Casual Creators
Kate Compton and Michael Mateas
Expressive Intelligence Studio
University of California, Santa Cruz
kcompton/mmateas@soe.ucsc.edu
Abstract
Many creativity tools exist to support task-focused creativity,
but in recent years we have seen a flourishing
of autotelic creativity tools, which privilege the enjoyable
experience of explorative creativity over taskcompletion.
Because these tools are much smaller in
scope, less commercially significant, and less ”serious”
than their larger siblings, they have been overlooked in
academic research. This paper coins the term ”Casual
Creators” for these tools, and provide a definition to
identify tools that belong to this category. We also identify
the particular design considerations that arise from
autotelic creativity, and propose a number of strong design
patterns that serve those considerations, patterns
which are demonstrated by case studies of software built
with those patterns. We believe that once this field is
identified and named, the currently-isolated practitioners
who make these casual creators will be able to share
knowledge, like these design patterns, and develop a
community of practice.
Introduction
An alternative design space
Often when we talk about tools to support creativity, the
‘creators’ exhibiting creativity are task-oriented professionals
or amateurs, who have a specific problem to solve, or
design task to accomplish. There exist many complex, powerful,
and frequently expensive tools for different kinds of
creative tasks: Maya for 3D modeling, FinalCutPro for editing
video, Ableton for music production, Photoshop for images.
These professional tools must support a broad range
of possible actions with a focus on efficient task completion,
as their users are typically being paid to complete predefined
design tasks. Design, as an activity, is ”goal-oriented”, ”intentional,
purposeful” (Gero 1994). ”Task” is an appropriate,
and common, term for the primary action of using these
tools, because the goal is to enable productive labor from the
user.
Is this the only way to use creativity tools? Is productivity
the only goal of creativity?
We propose the category of Casual Creators as an
alternate design space for tools which support creativity
as an intrinsically pleasurable activity, rather than as an
extrinsically-motivated way to accomplish tasks.
Creativity is often an autotelic activity: we paint, draw,
sculpt, sew, write and make music creatively, often with
complete disregard for the quality of the final product (much
less concern for task productivity), because the activity itself
is so enjoyable.
This autotelic, intrinsically-rewarded form of creativity is
psychologically quite distinct from creativity exhibited in
a environment with extrinsic motivation (Amabile, Goldfarb,
and Brackfleld 1990). We should expect that support
tools for autotelic creativity will be correspondingly different.
There is a thriving ecosystem of appropriately-designed
software tools to support task-focused creativity, so why
isn’t there a corresponding set of software to support autotelic
creativity?
One reason is economic: the labor of the creative person is
of commercial value to either their employer or their client,
so there is a market incentive to maximize that output with
effectively-designed tools, either to be purchased by the creative
person themselves (for an independent contractor) or
by their employer.
The other reason is the perceived “seriousness” of output.
This division between serious and frivolous output mirrors
the division in creativity research, between psychological
(“P”) creativity and historical (“H”) creativity (Boden
2009). Psychological creativity is the ‘everyday’ (Kaufman
and Beghetto 2009) creativity of average people, while historical
creativity is the ‘eminent’ creative ability ascribed
to famous world-changing creators and their innovative creations.
Historical creativity is more valued socially and economically,
so when building a tool to support creativity, tool
designers like to imagine that it will be used by the next great
artist or genius inventor (Shneiderman 2003), or at least used
to make some famous or commercially successful product.
We believe, however, that it is also important to build tools
supporting ”everyday” creators in enjoying pleasant and ful-
filling creative exercise, even if they never produce worldchanging
output.
Existing creators
Despite these reasons for not receiving ”serious” attention,
there are many small applications that do exist to support casual
creativity. The recently-developed app marketplaces for
mobile have provided a perfect haven for this sort of creativity
app. These apps each support creating only a single kind
Proceedings of the Sixth International Conference on Computational Creativity June 2015 228
of artifact, such as abstract generative pictures (Secretan et
al. 2011), virtual pottery (”Let’s Create! Pottery” 2014),
or 3d printable bracelets (System 2015). These apps create
artifacts from a greatly-reduced possibility space, compared
to the previously mentioned general-purpose professional
tools. The narrowness of the possibility space allows
the tool to provide greater support for the user, eliminating
potential bad artifacts and speeding the process of creating
good ones, at the expense of flexibility and versatility. This
loss is acceptable, however, as these products aren’t being
created in response to the exacting demands of a boss or
client, but rather because they are intrinsically fun to make.
In fact, the end product may be discarded entirely after completion!
(Nilsson 2003).
Other examples of autotelic creativity tools have come
from games, a field that has always valued pleasurable user
experience over productivity. Not every game contains creative
activities, but many games do feature the creation (and
curation) of a house, creature, or avatar, and these creativity
tools can end up being more fun than the eventual gameplay.
