The Simulation Cycle: combining games, simulations

E–Learning Volume 6 Number 1 2009 www.wwwords.co.uk/ELEA

The Simulation Cycle: combining games, simulations, engineering and science using StarLogo TNG ERIC KLOPFER, HAL SCHEINTAUB, WENDY HUANG, DANIEL WENDEL & RICAROSE ROQUE Massachusetts Institute of Technology, USA

ABSTRACT StarLogo The Next Generation (TNG) enables secondary school students and teachers to model decentralized systems through agent-based programming. TNG’s inclusion of a threedimensional graphical environment provides the capacity to create games and simulation models with a first-person perspective. The authors theorize that student learning of complex systems and simulations can be motivated and improved by transforming simulation models of complex systems phenomena (specifically this study examines systems including epidemics and Newtonian motion) into games. Through this transformation students interact with the model in new ways and increase their learning of both specific content knowledge and general processes such as inquiry, problem solving and creative thinking. During this study several methods for connecting the simulations to game dynamics were tried, ranging from student-created games, to altering existing games, to students playing premade games. This article presents the results of research data from two years of curriculum development and piloting in northern Massachusetts science classrooms to demonstrate the successes and challenges of integrating simulations and games. This article also explores the results of these interventions in terms of ease of implementation, student motivation and student learning.

Two teams are building models of virtual worlds. They each need to consider the relevant aspects of the world that they want to represent, focusing on what is important for their purposes, and what is superfluous. They also need to consider how they will provide appropriate inputs into their system and understand the output of their models, including whether the feedback that the models provide is clear. Each team needs to cleverly devise algorithms that appropriately represent the actions and behaviors of the inhabitants of their virtual world, and investigate the outcomes that they observe. In many ways the actions of these two teams are indistinguishable. However, as the products progress, the differences become more pronounced – one team is developing and studying a simulation of warming seas designed to help scientists save endangered species; the other is building a jet ski racing game designed to entertain. Both of these products require good initial models of fluid dynamics, tide flow, buoyancy, and many other physical parameters as a starting place. They may both incorporate information about how weather impacts the oceans – either to make the simulation more accurate or to make the game more exciting. The simulation requires important biological parameters to describe the ocean inhabitants, whereas the game requires important physical information to simulate the behavior of the jet ski under different ocean conditions. Of course there are distinct differences between the way the game and the simulation are developed and studied. These differences allow the simulation to be more predictive, and the game to be more engaging. But perhaps they are more similar than distinct. It is this connection between the design and building of games and simulations that motivates the research and development of our Simulations, Systems and Computational Literacy (SSCL) curriculum and the corresponding tool, StarLogo The Next Generation (TNG). Simulations are an 71

http://dx.doi.org/10.2304/elea.2009.6.1.71

Downloaded from ldm.sagepub.com by guest on August 21, 2015

Eric Klopfer et al increasingly important part of science and engineering. They have radically altered the scientific landscape, making possible a wide range of breakthroughs in basic and applied science. Simulations allow scientists to test and make predictions of nanoscale phenomena that happen in an instant as well as global processes that may take centuries to unfold. For scientists, simulations are not something that come packaged and shrink-wrapped delivered to their desktops but rather they are expressions of an individual scientist’s or a team’s hypotheses that are developed and tested through simulation. It is through an iterative process of design, implementation and testing that scientists work through their ideas. As much as we strive for school science to reflect the tools and practices of real science, serious incorporation of simulations in the school curriculum is rare. When simulations are incorporated into the curriculum it is often through pre-made simulations such as threedimensional (3D) animations or representations for visualizing phenomena normally difficult to conceptualize from formulas or abstract explanations alone – vector fields in electromagnetism (Belcher et al, 1999), particle behavior (Moodley, 2004), molecular dynamics (Pallant & Tinker, 2004), genetics (Buckley et al, 2004), or osmosis (Meir et al, 2005). Despite the use of environments that lower the entry barrier for simulation development, most work has focused on student use of simulations rather than development. A few exceptions are worth mention: the AgentSheets simulation (Sherell et al, 2005) for high school students in beginning programming classes allows students to experiment with input values and visualize the results instantly, enabling exploration and the discovery of patterns and formulas of combinatorics on their own. StarLogo T (Wilensky, 1997) and its successor, NetLogo (Wilensky, 1999), have been employed for modeling in many physical, chemical and biological processes, including GasLab (Wilensky, 2003), a series of gas law simulations designed to teach about gas dynamics at the atomic level, and Connected Chemistry (Stieff & Wilensky, 2003), designed to help students understand the connection between microscale processes and macro-scale observable phenomena in chemistry. Although these do not emphasize simulation development per se, they do provide access to the underlying code which could arguably provide students with insights to their development.

