RumbleBlocks: Teaching science concepts to young children through a Unity game

Abstract

RumbleBlocks was developed at the Entertainment Technology Center (ETC) to teach engineering principles of tower stability to children ages 4-7. The game features tower construction, tower piece removal, and tower comparison levels which were designed with feedback from early childhood educators and learning researchers, and iteratively improved with feedback from child play tests. This paper emphasizes the development process, and initial formative play tests with children. It was developed using the Unity3D game engine, allowing for export as a stand-alone application, web player, or to mobile devices. First results are promising in terms of educational effectiveness, with more studies planned for the future.

Full text

RumbleBlocks: Teaching Science Concepts to
Young Children through a Unity Game
Michael G. Christel, Scott M. Stevens, Bryan S.
Maher, Sean Brice, Matt Champer, Luke Jayapalan,
Qiaosi Chen, Jing Jin, Daniel Hausmann, Nora
Bastida, Xun Zhang
Entertainment Technology Center
Carnegie Mellon University
Pittsburgh, PA, USA
Vincent Aleven, Kenneth Koedinger, Catherine
Chase, Erik Harpstead, Derek Lomas
Human-Computer Interaction Institute
Carnegie Mellon University
Pittsburgh, PA, USA
Abstract—RumbleBlocks was developed at the Entertainment
Technology Center (ETC) to teach engineering principles of
tower stability to children ages 4-7. The game features tower
construction, tower piece removal, and tower comparison levels
which were designed with feedback from early childhood
educators and learning researchers, and iteratively improved
with feedback from child play tests. This paper emphasizes the
development process, and initial formative play tests with
children. It was developed using the Unity3D game engine,
allowing for export as a stand-alone application, web player, or to
mobile devices. First results are promising in terms of
educational effectiveness, with more studies planned for the
future.
Educational game, early childhood science education, game
development process, Unity game engine
I. INTRODUCTION
Young children have inquisitive minds, often asking their
family members and teachers questions about science.
Through thoughtful interactions, families and early childhood
professionals can foster a child's scientific thinking, stimulate
curiosity, and establish a foundation for a lifelong interest in
science [1]. Through powerful interactions of being present,
connecting with the child, and offering opportunities to extend
learning, an educator can foster a child's ability to explore,
think, and communicate [2]. The RumbleBlocks development
team began with the challenge of producing a game for
children ages 4-7 that would teach science concepts in ways
that can be tracked by educational researchers, with all the
positive aspects of a game promoting curiosity and
engagement. Can a game offer powerful interactions to the
child, while fostering measurable scientific learning?
RumbleBlocks is a collaborative development between two
departments at Carnegie Mellon University: the Entertainment
Technology Center (ETC), and the Human-Computer
Interaction Institute (HCII), with HCII faculty also a part of the
Pittsburgh Science of Learning. This paper discusses design
decisions for RumbleBlocks made by the ETC and HCII, the
iterative development process involving child playtesters, some
very early formative evaluation work, and concludes with the
next steps for the project as it heads toward more formal
educational evaluation. The template presented here can guide
other game development teams interested in early childhood
science education.
The ETC addressed game development (discussed in this
paper) through two semester-long projects, with the HCII
focused on educational evaluation (including much planned
future work). In the Fall 2011, RumbleBlocks began with the
ETC Illuminate project. In Spring 2012, the ETC Sci-Fri
project continued RumbleBlocks work as well as other science
game efforts. Both projects detailed their weekly progress in
online newsletters, with download links for RumbleBlocks as a
web version, stand-alone PC, Mac, or Android game nested
within the Illuminate pages [3]. The interested reader can
search out the newsletters and play the game for greater detail
and insight behind the points made in this paper.
II. SCIENCE PRINCIPLES UNDERLYING THE GAME
The game developers began with a review of various
United States government and educational advisory board
standards, guidelines, and documents addressing science
education for pre-school, kindergarten, and grades 1-3 (ages 4-
9). They interviewed teachers in this class range, and worked
closely with additional evaluation experts from the Pittsburgh
Science of Learning so that the domain choice would be one
where children's advances in science learning (or lack thereof)
could be adequately tracked during game play. They visited
classrooms for this age range, seeing what the children used for
science learning and play. Brainstorming sessions produced a
number of candidates, including machine repair, levers, states
of matter, day/night and seasonal cycles, causal reasoning, and
electric circuits.
