in a declarative way preventing syntactical errors. The
fact that the entire internal state of each component within
the toolkit can be written to a file also works as a logging
mechanism.
Many agent-based tools are designed to function as a
framework or as a stand-alone application that the game must
communicate with, using a specific protocol. In both of these
cases, the game developer has to accommodate the game
to how the agent tool specifies its communication protocol,
its execution cycle and its extensions points, instead of the
other way around. Moreover, given their opinionated nature,
agent-based frameworks are difficult or even impossible to
compose together. It was based on these limitations that we
applied a functional library design pattern in the development
of the toolkit. Consequentially, all the different components
were developed as libraries, i.e. a collection of functions with
well defined inputs and outputs, that the game developer can
directly import and explore more easily without having to

Fig. 1. Diagram of the Role Play Character Component.
worry about future compatibility issues with other tools.
The main functionality of FAtiMA Toolkit is divided in two
main components, the Role-Play Character and the Integrated
Authoring Tool.
A. Role-Play Character
The Role-Play Character (RPC) is the name given to the
component (see Figure 1) within the toolkit that manages
each character’s reasoning and emotional state based on a
perception-action mechanism, which can be described in the
following manner. Firstly, the events that occur in the game
world are sent as input to the Emotional Appraisal component,
which is based on a formalization of the OCC cognitive theory
of emotions [6]. This component then determines if the event
will trigger a new emotion for the character. Each character can
be configured with different appraisal rules that will result in
having different emotional outcomes for the same events. After
the emotional appraisal process is done, any resulting emotion
is added to the Emotional State. Events are also stored in the
character’s Autobiographical Memory along with any emotion
associated to them. The character’s Knowledge Base keeps
track of what the character believes as logical predicates such
as Weather(Outside) = Raining. These beliefs are also updated
according to the events sent by the game world.
After all the internal structures are updated, the RPC uses
the Emotional Decision Making component to select the next
action of the character. This is done using a rule-based mech-
anism that considers both the beliefs of the character as well
as its emotional state. In addition to regular beliefs that are
directly stored in the Knowledge Base, the decision-making
process also takes into account meta-beliefs, which are added
by Reasoning Components such as the Dialogue Manager
or the MCTS. Syntactically, meta-beliefs are expressed in
the same manner as regular ones. The key distinction is
that, rather than being stored, the values of these beliefs
is determined dynamically by the algorithm specified in the
reasoning component. This allows the combination of multiple
decision-making strategies into a unified rule-based system.
Developers can also register their own modules as additional
reasoning components and the meta-beliefs they introduce
will become available in the conditional rules of all other
components. For instance, consider a game with a specific
scoring mechanism for the player and the developer wants
to create a decision rule for NPCs to congratulate the player
whenever the player’s score reaches a certain threshold. This
could be achieved by registering the scoring mechanism as a
new Reasoning Component that would add Score(Player) =
[x] as a new meta-belief.
Game characters should have believable emotional re-
sponses to give the illusion of life. For applied games that rely
heavily on social interaction, it quickly becomes impractical to
manually script all the emotional reactions of each character
for each possible event. The RPC asset tackles this issue by
allowing game developers to create general profiles of how
characters respond emotionally in their games. They can test
and configure these profiles outside of the game and they can
naturally switch between profiles without having to recompile
the game source code.
B. Integrated Authoring Tool
The Integrated Authoring Tool is the other main component
of the toolkit that is designed to be the central hub for
game developers when creating a new storytelling scenario
or adapting existing ones. It allows the configuration of the
general aspects of the scenario and provides quick access to
the authoring tools of the Role-Play Character component.
However, the main feature of this component is that it contains
a dialogue editor that allows the developer to specify the
dialogue acts that are available for both the player and the
characters.
For the purpose of dialogue management, the author must
define the interaction state where each dialogue may occur as
well as define the next state if a certain dialogue is selected.
During runtime, all characters are informed about the existing
dialogue acts as well as dialogue states. Characters are then
able to use this information to decide what to say according
to their internal state and decision-making mechanisms. To
give an example, consider that the integrated authoring tool
informs a character that at the start of the interaction there
are two valid dialogues, one to greet the player respectfully,
another to greet the player in an angry manner. If the character
is angry, the emotional decision making asset will select the
second option. If not, then the first greeting will be selected
instead.