And yet, though these tools exist, there is no central community
in which their designers can communicate shared
knowledge. There is no set of best practices that can be referenced
by those who are attempting to make such a tool, or
even a name for them so that the tool-maker can describe
what they are making. The World of Warcraft (Blizzard
2004) character-creator and the ”Let’s Create! Pottery” virtual
potter’s wheel (to choose two out of many, many existing
tools) may not seem to have much in common at first
glance, or even share the same marketing category. We hope
that by identifying this software genre, including lessons and
design patterns learned from existing examples, we can de-
fine a distinct area for future research and tool implementation.
Introducing Casual Creators
Casual creators can be distinguished from other creativity
support tools by their goal of supporting autotelic creativity,
not task-focused creativity. From this initial difference
in goals arises a variety of other differences: in design considerations,
optimal design patterns, and the psychological
states that they encourage in the user. To that end, we propose
the following definition, which encapsulates the very
exciting alternate design space of Casual Creators.
Definition
A Casual Creator is an interactive system that encourages
the fast, confident, and pleasurable exploration of
a possibility space, resulting in the creation or discovery
of surprising new artifacts that bring feelings of
pride, ownership, and creativity to the users that make
them.
Casual Creators are interactive systems. There are historical
examples of non-digital casual creators, like the classic
generative art toy Spirograph or the knitting toy Knit Magic,
though digital software systems provide more affordances.
They do, however, need to be interactive, driven by the user,
because the learning and creating process is so core to the
psychological experience of using one. Computational creativity
can be used to assist the design process, but must be
in a mixed initiative partnership with the user.
Casual Creators are tools that create artifacts, of some
kind, which may be instances of virtual models or static images,
or more abstract artifacts like story grammars or AI
behaviors. Each creator has some possibility space, the set
of all possible artifacts that could be created using that tool.
The user creates (or discovers) artifacts by searching or exploring
the space for ’good’ artifacts. For the casual creator
to be successful tool, there must be a way for users to find
artifacts meeting functional and aesthetic criteria, avoiding
getting stuck in a space of bad artifacts.
The possibility space should be narrow enough to exclude
broken artifacts (such as models that fall over or break when
3D printed) but broad enough to contain surprising artifacts
as well. The surprising quality of the artifacts motivates the
user to explore the possibility space in search of new discoveries,
a motivation which disappears if the space is too
uniform. It also provides feelings of ownership and creativity
when the artifact is discovered. In a sufficiently multidimensional
possibility space, ‘search’ and ‘creation’ become
blurred, as the only way to arrive at a particularly interesting
artifact is to move through the space intentionally, rather
than randomly searching. The user will feel greater ownership
and creativity the more they attribute their discovery to
their own actions, and pride is increased further when they
feel that their discovered artifact is somehow special, surprising
within the possibility space.
How does the user navigate this possibility space? Do
they make tiny adjustments, tentatively inching through the
possibility space, or do they make wild jumps, from solution
to novel solution, exploring large regions of the space over
a short period of time? An optimal creative process is described
as making ‘creative leaps’, so we want to guide the
user toward a fast-moving and confident exploration of the
possibility space. The user’s experience should feel playful,
powerful, and pleasurable, like a flow state.
The user of a casual creator is a casual user, and the system
can expect no previous domain knowledge, no previous
technical experience, or adherence to a long learning process.
All of the learning and creativity described above must
occur in the first few minutes, and provide a good experience
even if the user never spends time to gain mastery.
This definition focuses on the design goals of a Casual
Creator, the experiences that Casual Creators are particularly
suited to create. How we can design tools that achieve
these goals is explained in the rest of this paper, through a
description of design patterns and case studies that tested
them. Some patterns come from existing Casual Creatorlike
tools, like the Spore Creature Creator, Nervous System’s
design tools (System 2015), and academic experiments like
Picbreeder (Secretan et al. 2011). More patterns come from
our current understanding of autotelic creativity, anticipating
design patterns that support such a psychological state. To
predict these new potential patterns, we draw from existing
theories of creativity, flow, and design.
Proceedings of the Sixth International Conference on Computational Creativity June 2015 229
Related Fields
Reflection-in-action and Direct Manipulation
During the creative design process, the user modifies the artifact,
moving quickly through a cycle of evaluation, planning,
modification and reevaluation. This process shows the
“Reflection-in-action” theory of learning, in which a learner
hypothesizes, acts, and reflects on the results as a way to iteratively
understand a domain or problem. The originator of
that theory, Donald Schon, also applied it to the process of
designing (Schon 1992) in which the designer ”sees, moves
and sees again”. The seeing and moving are grounded in the
materials themselves. This cycle cannot take place disembodied
in the mind but must be enacted in dialogue with the
artifact.
