Knowledge-Level Creativity in Game Design
Adam M. Smith and Michael Mateas
Expressive Intelligence Studio
University of California, Santa Cruz
{amsmith,michaelm}@soe.ucsc.edu
Abstract
Drawing on inspirations outside of traditional computational
creativity domains, we describe a theoretical explanation
of creativity in game design as a knowledge
seeking process. This process, based on the practices of
human game designers and an extended analogy with
creativity in science, is amenable to computational realization
in the form of a discovery system. Further, the
model of creativity it entails, creativity as the rational
pursuit of curiosity, suggests a new perspective on existing
artifact generation challenges and prompts a new
mode of evaluation for creative agents (both human and
machine).
Introduction
Paintings (Colton 2008), melodies (Cope 2005), and poems
(Hartman 1996) are familiar domains for artifact generation
in computational creativity (CC), and much established
theory in the field is focused on evaluating such
artifacts and the systems that produce them. In this paper,
we draw inspiration for a new understanding of creativity
from the less familiar (but no less creative) domain of
game design. In its full generality, game design overlaps
visual art, music, and other areas where there are many
existing results, but where it stands apart is in its unavoidably
deep, active interaction with the audience: in gameplay.

 Crafting gameplay is the central focus of game design
(Fullerton 2008). Play, however, is not an artifact to be
generated directly. Instead, it is a result that emerges from
the design of the formal rule system at the core of every
game (Salen and Zimmerman 2004, chapter 12), a machine
driven by external player actions.
 Where, in visual art, we might judge the creativity (as
novelty and value) of an artifact on the basis of the work’s
similarity to known pieces and its affective qualities
(Pease, Winterstein and Colton 2001), it is not so easy to
make direct statements about the properties of the artifacts
in game design. Desirable games are celebrated for their
innovative gameplay or the fun experiences they enable–
these are properties of the artifact’s interaction with the
audience, not of the artifact itself. The focus on predominantly
passive artifacts in CC, those which can be appreciated
via direct inspection rather than through interactive
execution, has masked what is obvious in game design:
that the desirability of artifacts is in their relationship to
their environment.
Armed with such an understanding, we seek a theoretical
explanation of creativity in game design–not the engineering
application of established design knowledge, but
the rarer experimentation that realizes new forms of gameplay
and original player experiences. This theory should
speak to both the artifacts and processes of game design,
and do so in a way that meaningfully explains game design
as done by humans as well as computational means. Towards
capturing the richness of existing human design activity,
we are most interested in a theory of transformational
creativity (Boden 2004) that explains how designers
build new conceptual spaces of game designs and reshape
them in response to feedback experiences observing play.
We introduce a new theoretical model that is amenable
to computational realization which describes creative game
design as a knowledge-seeking process (a kind of active
learning). Our broader contribution, creativity as the rational
pursuit of curiosity, can provide an explanation of
and suggest new questions for applications in traditional
CC artifact generation domains.
In the following sections we will review established
game design practices, draw an analogy between game
design and scientific discovery, review and apply Newell’s
concept of the knowledge level, and then introduce our
model of creativity. Finally we will conclude with a discussion
of the implications of this theory for game design
and the larger CC context.
Game Design Practices
In a standard text, Salen and Zimmerman (2004, p. 168)
introduce the “second-order” problem of game design
bluntly:
“The goal of game design is meaningful play, but play is
something that emerges from the functioning of the
rules. As a game designer, you can never directly design
play. You can only design the rules that give rise to it.
Game designers create experience, but only indirectly.”
Play includes the objective choices made by a player and
the conditions achieved in the game, along with the playProceedings
of the Second International Conference on Computational Creativity 16
er’s subjective reactions and expectations. At this point, it
is straight forward to adopt the first tenet of our theory of
creative game design: game designers are really designers
of play.