III. CAS E STU DY 1-SPACE MO DU LES INC
Space Modules Inc is a single player game where the player
takes on the role of a customer service representative for a
spaceship part manufacturer “Space Modules Inc”. The virtual
characters in the game play the role of customers that call
the player (see Figure 2) about hardware and software faults
they are experiencing. Some characters will be angry, others
uncooperative or stressed, and it’s up to the player to manage
the situation and decide how best to respond.
Players have to respond to situations by engaging in con-
versation with customers. This is done by having the player
pick one of the available dialogue options in response to the
character’s chosen dialogue. The process is repeated until the

Fig. 2. Space Modules Inc. Game Flow.
Fig. 3. Space Modules Inc - Dialogue Screen (left image) and Result Screen
(right image) Flow.
final state of the conversation is reached and then the player’s
score is passed to the review screen to be shown to the player
(see Figure 3). The customer satisfaction score depends on
how the player affected the emotional state of the character.
The idea is that each customer can have a different emotional
profile, thus providing a different challenge to the player.
From a pedagogical perspective, players must learn how to
manage intense emotions and how to respond to customers
in a professional manner in the best way. In other words,
the pedagogical goal of the game is to train players in being
able to identify a person’s emotional state through verbal and
nonverbal feedback and gain further experience in providing
effective emotional responses.
The emotional reactions of the customers in Space Mod-
ules are determined by the Role-Play Character component.
According to the selected emotional profile, this component
initializes the overall mood of the character to a given value
between -10 to 10. The component then updates this value
based on how it evaluates the option selected by the player. If
the player decides, for instance, to give the wrong solution for
the problem that the customer has, the component will gener-
ate a “Distress” emotion and the overall mood decreases. The
player can then repair the mood of the character by selecting
a dialogue that shows empathy for the character’s distress.
Fig. 4. Sports Team Manager Game Flow.
However, if this dialogue is selected when the character is not
feeling distressed, then it will be judged negatively instead and
the mood of the character decreases accordingly. The amount
by which the mood decreases or increases is also another
parameter that is possible to configure in the RPC component.
IV. CAS E STU DY 2-SPO RTS TE AM MA NAG ER
Sports Team Manager is an applied game also developed
by PlayGen with the assistance of the FAtiMA Toolkit. The
overall goal of the game is to have the player be able to
assemble together the most optimally performing sailing team
by resolving conflicts and managing the team’s interactions.
The player interviews virtual characters to identify their skills
and personalities. The team has a set of roles, each with
overlapping skill requirements. A successful sailing team is
not solely based on skill, but also on the social relationships
between team members. Players must communicate with their
team, deciding which members are placed into each position
per race and resolve conflict situations as they arise. Figure 4
shows the game flow during an individual race session.
The players must first review the positions they need to
fill on the boat, taking note of the required skills for each.
Next, they must meet with their NPC team members, taking
into account the skills and inter-team relationships already
known, asking questions where further information is needed.
Using this information they should, if required, recruit new
members into the team and place individuals into positions.
After racing with the selected line-up, players will occasionally
have to handle events with team members. After the event
stage concludes, using the result and pieces of feedback from
the race session, players begin the gameplay loop again, but
now with additional information to assist in their decision
making.
The Role-Play Character component is used here to model
the emotional state and decision making of each team member
based on their belief set. The component analyses the actions
of the player and determines their effect on the emotional state
of each NPC based on their current state and the emotional
weighting of the event in their perspective. To give an example,

Fig. 5. Sports Team Manager - Post-Race Event.
after each race session, it is possible for a team member to
come to the player in order to talk to them. The character
might for instance, ask why she was not picked (see Figure 5).
Players can then reply back to the team member by selecting
from a list of dialogue options. If the player selects an overly
aggressive reply, the character is likely to feel angry, affecting
its next response.
As mentioned previously, the Role-Play Character compo-
nent stores the beliefs of every NPC and saves these beliefs
over multiple play sessions. These beliefs are related to
information such as their last position in the team, skill ratings,
opinion ratings and event states. Furthermore, the events sent
to the characters are saved, meaning a history of events can
be preserved. This allows a history of every team selection to
be stored. As all of this information is stored regularly, it can
be also be reloaded in further play sessions, allowing for the
possibility of a persistent game.