Seeing may include the user’s visual perception of the artifact,
but is also a way to describe the evaluation of the artifact.
Is it aesthetically pleasing, stable, strong? How does
the designer predict that it will perform in its intended role?
Some of these evaluations could be performed or assisted
computationally. Schon surmises that while the reflective
design process itself is not well suited to unsupervised computational
processes, computation could provide new ways
of “seeing”, or provide constrained micro design spaces “extending
the designer’s ability to construct and explore them.”
He concludes that “The design of design assistants is an approach
that has not in the past attracted the best minds in AI.
Perhaps the time has come when it can and should do so.”
In Direct Manipulation (Shneiderman 1993), a UI concept
that parallels reflection-in-action, a complex software
system provides ”continuous representation of the object
of interest” and “[r]apid, incremental, reversible operations
whose impact on the object of interest is immediately visible”
and promises that “after obtaining reinforcing feedback
from successful operation, users can gracefully expand their
knowledge of features and gain fluency.”
The user can manipulate the system with rapid operations,
then evaluate the effect immediately, because the artifact is
always visible and responds immediately to the modification.
The actions are reversible so the user is encouraged to
experiment without anxiety, incremental, so subtle changes
can be observed, and rapid (and rapidly seen), so that this
learning cycle can operate continuously with each tiny iteration.
Flow
Csikszentmihalyi’s Flow theory is influential in games and
creativity studies, but seems particularly well suited to the
autotelic creativity of casual creators as“[i]deally, flow is the
result of pure involvement, without any consideration about
results.”(Csikszentmihalyi 2000) For flow to be achieved,
the activity must have goals to create a sense of progress,
immediate feedback so that progress can be sensed, and a
balance between their perceived skills and challenges. Flow
can be disrupted if the user feels frustrated, intimidated, or
overwhelmed by choices.
Flow has a complex relationship with goals. Though the
activity should be enjoyable in itself, without the pursuit of
an outside reward, goals provide the required direction and
progress. Goals can be provided as preset challenges, but often
it is better to encourage the user to develop their own internal
design goals. A good goal can be evaluated momentto-moment,
may change over time, and can be either highly
specific to the user, or come from knowledge of the design
space.
The flow state is very conducive to both creativity and
an autotelic experience, and so provides important design
considerations for potential Casual Creators, especially for
avoiding conditions which disrupt flow, like choice paralysis
or hard failures.
Creativity Support Tools
Lubart (Lubart 2005) identifies several categories of human
and computer collaboration in the creative process: the computer
can act as nanny, coach, pen-pal or colleague. Riedl
and O’Neill (Riedl and ONeill 2009) suggest ”audience” as
a fifth role for the computer. These categories provide a useful
taxonomy, but do not provide implementable patterns.
The field of Creativity Support Tools, of which Casual
Creators could be considered a subcategory focused on autotelic
creativity, provides many concrete design patterns.
Resnick et al (Resnick et al. 2005) identify many such patterns.
Some, like “Support Exploration” and “Make It As
Simple As Possible - and Maybe Even Simpler” are patterns
to support flow experiences and reflection-in-action styles
of learning. Other principles like “Support Many Paths and
Many Styles”, “Low Threshold, High Ceiling, and Wide
Walls” reflect how users will start with diverse goals and
skills, which will further evolve as they use the system.
Some principles, “Choose Black Boxes Carefully,” “Support
Collaboration,” and “Support Open Interchange”, ask
the designer to reflect on the communities in which creative
collaboration and learning occur, and how creativity develops
as multiple users share knowledge. When we look at
the creative communities that flourished for tools like Spore,
Scratch (Resnick et al. 2009), and Twine (Klimas 2012), it
becomes clear that, though the design of the single-user software
is important, the technology decisions of data format,
data interchange, hosting, and modifiability are equally critical
to enabling creativity and fostering ownership. Creativity
occurs between the user and client-side application, but
also in the communities of practice that develops outside of
the app, so creativity support tools must consider both sites,
personal and communal (Maher 2012).
Generative Methods and Computational Creativity
Computational creativity is the science (and art) of encoding
human-style creative process as automatable systems,
with the goal of building a system which “exhibits behavior
that would be deemed creative in humans.” (Colton et al.
2009). How ‘creativity’ can be detected in the finished artifacts
of these systems is its own difficult problem (Maher
2012), but the field has successfully built generators that can
design artifacts for domains as diverse as jokes (Petrovic and
Matthews 2013) and paintings (Colton 2012).
These systems create artifacts by encoding the process of
creating art (or literature, jokes, game levels, music, etc).