The ideas were collapsed down to a subset that seemed
fruitful to teachers, viable to educational researchers, and with
potential to appeal to both girls and boys in a slightly narrower
ages 4-7 demographic, a potential tested repeatedly in follow-
up interactions with children. In the end, the idea that won out
was a tower-building game. The underlying engineering
principles being taught by the game include the following:
 Expanded base: the base should be wider than the top of
the structure. Towers built more in the shape of a pyramid
tend to be more stable, so that when upper pieces shift,
they have room to do so.

 Symmetry: the structure should be symmetrical in at least
one dimension around the central axis; structures aligned
with equal distribution of weight tend to be more stable.
 Closed gaps: Blocks in the same story of the building
should be touching each other rather than spaced apart.
Interviews with 5-6 year olds that used contrasting cases
showed that these principles are not universally understood
already, i.e., the children made mistakes. Two towers were
shown on paper to a child and he or she was asked which is
more likely to fall. Children's answers were recorded and later
analyzed to uncover misconceptions that may be used to
produce instructional sequences with contrasting cases. The
use of contrasting cases designed to help students notice
information they might otherwise overlook dates to the 1950s
and perceptual learning, while also providing an opportunity to
maintain instructional fidelity and experimental control [4].
The contrasting cases helped to confirm the appropriateness of
the science concepts to be addressed by the game.
III. DEVELOPMENT PROCESS
The ETC offers a two-year professional degree, the Masters
of Entertainment Technology. In pursuit of this degree,
students spend their first semester with four core courses,
including "Building Virtual Worlds" (BVW), followed by three
semesters in which students tackle studio projects like
Illuminate with a small team of artists, game and audio
designers, and programmers. Courses like BVW and the studio
projects teach the value of rapid prototyping and iteration [5].
BVW groups students in teams of four that create 5 separate
prototypes over the course of the semester, all taking no longer
than three weeks from start to finish. This class embodies the
idea that the quicker you get a game fielded, the quicker you
can fail and discover where your initial ideas were ill-
conceived which lessens the impact of course corrections. The
ETC emphasizes the importance of early and frequent iteration
in game design, and the majority of semester studio projects
follow the Agile development process with weekly sprints that
break down tasks into small increments.
RumbleBlocks benefited greatly from iteration. First
designs were communicated through paper prototypes, leading
to insights that guided subsequent work. For example, the first
towers were constructed of a playful mix of lollipops, candy
corn, and chocolate bars, chosen for a strong candy theme with
rich color and variety of shapes. This early idea, loved by the
artists, met with skepticism from educators concerned about
promoting bad eating habits. After a number of follow-up
sketches and ideas, a space theme featuring stranded aliens was
adopted, once additional tests with boys and girls found
acceptance across both genders with a cute, non-threatening,
nurturing art style.
The award-winning Unity 3D game engine [6] was chosen
for delivering RumbleBlocks, in part because it supported
deployment to a variety of platforms found in the schools and
homes of the target demographic, as well as having a built-in
physics engine useful for animating tower building activity. A
related early playtest focused on whether 4 to 5-year-olds were
comfortable with 3D blocks manipulation (in three-
dimensional space) using a mouse as an input device. Many of
the art assets in RumbleBlocks are three-dimensional (the
spaceship, alien, blocks) for physics engine consideration, but
children had difficulty in maneuvering 3D objects, doing much
better in manipulating the objects when constrained to a single
perspective view along the z axis. While the blocks are boxes
and cubes, in such a view they appear as rectangles and
squares. Figure 1 shows the perspective view as seen by the
child players (left), with the assets constructing the view shown
at right in a top-down view in the Unity development
environment illustrating their three-dimensional nature.
Figure 1. RumbleBlocks screen shot (left), a single perspective along the z-
axis built with a mix of planar and 3D assets as shown in Unity Scene View
from a high angle looking down (right).
Additional playtest iterations refined the tower-building
levels, but the HCII team members expressed concern over the
evaluation of a sandbox exercise like that illustrated in Fig. 1:
when the player places a block, is it a good or bad move
showing understanding of the principles of Section 2?