Figure 1. The simulation cycle, showing an intersection of science (bottom) and engineering (top). Scientists and students can enter these two linked cycles in multiple places for a particular purpose and can move throughout the cycles over time as the situation calls for.

While pre-built simulations can provide students with accessible visualizations, immersive learning environments, and the opportunity to analyze data from virtual experiments, they cannot provide them with the freedom to express and explore their own understanding of a given phenomenon through simulation use, nor can they provide much insight into the use of simulations in the scientific enterprise generally. In order to do that, students must be able to experience the full 72


The Simulation Cycle spectrum of interactions with simulations which combine science and engineering, including: studying and analyzing existing simulations, understanding and redesigning simulations, and even building simulations themselves (see Figure 1). No single investigation needs to span the whole spectrum from design to building to data collection, but multiple experiences with different parts of this process contribute to an understanding of the role of simulations broadly. Limited work has been done on students actually developing simulations: Wilensky & Reisman (2006) conducted detailed studies of students building simulations using StarLogo T and found marked improvements in student understanding using detailed observational studies. Such work is difficult to scale, however, requiring substantial in-class support from researchers to enable student and teacher programming. Similarly, difficulty in scaling has been found with StarLogo (Scheintaub et al, in press) as well. One study (Kuch, 2007) used Arena, a professional simulation tool used in manufacturing, healthcare, and other industries, to integrate several parts of the simulation process in a high school course specifically about simulation. After examining pre-built simulations, students in the course designed and built simulations to investigate a problem of their own choice. One group of students chose to optimize the food queues in their school cafeteria by collecting data during lunchtime and building a simulation to determine what modifications and scenarios optimized traffic flow in the cafeteria. By providing them with a tool to create their own simulations and the skills to design models, students were able to augment their intuitions to develop testable hypotheses and scientific models of social and physical processes. There have been a number of barriers to incorporating the redesign and creation of simulations in classrooms, despite the potential of integrating simulation development (particularly of complex systems; see Jacobson & Wilensky, 2006) into science classes. Such barriers include the necessary scaffolding and motivation for both students and teachers and recognition of the value of programming and design processes for learning. Games, on the other hand, are inherently interesting to students. Thus, programming in the context of game design and construction could provide the motivation necessary for students and teachers to experience the power of algorithmic and design thinking. Such experiences could then become a stepping stone toward the use of programming environments as curricular learning tools. Without proper preparation, students see simulations as abstract. In a time when much of school activity is seen as abstract and unrelated to their lives, working with schematic representations of simulated systems (typically presented as two-dimensional ‘top-down’ schematic representations) can further disengage students. This is particularly true of students who have come to expect high fidelity 'after growing up interacting with nearly true-to-life 3D game environments. Moreover, most teachers are not adequately prepared to work with simulations and bring them to life for students. Developing the expertise necessary to engage students in a more rigorous curriculum of science simulations is something that most teachers do not have time and support for. In addition, most tools created for scientists are too complex for the classroom and would take significant mastery to use. It is just such challenges that motivated the development of StarLogo TNG. StarLogo TNG builds on the tradition of Logo-based languages designed to facilitate the development and study of simulated systems in classrooms. This latest version provides several key advances, including a graphical programming language and a game-inspired three-dimensional world. Its graphical programming language structures programming tasks through the use of programming blocks (Begel, 1996), relieving the major burden of syntax that scares many teachers away from previous StarLogo generations that use a traditional text-based language for programming. While the commands were relatively simple in earlier versions, they still required absolute precision in wording, arguments, and punctuation. For example, think about a task where your goal is to move pieces around on a chessboard. If you wanted a creature to move forward one space if there were a red space in front of it and move back one space if there were not a red in front, you would have to use the following code exactly: to move ifelse pc-ahead = red [forward 1] [back 1] end 73


Eric Klopfer et al While seemingly simple code at first, there are numerous ways to break it. If the spaces on either side of the equals sign are left out, the program will complain. If ‘ifelse’ is spelled with a space or a dash, it will complain. If parentheses or curly brackets are used instead of square brackets, it will complain. If pc-ahead is ahead-pc or any other variant, it will complain. If the second part of the ifelse command (the [back 1]) is left out, it will complain. Requiring such precision places a large cognitive load on both teachers and students, and they have to struggle to maintain focus on the systems and simulations principles rather than the programming language, which is intended as merely a means to an end. Programming blocks help to relieve such syntax-related problems. In StarLogo TNG, all of the commands are available to the programmer in an organized palette (see Figure 2). Using a painting metaphor, the commands stored in palettes are moved (‘painted’) onto the programming canvas.