The resulting algorithms must be able to create not only one
Proceedings of the Sixth International Conference on Computational Creativity June 2015 230
successful example, but a wide and interesting space of possible
valid artifacts, some of which should be able to surprise
even the person who wrote it. Such algorithms can be called
generative methods (Compton, Osborn, and Mateas 2013);
they use a range of technologies (genetic algorithms, grammars,
declarative modeling), but all share the goal of creating
large possibility spaces of valid-yet-surprising artifacts.
This is the optimal type of possibility space for computational
creativity systems, and also for Casual Creators.
Computational creativity and generative methods are often
a poor fit for productivity-focused creativity apps. Professional
creativity involves creating to very specific requirements,
requires complete control and the ability to fine tune
the resulting work. Generative methods create a lot of work,
very fast, but with minimal control over the output (compared
to hand-authored content) and often no way to iterate
on the output. A casual user, without the need for complete
control, is willing to trade a loss of control for the speed,
power, and surprise of generative methods.
The expressive range of such systems must always be balanced
with the need to produce valid content. A system
could produce a wide variety of mostly broken artifacts, or
produce a set of high-quality yet homogeneous artifacts, but
both of these are failures. We have found the phrase ‘1000
bowls of oatmeal’ useful to describe the common antipattern
of generating a set of artifacts which are technically distinct
to the computer, but perceived by humans as uniform.
Computational creativity systems usually run autonomously
and unsupervised by humans. Pairing these
methods with human users can add additional power to the
process (Davis et al. 2014), as humans provide aesthetic
evaluations and intuitive leaps to the rapid generativity
of the computation creativity processes. Mixed initiative
systems, in which the computer and human users operate
simultaneously or by turn-taking, support a creative cycle
in which each user reflects on the previous contributions
of their collaborator and modifies the artifact according to
their particular abilities. The end products of the creative
process are improved, and ideally the user enjoys the
experience of collaboration, if the system is well designed.
Interaction with a highly generative system has a particular
set of pleasures, whether in the context of a game or a
creativity tool. Chaim Gingold refers to such pleasurably
interactive systems as ‘Magic Crayons’ (Gingold 2003):
computational, accessible, sketchable, expressive systems
which invite the user to play with them and discover hidden
secrets and affordances.
Design Patterns
The definition of a Casual Creator as an autotelic creativity
tool provides an abstract guide for what we would want a
potential Casual Creator to accomplish. To actually design
such a tool, these high-level patterns must be interpreted into
concrete design patterns. We have identified a number of
these patterns, drawing from existing Casual Creators, and
from the related fields, and tested them by using them to create
a wide variety of systems, described in the Case Studies
section below. These design patterns are not exhaustive, but
are representative ones that are versatile, common, and easy
to apply.
Instant feedback Recall that both direct manipulation and
reflection-in-action require the user to observe the artifact,
make a change, and see the results, a process which allows
them to discover patterns and affordances in their possible
changes, mastering the system while iterating on an artifact.
In the instant feedback pattern, the changes should be immediately
visible in the modified artifact. However, just visually
regenerating the artifact in response to changes, even in
real-time, is not necessarily enough to provide appropriate
feedback.
‘Seeing’, in the reflection-in-action model, encompasses
more than just ‘looking at’. ‘Seeing’ actually encompasses
the entire process of sensing and evaluating the artifact’s fitness
according to both the potential use case and the user’s
own design model. For objects with a strictly aesthetic role,
this is easy: the user glances at it, and can instantly decide
their opinion of it. Other evaluations are complex, and must
be either mentally simulated by the user, or else evaluated
by the system. Requiring the user to mentally simulate complex
consequences will take a lot of time and attention, and
the evaluation could be inaccurate or flawed, slowing the iteration
process. The instant feedback pattern would recommend
computationally simulating and visualizing as much
as possible so that the user can get feedback at a glance.
The Chorus Line Named after the choreography concept
in which many dancers all execute the same routine simultaneously,
the chorus line pattern was used in Spore (Hecker
et al. 2008) as an internal tool to test animations on a wide
range of creature morphologies. The chorus line is a subpattern
for instant feedback, in situations where what is being
generated is not a single artifact, but a space of artifacts.
In that situation, the user should be able to ‘see’ (in
the reflection-in-action sense) the space of their creation, instantly.
Instead of generating one example, this pattern suggests
generating many examples, and overlaying them (spatially,
temporally, graphically) to make subtle differences
and similarities easier to spot.