Concerning the Integrated Authoring Tool, this component
is used to manage the configuration of the scenario, which
contains a list of all possible role-play characters that are
dynamically created at the beginning of and during each game.
The component also contains all of the dialogue options for
the player and the NPCs during various parts of the game,
such as team member meetings and post-race events.
V. G AME DEVELOPERS FEE DBACK
Game developers from PlayGen were independent in the
integration of the FAtiMA toolkit in their game code and were
successfully able to use the toolkit to support the intended
gameplay in the two games. They relied on the documentation
and examples created for the community and had full access to
the toolkit source code. We conducted an informal interview
to get their impression regarding the technical integration
and the usefulness of the toolkit. Contacts were made by
email and face to face. The conversation was around three
main questions: (1) How was the FAtiMA toolkit used in the
development of the game?, (2) What were the main benefits
of using the FAtiMA toolkit? and (3) What were the main
difficulties of using the FAtiMA toolkit?
Game developers reported that “the integration was not
difficult, but that a proper use of the toolkit requires a steep
initial learning curve”. The toolkit facilitated the creation of
mechanisms “to determine the change in emotional state and
mood depending on the dialogue chosen by the player” and
was also useful “to calculate the NPC response to the provided
piece of player dialogue, depending on their emotional state
and the type of player dialogue selected.” and to “decide how
a NPC should greet the player depending on their current
relationship with the player.”. They highlighted two main
benefits regarding the pedagogical value that the FAtiMA
toolkit provided. First, the use of the toolkit was “good because
players get immediate implicit (contextual) feedback”. They
mean that the emotional responses of the characters were
potentially very good cues for the players to assess if they
were playing well without the need to show explicit numeric
score. The second benefit, was the “ability to dictate the course
of conversation indirectly through using the toolkit’s dialogue
and NPC emotions systems, as these have made setting up
and controlling scenarios a much easier process as a result.”.
What is relevant, in the pedagogical sense, is the fact that the
definition and setting up of the scenarios was made directly
by the trainers who will apply the games. Hence, the game
can be configured and adapted by the people who have the
most knowledge about the content to be delivered in order to
achieve the learning goals of the game.
VI. STUDENTS GAME AI PRO JE CT S
The toolkit was also put to test in a course on Game AI at
IST, University of Lisbon in the fall semester. It was used in
the final project of the course (out of 4) that constituted 30%
of the grade. Sixty-eight students, working in groups of three,
were engaged. They had a workshop on the FAtiMA Toolkit
(of about 2 hours) before tackling the problem. They used a
version of the toolkit that is integrated with the Unity game
engine and uses components to realise the body and expression
of the characters developed by other members of the RAGE
project.
Each group was given the task of using the FAtiMA
toolkit to create two conversational scenarios, one with a
single character interacting with the player and another with
two characters engaging in conversation with the player at
the same time. Students were free to select any theme for
the conversation as long as the non-player characters had
believable emotional responses and could be configured to
have different personalities. All groups managed to finished
the project. Some of the scenarios created had quite interesting
and surprising themes. For instance, one group chose to create
a scenario where players were at the gates of heaven and had
to convince the gatekeeper to let them in. To be successful,
players had to avoid upsetting the gatekeeper too much. Other
groups opted for a more serious theme such as a job interview
(see Figure 6) or a shopping scene with a father, his son, and a
shopkeeper. With the student’s permission, these scenarios will
be publicly available as examples that are part of the toolkit.
From a software quality perspective, given the wide range of
scenarios explored by the students, we were able to identify
some issues with the toolkit, which were promptly fixed.

Fig. 6. Students’ Job Interview Demo.
VII. CONCLUSION
In this paper, we argued that the development of serious
games is faced with additional challenges that are related to
the pedagogical goals that the designers have in mind. For
instance, in games that are about teaching conversational skills,
developers have to figure out how to offer a rich interaction
space that supports the exploration and failure of different
communicative actions and their associated socio-emotional
effects.