The idea of adopting an iterative, “playcentric” (Fullerton
2008) design process, in which games are continually
tested to better understand their emergent (play) properties,
is corroborated by others like Schell (2008), who further
describes the supreme importance of “listening” in the design
process (being able to process feedback from the
player’s experience of candidate designs). Beyond initial
conceptualization of a game idea and tuning and polish of
the final product, the two most important practices of game
design are prototyping and playtesting, both of which are
intentionally focused on providing the designer with a better
understanding of play.
 Prototypes are playable artifacts, working models of a
game idea that permit asking and answering questions
about how a game will interact with its environment without
requiring the effort to create a complete, polished
game. The aesthetics of prototypes are very different than
for complete games: most artwork and sound is stripped
away from a design idea to produce an artifact that most
effectively elicits feedback on a designer’s current focus of
interest (often how the interaction of game mechanics affects
the trajectory of play).
 Prototypes must be set to interact with an audience to
gain the answers they are designed to provide. Playtesting
is the practice of playing a game or gameplay prototype
while observing the choices made, actions taken, or reactions
expressed by sample players (the designer, a friend, a
dedicated tester, or even a member of the target audience).
Observations made during playtesting can reveal objective
properties of a game such as unwritten-but-implied mechanics,
exploits, and alternative puzzle solutions, or subjective
properties such as the level of engagement, fun, or
hesitation expressed by the sample players (Smith, Nelson,
and Mateas 2009).
 Despite the ostensible purpose of game design being the
production of complete, desirable games for play by endusers,
the practices of playtesting and prototyping are centered
on providing feedback to the designer. Through the
rough generate-and-test process of iterative game design,
where several prototypes are created and playtested during
a single project, the underlying goal is to build up sufficient
skill and understanding to later produce the highquality,
final game artifact. Such a self-affecting process is
exactly what McGraw and Hofstadter call the “central loop
of creativity” (1993).
 Beneath the surface, the practices of game design are
almost exclusively about the collection of design knowledge,
knowledge regarding the relationship between the
component elements of a game system and that game’s
potential execution in interaction with a player. Such design
knowledge spans what design patterns to employ, how
to assemble them, and why such an assembly will produce
a certain play experience.
 Existing design knowledge can be applied to realize familiar,
well-understood play experiences, but creative
game design demands a continuous source of new design
knowledge. Thus, the second tenet of our theory is this:
creative game design is about seeking design knowledge.
An Analogy with Science
To expand on knowledge-seeking in game design, we want
to draw an extended analogy between game design and
science. Doing so will allow us to connect the creative activity
in the game design process with the activity carried
out by scientific discovery systems in CC.
 For design generally, Dasgupta claims “design problem
solving is a special instance of (and is indistinguishable
from) the process of scientific discovery” (1991, p. 353).
While Dasgupta focuses on explaining design activity specifically
in terms of finding a confirmable theory which
resolves a particular unexplained phenomena, our analogy
is intentionally softer, to enable applications in a variety of
CC domains that no not immediately appear as “design
problem solving” domains.
Scientific Practices
Roughly, the scientific method is a closed loop with the
following phases: A hypothesis is generated from a working
theory, the hypothesis drives the design of an experiment
(usually realized with a physical apparatus), data
from executing this experiment is collected, and conclusions
are drawn which can be integrated into the working
theory. A scientist will design an experimental setup, despite
already possessing a theory which makes prediction
about the situation, precisely because there is some uncertainty
about the result. This result, whether matching the
prediction or not, should provide informative detail about
the natural laws at play in the experiment’s environment.
 In our analogy, experiment design is prototyping; experiment
execution and subsequent analysis is playtesting.
The combination of the declarative knowledge of natural
laws and the procedural knowledge of operating laboratory
equipment is game design knowledge. Finally, the closed
loop of the overall scientific method corresponds to iterative
game design.
 Making the parallel clearer, Gingold and Hecker (2006)
talk about how gameplay prototypes should be informative,
answer specific questions, and be falsifiable. In the philosophy
of science, the notion of informative content (and its
relation to falsifiability) guides the evaluation of theories
and experimental designs.
 In their capacity to design and execute informative experiments
and produce coherent and illuminating explanations
of the anomalous results, scientists are clearly creative.