Simulation and approximating feedback Automated visualization
becomes especially important when the artifact
being generated would take minutes or even hours for the
user to evaluate, rather than milliseconds for an image, or
seconds for an animation. For artifacts such as game levels,
the artifact is judged by the many gameplay traces over
time that could be played on it, which cannot be visually
evaluated with much accuracy by a casual user. Nor can a
system show the user all possible gameplay traces, so the
user must be shown a proxy of the evaluation. When Riedl
and O’Neill (Riedl and ONeill 2009) add ‘computer as audience’
to Lubart’s categories, their simulation proposed to
accurately model how a human reader would evaluate generated
stories. In Sentient Sketchbook (Yannakakis, Liapis,
and Alexopoulos 2014), the system calculates “navigational
and topological properties” as the user interacts with it, providing
instant feedback for a complex artifact. This evaluation
does not fully encapsulate the actual gameplay imProceedings
of the Sixth International Conference on Computational Creativity June 2015 231
plications of the map, which for a finished level being put
in a game, could be a potential design issue. However, for
a Casual Creator, the goal of the evaluation is to provide
the sense of progress towards a goal necessary for achieving
flow. Only the perception of progress is necessary: as long
as the user perceives progress, the accuracy of the evaluation
is irrelevant.
Entertaining evaluations One nice benefit of relaxing the
need for accurate evaluations is that the evaluations can
themselves be pleasurable and entertaining. In the Spore
Creature Creator, when the user modifies their creature the
creature will respond by laughing and shaking the new body
part in appreciation, or, less commonly, expressing distress.
The choice of happy or sad reaction does not actually represent
any real system state, it just provides arbitrary feedback.
That feedback is psychologically significant, for encouraging
the flow state, but also for letting the user feel
pride in pleasing their little AI judges. Even if the user starts
with no particular design direction of their own (a common
issue with casual artists) having a simulated critic present
can suggest a direction for the user, even if they choose
to ignore it. The abstract generative art game BECOME
A GREAT ARTIST IN JUST 10 SECONDS (Brough and
McClure, 2014) waggishly compares the user’s glitch art to
classic masterpieces and rates it with a percent similarity,
an intentionally arbitrary metric that still serves to provide
optional direction to the casual user.
No blank canvas One benefit of focusing on these
intrinsically-motivated users is that they are often much
more flexible about the final product. In contrast to a system
like Maya, which must support extremely broad use cases
and a high degree of fine-tuning in order to make a very particular
finished product, a casual user will have more flexible
requirements for their product. They likely want it to be
functional and aesthetically pleasing, but are willing to consider
many more possible kinds of solutions, or may not even
start with any particular solution in mind (Nilsson 2003)
Professional artists know the terror that comes from facing
a blank canvas (Bayles, Orland, and Morey 2012), but
this experience is also intimidating and paralyzing to the
novice user. However, this can be very easily mitigated, by
providing either a starting shape (Spore) or a suggested challenge
(Let’s Create! Pottery). The first move is the hardest,
so this restricts the first move to a single decision: accept the
prompt, or discard it. Once this one move is taken, subsequent
actions are easier.
Limiting actions to encourage exploration This restriction
of actions can be useful even after the user has moved
past the blank-canvas moment. To achieve a flow state
(Csikszentmihalyi 2000), the user should be able to quickly
and confidently make decisions, which is easier if the available
choices are appropriate, limited in number, and their
consequences are clear (or at least suggested) to the user.
One strategy is modal interaction: limiting actions by the
particular mode that the creator is in. This approach is common
in character creators like that in World of Warcraft and
Spore Creature Creator, which have different modes corresponding
to user actions like painting or building and panels
with sub-actions within those modes, to choose hair or faces,
revealing only actions for the mode that the user is currently
in. Another approach is to limit the actions available, slowly
unlocking them in response to experience, challenges defeated,
purchases, or some other pacing mechanism. If the
possibility space is temporarily restricted, the ability to more
fully explore the space scaffolds the user’s understanding of
the possibility space.
Figure 1: Top: Mutant-shopping for images in Picbreeder.
Though the user can control the rate of mutation, they can
only ‘create’ an image by selecting the parents of the next
generation. Below: Sentient Sketchbook shows automated
evaluations, allows direct editing, and also provides some
alternate mutants on the right
Mutant shopping One feature that can help the user find
unexpected solutions in the possibility space is not a creative
‘action’ at all, but the availability of suggested alternatives,
like artifacts near the current one in the possibility space. In
some tools, the user is not given any way to edit the artifact,
and must navigate the possibility space by picking one of the
new options, as in Picbreeder (Secretan et al. 2011).
In other cases, as in the parametric tree modeler Dryad
(Talton et al. 2008), the user may use these alternative to
browse the space, but can also further edit the artifacts that
they discover in that way. A third framing of this pattern
is found in Sentient Sketchbook (Yannakakis, Liapis, and
Proceedings of the Sixth International Conference on Computational Creativity June 2015 232
Alexopoulos 2014), in which the user edits a game level normally,
but the system uses that information to generate additional
suggested artifacts that are ‘nearby’ for some more
abstract calculated metric, rather than ‘nearby’ in their underlying
representation.