In the mainstream gaming industry, dialogues are typically
handled through branching structures that limit the set of
possible interactions, by offering little flexibility in the way
characters respond to what the players say to them. Alterna-
tively, in the research field of virtual agents, researchers have
developed and proposed tools for the creation of conversational
agents that have rich socio-emotional models driving their
behavior. These agents have great potential for being applied
in serious games that teach soft skills, as their behaviors are
more procedural and less scripted. However, so far, agent
architectures are still far from being widely used in the serious
games industry due to, in large part, accessibility issues.
With those issues in mind, we took an existing virtual agent
architecture, FAtiMA Modular, and adapted it to a new toolkit
with the goal of making it more appealing to game developers.
For that effect, we adopted a functional library pattern instead
of a framework-based approach. Moreover, the functionality
was divided in two main components, the Role-Play Character
and the Integrated Authoring Tool. The first is responsible
for managing the character’s beliefs, memories and emotional
state as well as running a decision-making process for each
character according to its ascribed role. The second component
allows the developer to manage the list of all the characters
that are available in each game scenario as well as the available
dialogues that the characters, including the player’s avatar, can
select from at any given state of the interaction.
The resulting toolkit was then applied successfully by a
game studio, PlayGen, in the development of two serious
games. The first game was designed to teach players how to
properly communicate with emotional customers in a customer
service setting. The second game has the player managing a
sport sails team composed by multiple characters with different
role preferences. Both of these games benefited from the use
of the toolkit in adding emotional dynamics to their characters
that is reflected in their decisions. Additionally, a group of 68
students successfully developed projects for a Game AI course
using the toolkit. This experience was a good stress test on
the toolkit given the wide variety of scenarios explored by the
students.
As future work, we plan to conduct more formal user study
centered around the authoring capabilities of the toolkit. The
main idea will be to have participants watch a video tutorial
about how the toolkit works and then be instructed to change
an existing game scenario according to a set of predefined
goals. The feedback obtained will then be used to further
improve the toolkit.
ACKNOWLEDGMENTS
This work was supported by national funds through
Fundac¸˜
ao para a Ciˆ
encia e a Tecnologia (FCT) with reference
UID/CEC/50021/2013 and has been partially funded by the
EC H2020 project RAGE (Realising an Applied Gaming Eco-
System) Grant agreement No 644187.
REFERENCES
[1] R. Aylett, M. Vala, P. Sequeira, and A. Paiva, “Fearnot!–an emergent
narrative approach to virtual dramas for anti-bullying education,” in
International Conference on Virtual Storytelling. Springer, 2007, pp.
202–205.
[2] M. Cavazza, F. Charles, and S. J. Mead, “Character-based interactive
storytelling,” IEEE Intelligent systems, vol. 17, no. 4, pp. 17–24, 2002.
[3] F. Correia, P. Alves-Oliveira, N. Maia, T. Ribeiro, S. Petisca, F. S. Melo,
and A. Paiva, “Just follow the suit! trust in human-robot interactions
during card game playing,” in Robot and Human Interactive Communi-
cation (RO-MAN), 2016 25th IEEE International Symposium on. IEEE,
2016, pp. 507–512.
[4] N. Degens and G. Hofstede, “Traveller - Intercultural training with
intelligent agents for young adults,” Proceedings of the . . . , 2013.
[5] J. Dias, S. Mascarenhas, and A. Paiva, “Fatima modular: Towards
an agent architecture with a generic appraisal framework.” Emotion
Modeling, vol. 8750, pp. 44–56, 2014.
[6] J. Dias and A. Paiva, “Feeling and reasoning: A computational model
for emotional characters,” in EPIA, vol. 3808. Springer, 2005, pp.
127–140.
[7] T. Economist, “Why video games are so expen-
sive to develop,” The Economist Group Limited, 2014.
[Online]. Available: http://www.economist.com/blogs/economist-
explains/2014/09/economist-explains-15
[8] J. Gratch, J. Rickel, E. Andr´
e, J. Cassell, E. Petajan, and N. Badler,
“Creating interactive virtual humans: Some assembly required,” IEEE
Intelligent systems, vol. 17, no. 4, pp. 54–63, 2002.
[9] W. L. Johnson and A. Valente, “Tactical language and culture training
systems: Using artificial intelligence to teach foreign languages and
cultures.” in AAAI, 2008, pp. 1632–1639.