Specifically, can game mechanics be introduced that offer
discrete moves, where each move can be analyzed for
correctness? Discussion on such measurement of player
activity led to two additional activities being introduced into
RumbleBlocks: tower block removal (Fig. 2 left), and
contrasting cases (Fig. 2 right). In tower removal, the player's
mouse cursor is a sledgehammer that is clicked over blocks to
remove them (or finger-tapped, on touch devices), with the
goal of removing a set number of blocks without disturbing the
flying saucer (i.e., the "unidentified flying object" or UFO for
short; most often called "spaceship" by the players). In
contrasting cases, the player selects the more stable tower, with
a subsequent earthquake knocking down the less stable tower
after the player's selection (the UFO then lands on the more
stable structure).
Figure 2. Screen shots of Tower Removal (left) and Contrasting Cases.
Formative playtesting occurred throughout the game design
process. RumbleBlocks was designed toward achieving an
ideal "flow" [7], with a balanced challenge level to let the
student enjoy a rewarding enough experience to remain
engaged and feel a sense of achievement without undue

frustration or resignation. The development team knew that
proper level sequencing would take time, so rather than hard-
code a particular sequence of levels, the level progression was
set through an xml configuration file. Each level's art assets
could be laid out and tested by the development team using
Unity's development environment and Scene and Game views.
Children could test different sequences of the game by
varying the configuration file. Through frequent playtesting,
RumbleBlocks has evolved, with the goal of driving children
toward achieving high levels of understanding of the science
principles of Section 2, through appropriate challenges in
advancing levels. Such iterative playtesting is a staple of ETC
project studio development processes [5]. Specific "lenses" for
game design are useful to focus playtesting from the palette
presented by Schell [8]. For example, the Lens of Flow
addresses the challenge progression in levels and a play test
might focus on flow. The Lens of Surprise addresses appealing
surprises in the game, and a play test might focus on reactions
to embedded surprises like the alien having a protective shield
with audio and animation effects when the player tries to hit it
with a block.
IV. RUMBLEBLOCKS STORY NARRATIVE AND GAME
MECHANICS
First versions of RumbleBlocks lacked a narrative
framework: children were prompted to build towers up to a star
top block. ETC faculty encouraged the students to incorporate
a story premise that would help give young children a concrete
explanation of the goal and motivation to move through the
game successfully. Over many weeks the story developed with
a series of play tests with groups of 1-7 children, who noted in
actions and words which story elements worked and which
were still confusing. The last of the Illuminate project's play
tests with 11 children (7 boys and 4 girls, ages 6-8) confirmed
that the implemented story was presented clearly, understood,
and helped drive the player to success. Comments like "I want
to help the alien" peppered the videotaped interviews.
The Unity game includes opening and ending victory
videos, illustrated in part by the storyboards in Fig. 3. The
player is introduced to a mother ship in space which is hit by a
comet, with a number of UFOs then evacuating the damaged
ship for a variety of planets. The different planets serve as
backdrops for the levels the players work through: an ice
world, a volcano world, etc. In each planet the UFO crashes,
damaging the ship but depositing the alien safely on a ledge.
An energy tower must be built to raise the ship to a level where
the alien can be rescued from the ledge, and the ship is
energized if the tower's blocks are placed properly over the
blue energy balls. While the tower energizes, an earthquake
shakes the terrain, knocking down unstable towers but allowing
good structures to save the alien.
The energy balls (3 are shown in Fig. 1, with two properly
covered by blocks) are a means of giving the players
scaffolding in the task. If the level is teaching the wider base
concept with a high degree of scaffolding, the energy balls will
give strong cueing that a pyramidal form with wide base and
narrowing top is the solution. Levels with less scaffolding will
show fewer energy balls and allow for more degrees of
freedom in block placement. To complete a level, the player
moves and rotates the crashed UFO to the top of the built
tower. The tower energizes the UFO, and it flies away with the
alien and a cheer on success. The player then is presented with
the next level. On successful completion of all levels covering
the principles of Section 2, the player sees a video of the UFOs
flying back out to a rescue ship, where they then are presented
together on a congratulations screen (right panel of Figure 3).
Figure 3. Storyboard panes for left Intro sequence (mother ship damaged;
UFOs crash land; friendly aliens need help) and right Victory sequence (UFOs
return to rescue ship; aliens all dance happily and wave following player
helping them all back to their UFOs).
V. ADDITIONAL LESSONS LEARNED FROM PLAYTESTING
The target age range is broad, and more formal tests may
eventually narrow the optimal target to say first-graders (ages
5-6). Younger children found levels challenging, and tower-
stability principles confusing if too many were presented at
once. Careful design of levels introducing complexity over
time, with scaffolding provided through the energy balls,
helped the players. Experienced older players worked through
the easy levels quickly without complaint as they focused on
the narrative of helping the alien. Younger players made use of
the energy balls to guide block placement.