Figure 2. The StarLogo TNG programming interface consisting of a palette of commands that can be painted onto the programming canvas.

Each command is placed on appropriately shaped blocks which are shaped such that they only fit together with other commands with which they work. For example, the above code from the original StarLogo would be represented in StarLogo TNG as the following:

Figure 3. The StarLogo TNG programming blocks for the above example where the goal is to move a creature forward one space if there is a red space in front of it and backward one space if there is not.

74


The Simulation Cycle Note that there is no longer any punctuation and it is now very difficult to leave out any necessary commands. The blocks show where commands need to be entered and suggest what kind is required, making it possible to focus on what the commands should be doing rather than syntax. Such change has made a dramatic difference: while few (if any) science teachers we worked with previously felt prepared to introduce the simulation development activities into their classes with previous versions of StarLogo, most now express that they feel comfortable with it. The second significant advance in StarLogo TNG is the rich three-dimensional world called Spaceland that inherits many of its characteristics from video games. It has several 3D views, affording multiple perspectives of the world. By default Spaceland is in third-person aerial view (Figure 4), where users can move their camera around the world. Alternatively, a world view gives a top-down perspective, while a first-person view (over the shoulder of one’s individual creature) lets the user focus in on a single agent (see Figure 5).

Figure 4. The aerial view of Spaceland in StarLogo TNG provides a systems-level view of the world.

StarLogo TNG also includes controls common to most simulations such as start and stop buttons, sliders that control variables, and data output mechanisms such as tables, graphs, and instant readout monitors. These are combined with game-like controls including keyboard and joystick inputs and onscreen readouts of scores or other parameters. The combination of traditional simulation outputs and representations with recognizable game aesthetics and inputs provides a rich environment that appeals to both students and teachers. It motivates them to build models of systems yet grounds their experience in the more familiar realm of games, demanding movement between local perspectives (for example, as a fire fighter moving through a burning forest to clear trees) and global perspectives (for example, as they measure the impact of tree clearing on the fire overall). We have begun development and testing of curricula across a number of disciplines around this principle, including biology, physics, middle school science, and middle school mathematics. In this article, we provide two case studies. The first case describes the use of StarLogo TNG in a middle school science curriculum; the second, in a high school physics curriculum.

75


Eric Klopfer et al

Figure 5. A first-person perspective in StarLogo TNG shows the world from the perspective of an individual agent.

Middle School Science Curriculum Design Case Study The design and pilot of the StarLogo TNG middle school curriculum follows two paradigms of integrating games and simulations. These are Simulation = Game, where students explore a simulation model by playing a pre-built game based on it, and Simulation Æ Game, where students explore a simulation model by using scientific inquiry processes (such as modifying parts of the model) and then applying the knowledge gained to solve game-like challenges based on that model. For each paradigm, we detail activities that illustrate the curriculum design process, present results of pilot testing, and discuss the implications for curriculum development and teaching practice. Simulation = Game An example of this paradigm is a game we call the ‘Planets Game’ which focuses on gravity, a topic selected from an eighth-grade science teacher’s curriculum after collaborative discussion on the feasibility of several topics. Following Wiggins & McTighe’s (2001) ‘backward design’ approach, we started with the learning goals: what are the important things that we want students to learn? We then turned to the Massachusetts science framework and located the relevant standard: differentiate between weight and mass, recognizing that weight is the amount of gravitational pull on an object. We designed an assessment to administer before and after the activity, inspired by Keeley et al’s (2005) Uncovering Student Ideas in Science series, a collection of child-friendly science probes that help to reveal common misconceptions. The moon’s gravity is less than the earth’s. If apples cost $0.99 per pound, which do you think will be the better deal: a. A bag of apples that weighs 5 pounds on earth b. A bag of apples that weighs 5 pounds on the moon c. They’re both equally good deals Explain the reason for your choice.