Although this process has a lot in common with evolutionary
algorithms (specifically human-guided evolutionary
algorithms (Klau et al. 2010)), the focus is not on producing
an optimal specimen, but on the enjoyment that the user
feels from this process. For this reason, we named this pattern
mutant shopping to capture the psychological pleasures
and motivations of a less-directed browsing and discovery
process like shopping.
Modifying the meaningful In Spore, parts can be placed
anywhere on the creatures, then modified by rotation or
pulling on their morph handles. In a traditional sculpting
program like Maya, these handles would be expected to control
a clear parameter like z-scaling, for maximum control
over the changes. The Spore designers discovered that it
was more interesting to have these handles control higher
level changes, like shifting a jaw from top-heavy overbite to
jowly underbite, or extending a foot’s shape from round toes
to pointed claws. Higher-level modifications like these give
the user a more meaningful space to explore.
Saving and sharing As noted in the “Design principles
for tools to support creative thinking” report, the client-side
application where the user is editing their artifact is only one
site where creativity occurs, and designers of Casual Creators
should also consider how they support creativity outside
of their app. One example of this principle is the use
of common, free, human-readable filetypes for saving data,
such as JSON or images. Spore embedded the creature’s
save data stenographically in a PNG image, and the latest
version of Twine 2 embeds the editable hypertext into the
HTML that plays the Twine game. Even if the client app
is still necessary to rebuild the content from the saved data,
as in Spore and Twine, users can share their data using existing
platforms. Most hosting sites allow text and images,
but not arbitrary files. If users can easily host their save files
on such hosting sites, they can build communities independently
from the makers of the original casual creator app.
Hosted communities An alternate pattern is to provide a
hosted community that is tied more closely to the client app,
as Picbreeder and Let’s Create! Pottery do. Casual creators
should encourage the user’s pride in their discovered or
created artifacts, so providing a showcase where user’s can
publish their work to share it with others supports this feeling
of ownership. Creations are often annotated or tagged,
and usually there is a commenting and messaging system,
enabling a large community to communicate within itself.
Modification is its own form of communication, so if the
system supports modification of artifacts, they should show
their ancestry, and notify the original creators so that they
can take pride in their influence.
Modding, hacking, teaching Users of casual creators will
quickly find that the tool does not support every action that
they want. The tool and its surrounding community support
should facilitate users in teaching each other mods and hacks
that expand the boundaries of what’s possible with the tool.
The previous two patterns support this pattern, as this teaching
can happen on external sites, or internal ones, but the
easier it is to find a clever hack, import it into the tool, and
modify it and republish the new results, the quicker these
ideas will spread through the community.
Case Studies
Instant Feedback: PendantMaker
PendantMaker is an online design tool for creating 3D printable
pendants. We observed that although 3D printing is interesting
to many people, the tools to create printable content
are difficult, with many potential pitfalls for making unprintable
and broken content. By restricting the domain space to
extruded tubes, we could guarantee that our generated geometry
would be valid for printing, and print reliably on a cheap
printer (a difficult set of physical constraints). When combined
with turning sliders, supporting the Direct Manipulation
patterns of “rapid, incremental, reversible operations”
(Shneiderman 1993), PendantMaker provided a very ‘safe’
place for the user to experiment without fear of failure.
We also noted that casual users often doubt their drawing
ability (Bayles, Orland, and Morey 2012) and lack direction,
so we designed a generative algorithm in which undirected
scribbles from the user would be reflected around an
axis, creating a design of surprisingly attractive symmetry.
We provide a canvas for the user to draw a line, which is
extruded, shaped, and reflected into the many intersecting
tubes on the right, creating the printable pendant in real time.
This very immediate feedback was critical: users could draw
aimlessly, but notice when the reflections would intersect or
join together, allowing the users to easily create a complicated
knotwork of intersecting tubes that would be impossible
to predict without feedback. We also added sliders for a
variety of tuning values, reflecting the Modifying the Meaningful
pattern above. Some sliders corresponded to clear values
like thickness and arm count, but ‘bloom’ performed a
complicated sculptural task of flaring the outermost tubes
in a curved shape. Complicated tools like bloom are only
usable with rapid feedback: their action is indescribable to
the user, but with a little experimentation, the user quickly
learns how to use them artistically.
Sharing and Ownership: IceMaker
IceMaker was an evolution of PendantMaker’s design, to
create extruded 3D snowflakes, and similarly uses tuning
sliders, symmetry, and extrusion to create complex geometry
that is both modifiable and guaranteed valid, with immediate
feedback. The extrusion path is not controlled by
the user’s drawing, as in PendantMaker, but rather by a particle
simulation. The behavior of the particles would be very
hard for a casual user to program, so instead we provide sliders
for values that represent the resulting appearance of the
path (‘complexity’, ‘wiggle’, ‘sharpness’) allowing the user
to explore the possibility space while not having to understand
the complex process behind it (Fig. 2).