The game designers paid careful attention to the Lens of
Pleasure and Lens of Juiciness from Schell [8], rewarding the
player's actions in many ways at once with audio and visual
cues. Through tests with children, decisions were made on
how to increase the pleasure and juiciness of the game. The 48
levels of the game are distributed across four different planet
worlds, each with its own alien to be rescued, audio
background track, and visual style. The blocks were picked up
from an inventory shelf (via mouse or touch interactions) and
made to interact with the alien world, rather than float in front
of it. The block being moved by the player can crash into other
blocks, turn the spaceship, or even interact with the alien's
force shield, accompanied with playful sound and visual
effects. The alien fidgets and babbles on the ledge for the
player's amusement. The earthquake challenges the player's
tower with audio rumbles and shakes that produce shifting and
clanging of blocks that move according to Unity's physics
engine. The energy balls light up and via lightning effects
energize the UFO on successful survival of the earthquake test,
as shown in Fig. 4. Children's reactions to various particle
effects, from level completion to UFO energizing, as well as to
the above-mentioned effects, were monitored in numerous
playtest sessions. Their smiles, focus on the screen
interactions, immediate discussion with other child players, and
follow-up interview comments confirmed their pleasure with

the interface modifications and underscored the importance of
juicy interfaces in science games for young children.
Figure 4. Zoomed-in view of the UFO becoming energized on successful
construction of a tower that survives the test earthquake; a follow-up
animation shows the alien and UFO fly off before proceeding to the next
level.
There was concern about whether the children would
understand how to move the blocks, using the mouse on a
personal computer, or a touch screen on phones or tablets.
Visual and audio cues were added and tested to provide
interactive feedback. A tutorial heavy in symbols was added to
the very first level that a child performed. If he or she repeated
the action asked for in the tutorial, it would move on; if not, it
repeated the instruction with more detail. One playtest
examined the utility of the tutorial, and found that even without
voice-over accompaniment (in English) describing the actions,
the illustrated action storyboard was good enough to
communicate how to move and rotate blocks and the UFO to
help the alien.
Children are familiar with real-world blocks and their
interactions. An early playtest with a few children examined
whether a picked up block should float over and through all
other objects until it is released into the world (and presumably
made a part of a tower). An alternate system was tested
whereby a selected block moved with the mouse but with
physics and colliders working on it so that if it banged other
blocks, they would move, if it banged the cliff it would stop, if
it banged the UFO it might rotate, all under realistic control of
the Unity physics engine. This sandbox of block actions that
behave like real blocks was found to be much more intuitive
and appealing.
Children did not react negatively to the mix of tower
building, tower block removal, and contrasting case judgment
levels in the game. They did notice a disconnect in that early
versions of contrasting cases had single block structures, rather
than towers constructed of pieces as shown in Fig. 2. For this
age group, a strong narrative and consistency were found to be
important.
For consistency, the Sci-Fri team made the changes to
contrasting cases interface as shown in Fig. 2, i.e., replacing
monolithic single block towers with towers constructed from
multiple blocks. The team then tested the work to look toward
educational effectiveness in the spring of 2012. A test with six
6-7 year olds began with six contrasting cases, then a mix of
tower construction and tower block removal levels (up to 33
levels, up to 30 minutes of play time), then once either the time
limit or all levels were reached a final set of six different tower-
pair contrasting cases. The results are shown in Fig. 5.
Students scored 42% on the pre-test and 67% on the post-test.
This difference suggests that the game produced substantial
learning. We believe it to be unlikely, but we cannot exclude
the possibility that differences in the pre- and post-test forms
produced or contributed to this difference. Hence, we will
follow up with more formal testing with many more subjects.
Specifically, we acknowledge that the results from Fig. 5
are from a very small sample set. Evaluation plans include
testing with students ages 4-9 to determine if there are
differences across these ages with respect to the engineering
principles being taught in the tower building, tower
deconstruction, and tower contrasting levels. There will be
pre-tests and post-tests given bracketing the game as was done
here, to establish whether the game promotes learning outside
of the game itself. Of course, the game will be instrumented as
well with robust logging and in-game assessments that will
document the achievements that take place within the game.