Because gravity is less on the moon than on earth, the mass of five pounds of apples on the moon must be greater than the mass of five pounds of apples on the earth, and thus the five-pound bag of apples on the moon would be a better deal than the five-pound bag of apples on the earth. Our 76


The Simulation Cycle assessment was designed to measure whether our planets game helped students improve their understanding of these relationships. After much brainstorming, we devised the following game narrative: You are an ant. You live with your friends in a house with a happy human family (‘the Simpsons’) and their pet fish, ‘Nemo.’ It’s a wonderful arrangement because the family is messy and leaves crumbs around, so the ants have plenty of food to eat. At the same time, the ants help keep the house clean, so the humans are happy. However, there’s a problem: an evil villain, just to be mean, kidnaps the family and the fish! This villain is both very smart and very dumb. He’s smart enough that he can fly the mom, dad, and fish each to separate planets! He also rigged up special jail cells with scales, and made the doors only open for his exact weight, so he would be the only person who could let the prisoners out. However, he is also really dumb; he set up the jail cells on Earth, and forgot to take into account the fact that different planets have different gravities, so his weight is different for each planet. Now the family and fish have a real problem, because they’re stuck on different planets, and not even the villain can open his own jail cells to let them out! That’s where you, the ant hero, come in. You need to search the different planets for the jail cells and free the prisoners! Luckily you have a lot of ant friends you can invite to stand on the jail scales with you to try to reach the right weight to open the doors. Remember, each jail is set to open with the same weight (the villain’s weight), but because gravity is different it might take different numbers of ants to weigh that much on different planets.

Through game-play (which is really a modified version of the scientific method in the bottom half of Figure 1, where students Play, Observe, and Generate Strategies instead of Test, Collect Data, and Generate Questions), students realize that the same weight opens the door to the jail cell on every planet, but because each planet has a different gravity, a different mass is needed to open the door on each planet. To exploit every learning opportunity conceivable, we designed the game so that the game protagonist can jump on each planet, making it not only fun to move around but also easier to get an intuitive sense of the gravity on each planet relative to the others. To help students plan their strategy, we also included an activity sheet with a table for tracking their guesses of mass and the resulting weight on each planet and reflective questions to guide their conceptual development. Our pilot took place at a K-8 charter school in a low-income community and was conducted with the entire eighth grade (47 students). We ran the one-hour activity with four groups of students during a school day, with students working in pairs throughout the activity. Each session started with a short survey and pre-activity assessment. Students then viewed a short video of an astronaut skipping across the surface of the moon and shared observations and hypotheses about it. The students in the study had already studied gravity for the past two weeks so they were able to make statements like, ‘the gravity is less on the moon so the astronaut weighs less on the moon’. After this articulation, we then introduced the game narrative and characters. Students were immediately engaged with the activity. They enjoyed controlling an ant in the first-person perspective and walking around to find the jail cells. Although they struggled a bit with the keyboard controls (which were awkward due to the way we implemented the underlying movement model), they seemed to enjoy the challenge and loved jumping on different planets to see how high their character went. Assessment results. Approximately half of the student pairs were able to open every jail cell, and every pair recorded trials on at least two planets. Only the pairs that opened all of the jail cells responded to the written questions on the worksheet. Of these responses, many showed a good understanding of the system. For example, when asked why an ant weighs more on one planet than on earth, one pair replied: ‘It mean[s] that the gravity on that planet is higher than the one of earth’. In response to the next question, which asked if a mass can have different weights on different planets, another wrote: ‘Yes because it’s different gravitational forces’. Many of the students began to see the relationship between gravity and weight and realized that they could divide the target weight by the weight of one ant to find the mass needed to reach the target weight in the current gravity. As luck would have it, the day of our pilot was also the day that the mathematics teacher was beginning a unit on solving problems using proportions. Later, 77


Eric Klopfer et al the science teacher told us that when the mathematics teacher was explaining proportions to the class, one student said that he could have used this method to help him solve the gravity game puzzles. Five weeks after the activity, we gave the students the same assessment question that we had given them before the activity, resulting in 45 matched pre-test/post-test assessments. Of the 45 pre/post pairs, 25 correctly stated both before and after the activity that five pounds of apples on the moon is more apples than five pounds of apples on the earth and hence a better deal. Of those 25, however, four incorrectly stated that apples would be weightless on the moon before the activity and two of those four retained this misconception even after the activity. Fourteen assessments were incorrect both before and after the activity, although the rationales for their answers changed at post-test, and six students chose a different answer from pre-test to post-test: four corrected misconceptions and two chose incorrect answers. Table I shows the rationales students provided for their answers. Consistently incorrect responses 1. Same weight in both places means everything is equal. 2. Different price per pound on the moon. 3.Nonsensical (e.g. ‘It would be good to have apples on Earth because they won’t float away’)

16 total 6 pre/6 post 5 pre/7 post 5 pre/3 post

Responses that became correct 1. Mass is less on the moon. 2. Earth-pounds are heavier than Moon-pounds. 3. Same weight also means same mass.