In Colton’s terms (2008), we can easily perceive this
creativity in the skill of precise experimental design, the
appreciation of unexpected result in the context of a working
theory, and the imagination of previously difficult to
consider alternative theories and the invention of new instruments.
By the analogy above, a scientist’s kind of creativity
can apply to the game designer as well, prompting
Proceedings of the Second International Conference on Computational Creativity 17
our third tenet: designers act as explorers in a science of
play, an “artificial science” in Simon’s terms (1996, Ch. 5).
Automated Discovery
Though largely distinct from artifact generation, automating
discovery in science and mathematics is an established
CC tradition (Langley et al. 1984; Lenat 1976). Within
these systems it is common to find subprocesses which
generate artifacts as part of the larger discovery process.
 The GT system (Epstein 1988), an automated graph
theorist, would periodically “doodle” random graphs within
a specific design space as a means of generating relevant
data which might spark a new conjecture about the desired
area of focus. With our analogy in mind, such doodling is
reminiscent of the exploratory, rapid prototyping process
sometimes used in game design (in what Gingold and
Hecker call the “discovery” phase of development). A heuristic
to generate new concrete examples of abstract concepts
was even present in the original automated mathematician,
AM, working with number theory (Lenat 1976).
 Beyond generating artifacts with only the indirect intention
of knowledge gain, more recent discovery systems
internally optimize expected knowledge gain when deciding
which experimental setup to test next, realizing an active
learning process (Bryant et al. 2001). Where artifact
generation provides opportunitistic benefits in mathematical
domains (in which graphs and conjectures are noninteractive,
static artifacts), discovery systems working in
the physical sciences fundamentally cannot avoid artifact
generation (as experimental design) during active exploration
of their domain.
A notion of “interestingness” is the glue that binds the
various subprocesses of automated discovery (artifact generation
included) together into an overall control flow
(Colton and Bundy 2000). In many cases, interestingness
measures the likelihood or quality of knowledge expected
to be discovered by taking a particular action (e.g. searching
for a counterexample) or focusing on a particular concept
(e.g. looking for new examples of special graphs). A
system’s overall notion of interestingness can be used to
induce a measure of value for artifacts generated in its artifact
generation subprocesses, a measure related to potential
for knowledge gain as opposed to aesthetic value.
 Returning to game design, our fourth tenet holds that
automated discovery systems inspire a computational
model of creative game design that explains the prototypes
produced in exploratory game design as the doodles produced
trying to flesh out design theories residing in the
designer’s head, motivated by an interest in designs that
have the potential to reveal new patterns.
Newell’s Knowledge Level
To complete the image of the creative game designer as a
discoverer, we need some better vocabulary for talking
about a designer’s knowledge, around which the entire
discovery process revolves.
 Newell (1982) describes the “knowledge level” as a systems
level set above the symbolic, program level. At the
knowledge level, we find agents with bodies of knowledge
that can take actions in some environment to make
progress towards goals. The actions taken by an agent are
said to be governed by a principle of rationality that states
“If an agent has knowledge that one of its actions will lead
to one of its goals, then the agent will select that action.”
This sense of rationality is distinct from decision theoretic
rationality in that it does not necessarily imply that an
agent must optimize anything.
 While this radically underspecifies an agent’s behavior
from a computational perspective, the constellation of concepts
at the knowledge level is useful for making statements
about our game designer. The intention of knowledge-level
modeling is to explain the behavior of knowledge-bearing
agents (be they human or machine) without
reference to how that knowledge is represented or access to
an operational model of the agent’s mode of processing.
 Understanding the game designer, at the knowledge level,
starts with making an assumption about what is known
and what is sought. But from our theory so far, we can
safely assume an important body of knowledge possessed
is that of tentative game design knowledge. This knowledge
permits the designer the use of tools such as paper
and trinkets for physical gameplay prototypes (often styled
after board games) and programming languages and compilers
for more detailed, computational prototypes. This
same knowledge permits understanding to be gained from
the observation of game artifacts in play, and suggests a
tentative vocabulary for composition of those artifacts (i.e.
knowledge of design patterns for game rules). The creative
designer’s goal, per our analogy with science, is clearly to
gain more design knowledge.