Figure 5. Results from six children on educational assessment of
RumbleBlocks, showing 95% confidence bars and accuracy scores on
contrasting cases pre- and post-tests.
VI. CONCLUSION AND FUTURE WORK
The results of Fig. 5 suggest that thirty minutes with a game
may change understanding of scientific principles regarding
tower stability as measured by contrasting cases. (It is
important to note though that results from such a small number
of students do not always generalize to large segments of the
student population. Stronger evidence will come from larger-
scale studies on which we have embarked.) Obviously, the
work is early in the formative stages. Will these results hold
for thousands of students? What is the role of the tower
construction activity found to be so appealing by the students
that they will stay with the game for a full 45 minutes? The
role of tower piece removal? The influence of a varying
number of contrasting cases interspersed with the other types of
levels? The role of showing center of mass perhaps visually on
the screen as a point that changes with each block placement?
The role of showing which tower falls due to being less stable
during an earthquake in the contrasting case pairings? These

and other questions will be considered by the HCII learning
researchers using classrooms of tens of students and making
use of DataShop logging that has served well for intelligent
tutor evaluations [9]. Eventually, a number of varying
configurations of RumbleBlocks will be deployed widely
through the web to fine-tune level choices, much like the game
Refraction has tested play time, progress, and return rate across
varying versions of their game [10]. Such broad deployment
across the web is facilitated by Unity's web player. Testing
various level compositions and sequencing is facilitated by
game logic that makes use of configuration files.
This paper has emphasized the early design decisions,
prototypes, and quick play tests with small sets of children
which led to the development of RumbleBlocks. From the
choice of art style to the inclusion of a narrative, from the need
for symbolic communication of a tutorial to target 4-7 year-
olds (who may not yet know how to read) to tweaking game
elements of fun and surprise to keep the players engaged, the
paper has overviewed the improvement of the game over time.
The interested reader is welcome to see more background on
the reported work and play RumbleBlocks via links from the
ETC [3]. RumbleBlocks appears to have measurable
educational effectiveness for children and is a fun game.
Future work will scale the evidence, field test more broadly,
and report modifications made based on such testing.
ACKNOWLEDGMENT
This work has benefitted from the work of ETC students
John Balash, Chandana Bhargava, and Weiwei Huo, who are
also Sci-Fri team members. RumbleBlocks is supported by
DARPA's ENGAGE program.
REFERENCES
[1] M. Fleer, "Fusing the Boundaries Between Home and Child Care to
Support Children's Scientific Learning," Research in Education, vol.
26(2), pp. 143-154, 1996.
[2] A.L. Dombro, J. Jablon, and C. Stetson, Powerful Interactions: How to
Connect with Children to Extend their Learning. Washington, D.C.:
National Association for the Education of Young Children, 2011.
[3] Entertainment Technology Center ENGAGE Projects Illuminate and
Sci-Fri, Carnegie Mellon University, http://www.etc.cmu.edu/engage,
2012.
[4] D.L. Schwartz, C.C. Chase, M.A. Oppezzo, and D.B. Chin, "Practicing
versus Inventing with Contrasting Cases: The Effects of Telling First on
Learning and Transfer," Journal of Educational Psychology, vol. 103(4),
pp. 759-775, 2011.
[5] R. Pausch and D. Marinelli, "Carnegie Mellon's Entertainment
Technology Center: Combining the Left and Right Brain,"
Communications of the ACM, vol. 50(7), pp. 51-57, 2007.
[6] M. Totty, "The Winners, Category by Category (Most Innovative
Technologies),", The Wall Street Journal, Sept. 27, 2010; see also
http://unity3d.com/.
[7] M. Csikszentmihalyi, Flow: The Psychology of Optimal Experience.
New York: Harper and Row, 1990.
[8] J. Schell, The Art of Game Design: A Book of Lenses. Burlington, MA:
Morgan Kaufmann, 2008.
[9] Pittsburgh Science of Learning Center: DataShop: a Data Repository and
Analysis Service for the Learning Science Community,
http://www.learnlab.org/technologies/datashop/, 2012.
[10] E. Andersen, Y.-E. Liu, R. Snider, R. Szeto, S. Cooper, and Z. Popović,
On the Harmfulness of Secondary Game Objectives, Proceedings of the
Conf. on Foundations of Digital Games (2011), pp. 30-37.