Proceedings of the Sixth International Conference on Computational Creativity June 2015 233
Figure 2: Ice-Maker, a 3D snowflake maker, guides the user
to create a snowflake and further personalize it with a message,
then embed the design into a single URL that the user
could share.
Since this interaction provides less agency than the drawing
interface in PendantMaker, we wanted to augment IceMaker
with other ways to declare ownership of the discovered
snowflake. Following the Saving and Sharing pattern,
we encoded the snowflake into a unique URL which the user
could share, post or send as a ‘saved’ version of their artifact.
Search and Discovery: Funky Ikebana and Tiny
Dancer
Creativity-as-discovery is further explored in Funky Ikebana,
in which L-system flowers are are generated from a
‘DNA’ of floating point tuning values. Similar to exploration
process in Dryad (Talton et al. 2008), the user iteratively
selects the flowers that they like, and the system generates
more nearby examples. This human-guided evolutionary algorithm
allows for ‘optimization’ of the flower, but as this
was designed as a Casual Creator, we focus on the pleasures
of mutant shopping more than potential optimization. The
flowers are arranged together, which makes different ones
easy to spot, so the user can pick from flowers that are very
different, or mostly the same. Regeneration when one flower
is selected and its children repopulate the space is instantaneous,
so the user can very quickly move through the space
of flowers. Picking from one of 10 children limits the number
of actions that the user can take, so choice paralysis does
not occur, as shown in previous mutant shopping examples
like Picbreeder (Secretan et al. 2011).
Unlike the Dryad and Picbreeder systems, we were also
able to use the L-system to create a simple animation for
the flowers, causing them to ‘dance’. Flowers danced differently,
an emergent property of their morphology, and users
could selectively evolve flowers for their movement instead
of just shape. Because we used the chorus line pattern
to show many flowers dancing at once, the user was able
to notice particularly graceful or vigorous ones, and select
for that. Tiny Dancer takes this idea further by simultaneously
evolving the morphologies of ragdoll dancers and their
dance-responses to music, so that the dances can also be selected
by mutant shopping on a chorus line.
Figure 3: Iteratively evolving smaller flowers in Funky Ikebana,
starting with the center flower in the first image. The
user’s current heuristic is to pick small simple flowers, but
that heuristic can change each time the user spots a flower
style that they like better.
Interventions: BotPrint and Binary Fission
The Casual Creator framework has been usefully applied as
an intervention in two existing designs, successfully modifying
the designs to improve the user’s creative experience.
BotPrint was an existing application to design lasercuttable
robotics kits for children. Users could drag handles
to shape the outline of the bot’s chassis, and some automation
would occur to figure out placement of components.
Unfortunately, the implications of moving components and
changing chassis size were not visible to the users, so making
modifications felt meaningless. Using casual creator design
patterns, we updated the system to simulate the bots
moving in an ‘arena’ with many other similar bots. This
provided a way for the user to evaluate the behavior of the
bots visually (chorus line), see and select variants (mutant
shopping) and enjoy watching the bots struggle for victory
(entertaining evaluations), while also directly modifying the
bots and then rereleasing them into the arena.
Binary Fission is a game designed to help the user make
binary decision trees to filter loop invariant data for a crowdsourced
science task about software verification. At first,
this does not seem to be a creative task, but by using casual
creator patterns to emphasize the creative side of selecting
the filters to build the filter tree, users enjoyed the
task and were able to explore the possibility space of trees
much faster. The biggest insight provided though the casual
creator lens was to show many filters for each choice point.
Calculating how well a filter would behave is a very arduous
evaluation for the user to perform themselves, so we colored
Proceedings of the Sixth International Conference on Computational Creativity June 2015 234
each by how well it filtered data at that point. Users were
able to glance through this potential ‘filter space’ for suitable
filters, and were able to apply them, see their implications,
and rebuild trees very quickly, turning what could have been
an opaque and arduous task into a fun reflection-in-action
learning experience.
Conclusion
This paper defines a new term, Casual Creators, to identify
a category of interactive systems which prioritizes the
experience of autotelic creativity above productive output,
an exciting new design space that is distinct from existing
productivity-focused creativity support tools. We have illustrated
the distinct design considerations of Casual Creators
by identifying and describing representative design patterns
drawn from theories of creativity and current successful systems.
These patterns were used to design several new systems,
and to evolve some existing designs to better support
casual creativity. From these case studies, we learned that
these patterns do clarify and inspire the process of building
systems to support casual creativity, as it was easy to identify
new system features from the patterns. Additionally, using
the lens of Casual Creators enabled us to easily find examples
of those features in a wider range of existing systems
than would have otherwise been possible.