4 total 2 pre 1 pre 1 pre

Responses that became incorrect 1. Apples weigh less, which means more apples, which means more money.

2 total 2 post

Table I. The rationales students provided for their answers, whether consistently incorrect, initially incorrect and then fixed, or initially correct and then messed up.

While a one-hour activity cannot hope to correct every misconception, we did find measurable improvement. From the results of the pre- and post-tests it appears that several students were able to overcome misconceptions and gain a better understanding of the relationship between mass, gravity, and weight from this activity. Our pilot also suggests, however, that we might want to rephrase the assessment question itself. Without intending to, we introduced an added level of complexity by asking students to interpret what was meant by a ‘better deal’. Additionally, it takes a certain amount of mathematical reasoning to determine that, given a fixed per-pound price, individual apples are cheaper if they are lighter. Perhaps, then, a better way of assessing students’ understanding of gravity, mass, and weight is to ask whether a five-pound bag of apples on the earth has the same amount of apples as a five-pound bag of apples on the moon and, if not, why. Implications. The ‘simulation-as-game’ paradigm’s advantage in this work is that integrated StarLogo TNG simulation/game projects such as those discussed here can be easily added to relevant parts of a standard school science curriculum to provide students with an engaging way to experience a simulation model. The disadvantage is that such integrated projects are very labor-intensive to develop. It may not always be possible to design truly integrated projects that are both fun and educational. Moreover, this paradigm tends toward the science half of the simulation cycle (bottom of Figure 1 – adapted for play) and does not naturally invite movement into the engineering half (top of Figure 1). In ‘Planets Game’, the model is more or less a black box because the code is too complex for beginner programmers to understand or modify. From Simulation to Game The ‘from-simulation-to-game’ paradigm traverses three loops – Scientific Method, Engineering Design, and finally, Play, as shown in Figure 6 (a modified version of Figure 1 that includes an 78


The Simulation Cycle additional loop), moving back and forth between the scientific method using simulation models, a gamer’s version of the scientific method applied to simulation-based games, and designing/building both the simulation and game models. Such movement also means much more classroom time compared to the ‘simulation-as-game’ paradigm. We will discuss the design and classroom pilot of a unit containing four activities on forest fires. We began with research on variables that affect the spread of a forest fire. We wrote these in the form of ‘enduring understandings’ and ‘essential questions’ (Wiggins & McTighe, 2001) to clarify the learning goals for our own understanding and to guide the development of the activities. We also compiled relevant standards from the Massachusetts science frameworks (Massachusetts Department of Education, 2006), American Association for the Advancement of Science standards (American Association for the Advancement of Science, 1993), and the National Research Council standards (National Research Council, 1996). For the assessment, we wanted students to demonstrate their understanding of the usefulness and essential features of simulations by comparing them to other computer-aided visualization tools such as animated flash movies. We showed students an animation about the beneficial role of fires on an ecosystem and a StarLogo TNG simulation model containing variables like wind that can be adjusted. Afterwards, we asked students to answer questions comparing the two, expecting that their answers prior to the workshops would focus on superficial features (such as differences in graphics) and that, following the intervention, their answers would show more nuance (such as noticing that the animation is always the same each run but the fire starts in a different place each time in the simulation).

Figure 6. The simulation cycle shows three loops. The activity starts with testing and tinkering and then goes through a loop of the scientific method before crossing into a design phase. The design phase crosses back through play into a phase of scientific method of play.

The first activity was designed around an existing StarLogo TNG forest fire model that has two variables: wind and tree density. After an introduction to the topic (a clip from Bambi and a brief discussion of students’ prior knowledge about forest fires), we introduce the simulation model by asking students to predict which would result in a more destructive fire: a less dense forest or a more dense forest. We ask them to find out by tweaking the variables in the model and running trials, keeping records of their experiments. These records are then analyzed and used as the basis for their report to ‘decision-makers’ such as government officials about what conditions lead to a destructive fire. In the second half of the activity, students program a third factor (topography) into the model by following a step-by-step tutorial and observe how the new factor influences the behavior of the system. Within three hours of class time, students traverse the entire simulation cycle shown in the first and second loops of Figure 6: running a simulation model, observing its 79