 Given this, we expect the designer to rationally (specfically
in the knowledge level sense) go about the practices
of game design as part of taking actions that lead towards
the gain of design knowledge. That is, game design activity
can be explained as the rational pursuit of design
knowledge gain.
Creativity as Rational Curiosity
The knowledge level lets us talk about a kind of rationality,
one that gives an explanation for why game designers take
the actions they do. But not all game design activity is creative,
no more than all of science being creative. So where
does creativity come in?
 The most creative parts of game design, we claim, are
the ones where the designer’s behavior is best explained by
the direct intention to gain new knowledge, to satisfy curiosity.
The bulk efforts of game production are a kind of
engineering which applies the knowledge gained in the
curiosity-driven creative mode.
As the motivation to reduce uncertainty and explore
novel stimuli (Berlyne 1960), curiosity has long been
known to be intertwined with the judgment of aesthetics
(Berlyne 1971). Saunders’ “curious design agents” (2002)
generate aesthetic artifacts according to their potential to
satisfy an internal measure of curiosity, doing so in order
to learn about an outside environment. This framework has
Proceedings of the Second International Conference on Computational Creativity 18
also been used to drive the behavior of a simulated society
of curious visual artists and even a flock of curious sheep
in a virtual world (Merrick and Maher 2009).
How curiosity-driven behavior can explain the various
processes and artifacts of human creativity at a high level
has been demonstrated at great length (Schmidhuber 2010).
Our unique claim, that creativity is a knowledge level phenomenon,
gains similar explanatory power without reference
to algorithmic details (such as the use of reinforcement
learning or optimization procedures) or human psychology,
as in Loewenstein’s comprehensive review of
research on human curiosity (1994).
 Looking concretely at curiosity in the domain of game
design, consider the example of speed runs, gameplay traces
that demonstrate a way of doing something in a game
(completing a level or collecting certain items) much faster
than a designer previously expected. Speed runs are interesting
to game designers, from a curiosity perspective, because
they often represent a novel stimulus and quickly
increase uncertainty about what is possible in a game,
creating an urge to seek out related gameplay traces that
would illustrate the general pattern by which the run was
achieved. With additional experience, the designer can
learn to either design-out such speed runs by adjusting the
game’s rules, or create new mechanics that reward them.
 Putting together curiosity about design knowledge with
knowledge-level rationality, we have our complete theory
of creativity in game design:
Creativity is the rational pursuit of curiosity.
 This claim applies to human and machine design agents
and gives a goal-oriented explanation to sequences of design
activity that result in design knowledge gain (clearly
including prototyping and playtesting). A creative game
designer makes games, not because that is their function,
but because they want to learn things about play that require
experimentation with certain artifacts to illuminate.
 We call this theory rational curiosity because it is a
knowledge-level treatment of the concept of curiosity that
focuses on how curiosity explains the selection of actions
towards a known goal. It claims that curiosity, applied rationally,
will result in behavior recognizable as creative
design activities.
Transformational Creativity in Game Design
Consider our model creative game designer, over time,
producing increasingly complex and refined playable artifacts
that are in line with the complexity of their currently
operating design theory. That playable artifacts are produced
is just an externally visible byproduct of the more
interesting process going on in the designer’s thoughts, the
growth and refinement of design knowledge.
 Taking a snapshot at any one time, the designer’s knowledge
is fixed. The present knowledge describes a “conceptual
space”, in Boden’s terms, of game designs and play
possibilities. Combinational creativity within this pace
would entail the generation of artifacts from known structures
and construction constraints or, perhaps, the enumeration
of explanations of a player’s behavior with respect
to known patterns. Taking a series of steps in terms of the
current design theory, producing a new game using a design
pattern of interest, producing a prediction of player
behavior, and then performing a playtest and comparing
the results with the prediction is an example of exploratory
creativity in this space. These activities are weakly creative
in the rational curiosity view because, though they might
indeed be motivated by potential knowledge gain, neither
realizes an actual change to the designer’s personal theory.