<references_biblio/>
References
Amabile, T. M.; Goldfarb, P.; and Brackfleld, S. C. 1990.
Social influences on creativity: Evaluation, coaction, and
surveillance. Creativity research journal 3(1):6–21.
Bayles, D.; Orland, T.; and Morey, A. 2012. Art & fear.
Tantor Media, Incorporated.
Boden, M. A. 2009. Computer models of creativity. AI
Magazine 30(3):23.
Colton, S.; Lopez de Mantaras, R.; Stock, O.; et al. 2009. ´
Computational creativity: Coming of age. AI Magazine
30(3):11–14.
Colton, S. 2012. The painting fool: Stories from building
an automated painter. In Computers and creativity. Springer.
3–38.
Compton, K.; Osborn, J. C.; and Mateas, M. 2013. Generative
methods. In The Fourth Procedural Content Generation
in Games workshop, PCG.
Csikszentmihalyi, M. 2000. Beyond boredom and anxiety.
Jossey-Bass.
Davis, N.; Popova, Y.; Sysoev, I.; Hsiao, C.-P.; Zhang, D.;
and Magerko, B. 2014. Building artistic computer colleagues
with an enactive model of creativity. In Proceedings
of the 5th International Conference on Computational
Creativity.
Gero, J. S. 1994. Towards a model of exploration in
computer-aided design. In Formal design methods for CAD,
315–336.
Gingold, C. 2003. Miniature gardens & magic crayons:
Games, spaces, & worlds. Ph.D. Dissertation, Georgia Institute
of Technology.
Hecker, C.; Raabe, B.; Enslow, R. W.; DeWeese, J.; Maynard,
J.; and van Prooijen, K. 2008. Real-time motion retargeting
to highly varied user-created morphologies. In ACM
Transactions on Graphics (TOG), volume 27, 27. ACM.
Kaufman, J. C., and Beghetto, R. A. 2009. Beyond big
and little: The four c model of creativity. Review of general
psychology 13(1):1.
Klau, G. W.; Lesh, N.; Marks, J.; and Mitzenmacher,
M. 2010. Human-guided search. Journal of Heuristics
16(3):289–310.
Klimas, C. 2012. Twine. Accessed: 2015-02-24.
Lubart, T. 2005. How can computers be partners in the creative
process: classification and commentary on the special
issue. International Journal of Human-Computer Studies
63(4):365–369.
Maher, M. L. 2012. Computational and collective creativity:
who’s being creative? In Proceedings of the 3rd International
Conference on Computational Creativity, 67–71.
Nilsson, B. 2003. ‘i can always make another one!’–young
musicians creating music with digital tools. Musicianship
in the 21st century: issues, trends and possibilities (Sydney,
Australian Music Centre).
Petrovic, S., and Matthews, D. 2013. Unsupervised joke
generation from big data. In ACL (2), 228–232.
Resnick, M.; Myers, B.; Nakakoji, K.; Shneiderman, B.;
Pausch, R.; Selker, T.; and Eisenberg, M. 2005. Design
principles for tools to support creative thinking.
Resnick, M.; Maloney, J.; Monroy-Hernandez, A.; Rusk, ´
N.; Eastmond, E.; Brennan, K.; Millner, A.; Rosenbaum, E.;
Silver, J.; Silverman, B.; et al. 2009. Scratch: programming
for all. Communications of the ACM 52(11):60–67.
Riedl, M. O., and ONeill, B. 2009. Computer as audience:
A strategy for artificial intelligence support of human creativity.
In Proc. CHI Workshop of Computational Creativity
Support.
Schon, D. A. 1992. Designing as reflective conversation
with the materials of a design situation. Research in Engineering
Design 3(3):131–147.
Secretan, J.; Beato, N.; D’Ambrosio, D. B.; Rodriguez, A.;
Campbell, A.; Folsom-Kovarik, J. T.; and Stanley, K. O.
2011. Picbreeder: A case study in collaborative evolutionary
exploration of design space. Evolutionary Computation
19(3):373–403.
Shneiderman, B. 1993. Direct manipulation: a step beyond
programming languages. Sparks of innovation in humancomputer
interaction 17:1993.
Shneiderman, B. 2003. Leonardo’s laptop: human needs
and the new computing technologies. Mit Press.
System, N. 2015. Nervous system.
Talton, J.; Gibson, D.; Hanrahan, P.; and Koltun, V. 2008.
Collaborative mapping of a parametric design space. Technical
report, Citeseer.
Yannakakis, G. N.; Liapis, A.; and Alexopoulos, C. 2014.
Mixed-initiative cocreativity. In Proceedings of the 9th Conference
on the Foundations of Digital Games.
Proceedings of the Sixth International Conference on Computational Creativity June 2015 235