[10] F. Khatib, S. Cooper, M. D. Tyka, K. Xu, I. Makedon, Z. Popovic,
D. Baker, and F. Players, “From the Cover: Algorithm discovery by
protein folding game players,” Proceedings of the National Academy of
Sciences, vol. 108, no. 47, pp. 18 949–18 953, 2011.
[11] J. M. Kim, R. W. Hill Jr, P. J. Durlach, H. C. Lane, E. Forbell, M. Core,
S. Marsella, D. Pynadath, and J. Hart, “Bilat: A game-based environment
for practicing negotiation in a cultural context,” International Journal of
Artificial Intelligence in Education, vol. 19, no. 3, pp. 289–308, 2009.
[12] G. Koo and S. Seider, “Video Games for Prosocial Learning,”
Ethics and Game Design, pp. 16–33, 2010. [Online]. Available:
http://128.197.153.21/seider/Consolidated papers/Prosocial Learning
in Video Games Final Version Ko.pdf%5Cnhttp://services.igi-
global.com/resolvedoi/resolve.aspx?doi=10.4018/978-1-61520-845-
6.ch002

[13] A. Martens, H. Diener, and S. Malo, “Game-based learning with
computers–learning, simulations, and games,” Transactions on edutain-
ment I, pp. 172–190, 2008.
[14] M. Ninaus, K. Kiili, J. McMullen, and K. Moeller, “A Game-
Based Approach to Examining Students’ Conceptual Knowledge
of Fractions,” in Games and Learning Alliance: 5th International
Conference, GALA 2016, Utrecht, The Netherlands, December 5–7,
2016, Proceedings, R. Bottino, J. Jeuring, and R. C. Veltkamp, Eds.
Cham: Springer International Publishing, 2016, pp. 37–49. [Online].
Available: http://dx.doi.org/10.1007/978-3-319-50182-6 4
[15] N. Padilla-Zea, F. L. Gutirrez, J. R. Lpez-Arcos, A. Abad-
Arranz, and P. Paderewski, “Modeling storytelling to be
used in educational video games,” Computers in Human
Behavior, vol. 31, pp. 461 – 474, 2014. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0747563213001295
[16] P. Sajjadi, E. El Sayed, and O. De Troyer, “On the Impact of
the Dominant Intelligences of Players on Learning Outcome and
Game Experience in Educational Games: The TrueBiters Case,” in
Games and Learning Alliance: 5th International Conference, GALA
2016, Utrecht, The Netherlands, December 5–7, 2016, Proceedings,
R. Bottino, J. Jeuring, and R. C. Veltkamp, Eds. Cham: Springer
International Publishing, 2016, pp. 221–231. [Online]. Available:
http://dx.doi.org/10.1007/978-3-319-50182-6 20
[17] A. Santos, F. Strada, and A. Bottino, “Games and Learning Alliance,”
in Games and Learning Alliance: 5th International Conference, GALA
2016, Utrecht, The Netherlands, December 5–7, 2016, Proceedings,
ser. Lecture Notes in Computer Science, A. De Gloria, Ed. Cham:
Springer International Publishing, 2015, vol. 9221, no. April 2016, pp.
73–82. [Online]. Available: http://link.springer.com/10.1007/978-3-319-
22960-7
[18] U. Spierling and N. Szilas, “Authoring issues beyond tools,” in Joint
International Conference on Interactive Digital Storytelling. Springer,
2009, pp. 50–61.
[19] W. Van Der Vegt, W. Westera, E. Nyamsuren, A. Georgiev, and I. M.
Ortiz, “RAGE Architecture for Reusable Serious Gaming Technology
Components,” International Journal of Computer Games Technology,
vol. 2016, 2016.
[20] N. Vannini, S. Watson, K. Dautenhahn, S. Enz, M. Sapouna, D. Wolke,
S. Woods, L. Hall, A. Paiva, E. Andr´
e, R. Aylett, and W. Schneider,
“”FearNot!”: A computer-based anti-bullying-programme designed to
foster peer intervention,” European Journal of Psychology of Education,
vol. 26, no. 1, pp. 21–44, 2011.
[21] L. von Ahn and L. Dabbish, “Labeling images with a computer
game,” Proceedings of the 2004 conference on Human factors in
computing systems - CHI ’04, pp. 319–326, 2004. [Online]. Available:
http://portal.acm.org/citation.cfm?doid=985692.985733