Transformational creativity in game design, then, is design
activity which results in a redefinition of design theories.
In iterative game design, where many prototypes are
produced in succession in response to feedback from playtesting,
is an intensely transformative process. Such transformations
can include the definition of a new design pattern
which simplifies the explanation of how another designer’s
game was constructed, a constraint which limits
the use of two patterns together, or a rule which predicts a
certain kind of player’s behavior when a certain combination
of pattern is present.
Discussion
Having proposed a theoretical explanation of creativity in
game design, let’s look at what it entails.
Computational Creativity in Game Design
The theory of rational curiosity in game design can be realized
computationally along two major paths: the development
of a game design discovery system, and new, knowledge-oriented
creativity support tools. More generally,
however, it suggests new elements that need to be modeled
computationally in support of either path.
Game Design Discovery Systems Recalling the analogy
with scientific and mathematical discovery systems, we
can imagine the design of a new kind of discovery system
that would work in the domain of game design knowledge.
This discovery system would produce games as part of its
experiments in exploring play, but it would also allocate
significant attention to decomposing games made by other
designers and producing explanations of observed human
player actions.
 The notion of interestingness in this discovery system
would correspond to a symbol-level realization of the
agent’s knowledge-level goal: the satisfaction of curiosity
about design knowledge. By selecting actions (such as the
construction of a prototype, the simulation of a playtest
using a known player model, or the analysis of a previously
produced game in light of a refined theory) according to
their calculated prospects for improvement of the working
design theory (a library of design patterns, predictive rules
for player behavior, and constraints on play-model construction),
the system’s behavior would implement a rational
pursuit of curiosity, creativity in game design.
 Constructing such a system would require new research
into adapting symbol-level representations of design knowledge
for use in game design, the development of a task
decomposition of creative game design into subgoals and
Proceedings of the Second International Conference on Computational Creativity 19
actions (such a concrete design methodology which would
be of interest to human designers as well), and the identification
of the relevant external tools of game design (certainly
paper prototyping materials and programming environments
are some of these tools, but where are the CAD
systems for games?).
Knowledge-oriented Creativity Support Tools
With recognition that design knowledge gain is the designer’s
goal, creativity support tools in game design should
focus on easing this process. In terms of Yeap’s desiderata
for creativity support tools, these tools should focus on
ideation and empowerment (2010). In game design, these
translate to the generation of candidate design knowledge
for the designer to consider and then leaving the adoption
of the new knowledge up to the designer (without undue
interference). Knowledge-oriented creativity support tools
should attempt to remove bottlenecks in the discovery
process.
 We created a gameplay pattern language and corresponding
search tool which is intended to accelerate the
extraction of feedback about game designs from prerecorded
traces of play (Smith and Mateas 2011). The system
is also capable of compiling patterns into a lower-level
form that can be used to search for additional evidence of
gameplay patterns with the machine playtesting tools included
in the BIPED early-stage computational prototyping
tool (Smith, Nelson and Mateas 2009). Neither of these
systems are themselves creative, but they are designed to
provide new actions to the creative designer for rational
selection in the service of knowledge gain.
 In another project (Smith and Mateas 2010), we captured
a design space of mini-games in a logic program and
used model-finding techniques to automatically generate
artifacts from the described space. Use of the logic programming
tools automates a small slice of game design
(the literal construction of artifacts) and provides a convenient
symbolic representation for some types of design
knowledge. While this project automates combinational
creativity in game design, the design space representation
and sampling tools are intended to support an external designer’s
transformational creativity in the realization of
new forms of gameplay through rapid exploration and expressive
redefinition of the mini-game design space.
New Perspective for Computational Creativity
We have mostly focused on game design, but rational curiosity
is intentionally worded so as to apply to other CC
domains. In fact, it should apply even to domains with apparently
non-interactive artifacts. Where, in game design,
we were concerned with the implications of game rule system
on player actions and reactions, in the domain of music
we should explore the implications for sound patterns
on audience anticipation and mood, in visual art the implications
for perceptual details on where the viewer’s eye
lingers or flees, and in sculpture the implications of geometric
arrangements on audience interest from particular
viewpoints. Such domains are not as interactive as game
design, but they could be equally deep in the subtlety of
how an audience reacts to an artifact – depth enough to
keep the rationally curious artist busy producing experiments
for quite some time.
 The knowledge-level analysis of creativity suggests new
questions to ask of CC systems: What does this system
want to learn? How is knowledge represented in this domain?
Is the system experimenting with the affordances of
the raw medium or focusing on audience reactions achievable
through it? (Both are equally creative assuming the
desired kind of knowledge is gained.)
 Consider NEvAr (Machado and Cardoso 2002), a creative
system in the domain of visual art. Unlike the
straightforward interactive genetic algorithm in PicBreeder
(Secretan et al. 2008), NEvAr does not ask its audience for
feedback on every artifact it internally considers. The system
summarizes sparse feedback from its human audience
in the form of a neural network which becomes a proxy for
their ratings in an internal evolutionary process. Rational
curiosity would describe NEvAr as a creative system, not
because it produces novel and valuable images, but because,
at the knowledge level, the system appears to be
rationally soliciting reactions on believed high-valued images
and incorporating the responses in a way that transforms
the space of images that the system will next produce–it
behaves consistently with the explanation that it is
rationally pursuing its curiosity (albeit with limited design
knowledge storage capabilities). If redesigned from scratch
with rational curiosity in mind, the system might incorporate
a more interpretable representation of learned knowledge
(for which it is easier to read as a design theory that
improves over time) and put more computation into experimental
design, reasoning over the expected knowledge
gain from enticing the human audience to provide feedback
on a particular work rather than always trying to display
the estimated-best available artifacts. Orienting the system
around active learning, we predict, would improve the system’s
apparent creativity.
 From our perspective, it is natural to ask what a system
learns as it runs. Though established techniques in computational
visual art such as design grammars and iterated
function systems can, in some cases, produce very interesting
(valuable) images, the static nature of these techniques
in isolation implies that, over time, our sense of novelty of
the kinds of artifacts these techniques produce will necessarily
wane because these techniques do not learn. While
rational curiosity would deem a technique that merely
samples a fixed generative space uncreative, these techniques
are still valuable to us–they encode very rich design
spaces that, upon gaining experience through experimentation,
a creative agent can alter as part of large-scale experiments
in the design of these generative spaces.
Conclusion
We have followed the clues embedded in the practices of
human game designers to a set of building blocks for a
theory of creative game design. To recap:
1) Game designers are really designers of play.
2) Creative game design is about seeking knowledge.
Proceedings of the Second International Conference on Computational Creativity 20
3) Designers act as explorers in a science of play.
4) Automated discovery systems inspire a computational
model of creative design.
5) Game design activity can be explained as the rational
pursuit of design knowledge gain.
This has led us to a new statement about creativity that
can apply to human or machine design agents in any artifact
generation domain: creativity is the rational pursuit of
curiosity, a knowledge-level phenomenon.
 We hope this model of creativity will inspire the exploration
of discovery system architectures for artifact generation
systems and the development of a new space of knowledge-oriented
creativity support tools.
Acknowledgements
This work was supported in part by the National Science
Foundation, grant IIS-1048385. Any opinions, findings,
and conclusions or recommendations expressed in this material
are those of the authors and do not necessarily reflect
the views of the National Science Foundation.
<references_biblio/>
References
Berlyne, D. E. 1960. Conflict, arousal and curiosity,
McGraw-Hill.
Berlyne, D. E. 1971. Aesthetics and psychobiology, Appleton-Century-Crofts.

Boden, M. A. 2004. The creative mind: myths and mechanism,
2nd ed., Routledge.
Bryant, C. H.; Muggleton, S. H.; Oliver, S. G.; Kell, D. B.;
Reiser, R. D.; and King, R. D. 2001. Combining inductive
logic programming, active learning and robotics to discover
the function of genes. Electronic Transactions on Artificial
Intelligence 5(B1):1-36.
Colton, S. and Bundy, A. 2000. On the notion of interestingness
in automated mathematical discovery. International
Journal of Human Computer Studies.
Colton, S. 2008. Creativity versus the perception of creativity
in computational systems. Creative Intelligent Systems:
Papers from the AAAI Spring Symposium, 14-20.
Cope, D. 2005. Computer models of musical creativity,
MIT Press.
Dasgupta, S. 1991. Design theory and computer science,
Cambridge University Press.
Epstein, S. L. 1988. Learning and discovery: one system's
search for mathematical knowledge. Computational Intelligence
4(1):42-53.
Fullerton, T. 2008. Game design workshop: a playcentric
approach to creating innovative games, 2nd ed., Morgan
Kaufmann.
Gingold, C. and Hecker, C. 2006. Advanced prototyping.
GDC 2006 lecture.
Hartman, C. O. 1996. Virtual muse: experiments in computer
poetry. Wesleyan University Press.
Langley, P.; Bradshaw, G.; and Simon, H. A. 1984. Rediscovering
chemistry with the bacon system. In Machine
Learning: An Artificial Intelligence Approach, 307-329.
Lenat, D. 1976. AM: an artificial intelligence application
to discovery in mathematics. PhD Thesis.
Loewenstein, G. 1994. The psychology of curiosity: a review
and reinterpretation. In Psychological Bulletin,
116(1), 75-98.
Machado, P. and Cardoso, A. 2002. All the truth about
NEvAr. Applied Intelligence 16, 101-118.
McGraw, G. and Hofstadter, D. 1993. Perception and creation
of diverse alphabetic styles. AISBQ, (85):42-49.
Merrick, K. and Maher, M. L. 2009. Motivated reinforcement
learning: curious characters for multiuser games,
Springer.
Newell, A. 1982. The knowledge level. Artificial Intelligence,
18(1):87-127.
Pease, A.; Winterstein, D.; and Colton, S. 2001. Evaluating
machine creativity. In Workshop of Creative Systems.
Salen, K. and Zimmerman, E. 2004. Rules of play.
Saunders, R. 2002. Curious design agents and artificial
creativity. PhD Thesis.
Schell, J. 2008. The art of game design: a book of lenses,
Morgan Kaufmann.
Schmidhuber, J. Formal theory of creativity, fun, and intrinsic
motivation (1990-2010). IEEE Transactions on Autonomous
Mental Development, 2(3):230-247.
Secretan, J.; Beato, N.; D'Ambrosio, D. B.; Rodrigues, A.;
Campbell, A.; and Stanley, K. O. 2008. Picbreeder. In
SIGCHI Conf. on Human Factors and Computing Systems.
Simon, H. 1996. Sciences of the Artificial, 3rd ed., MIT
Press.
Smith, A. M.; and Mateas, M. 2010. Variations Forever:
flexibly generating rulesets from a sculptable design space
of mini-games. In Proc. of the 2010 IEEE Conf. on Computational
Intelligence and Games (CIG 2010).
Smith, A. M.; and Mateas, M. 2011. Towards knowledgeoriented
creativity support in game design. In Proc. of the
2nd International Conference on Computational Creativity
(ICCC 2011).
Smith, A. M.; Nelson, M. J.; and Mateas, M. 2009. Computational
support for play testing game sketches. In Proc.
of the 5th Annual AI and Interactive Digital Entertainment
Conference (AIIDE2009).
Yeap, Wai K.; Opas, Tommi; and Mahyar, Narges. 2010.
On two desiderata for creativity support tools. In Proc. of
the Intl. Conference on Computational Creativity, 180-189.
Proceedings of the Second International Conference on Computational Creativity 21