Eric Klopfer et al behavior and collecting data, generating ideas about forest fires, and introducing a new variable into the system using the design–build–test cycle. The next two activities are similar in format and movement through the cycle. Both start with the scientific inquiry loop, enter the game-play loop, and touch briefly on the design/build loops (see Figure 6). The firescaping activity and the prescribed burn activity address two important forest fire prevention techniques. The first activity is designed specifically to show how firebreaks are useful to stop a fire from spreading by eliminating fuel in a particular area, and the latter activity shows how small fires can be useful to clear sick/dead plant matter in order to prevent larger fires. Both activities have two parts – the first part consists of scientific experimentation to find out something useful about forest fires that can then inform the strategies to do well on a game-like design challenge in the second part. In the firescaping activity, students experiment with different firebreak widths to find the smallest cutting radius that can successfully keep fire from ‘jumping the gap’. In the process, students realize that they must consider probabilities to account for the randomness in the model. A certain radius may work once but not the second or third time. When is a particular radius value ‘good enough’ to be used? Thus, students must analyze risk and decide how many successful trials are needed to establish confidence. Students turn the simulation into a game by programming keyboard controls for a forest manager agent who can respond to key presses to move around the forest, build houses, chop and replant trees, and start a fire. The object of the game’s first level is to ‘design’ a forest that contains 400 trees and one house that can survive if a fire starts in a random location. Students’ score depends on how many houses and trees are left standing. Thus, students are encouraged to optimize their strategies using the score as feedback. Figure 6 shows the simulation cycle. In the firescaping activity students start in the first loop of Figure 6 as they conduct their scientific investigation, then move to the second loop where they make changes to the program, and finally enter the third loop as they play the resulting game. In the prescribed burn activity, students are introduced to a modified forest fire model where a new variable – moisture – has been added; trees turn brown over time to show that they got sick and died and thus lose all their moisture. Trees with low moisture are more flammable than tress with high moisture. Students are asked to observe how this model’s behavior differs from the previous ones and to find the highest percentage of sick trees that a forest can tolerate before a random fire will wipe out most of the forest. Students then modify the simulation to turn it into a game by adding a firehouse that can deploy firefighters that set small fires when they run into sick/dead trees. Small fires kill brown trees but not green trees with high moisture content. Here, the game is to send out firefighters to set small controlled burns that get rid of the sick/dead trees with key presses; the goal is to keep a forest healthy for 90 seconds, after which time a random fire starts. Each firefighter costs money so students need to think strategically about when to send out firefighters and how many are needed and to continually use feedback from the simulation monitors to inform their moment-by-moment decision making. The winning objective is to save at least 85% of the forest while having the lowest cost. Thus, similar to firescaping, students start on the lower half of the simulation level diagram (Figure 6), using scientific experimentation to learn useful information from the model about the tipping point percentage of sick/dead trees that can devastate a forest if a fire starts, and then applying those results to maintain a healthy forest. In the final ‘Save the Bunny’ activity, students control a bunny agent from a first-person perspective to run to safety within a burning forest, guiding as many animals as possible to safety along with them. Students then design difficulty levels for the game by adjusting the values of variables, including tree density, wind, percentage of sick trees, and seconds until the fire starts. Students must design an easy level and a hard level, requiring them to synthesize all the different aspects of forest fires that they have learned about in the previous three activities into a single model. Designing game levels functions as a performance assessment to see if students can express their understanding of how the variables affect the spread of a fire. After testing their own designs, students play each other’s game levels. Thus, the activity traverses the design–build–play loops (two and three) of the simulation cycle (Figure 6), though there is little actual building beyond changing the values of variables to achieve two difficulty levels. Implementation. The forest fire pilot activities took place over the summer in a sequence of four 1.5hour sessions as the enrichment component of a summer test preparation program, with students 80


The Simulation Cycle working in pairs. Attendance was optional and we had an average of 15 rising eighth-grade students per session with a core group of roughly 10 students participating regularly. Each pair of students included a ‘driver’ and a ‘navigator’; the student controlling the computer (the ‘driver’) was told what to do by the student holding the worksheet (the ‘navigator’) and students switched roles periodically. Forest fire activity 1. The first activity (discussion of a clip from the movie Bambi) focused on the causes of forest fires and their impact on people and animals. We asked the students to imagine that they were forest fire experts who had been approached by park rangers about managing a forest’s density to prevent devastating fires. It was also helpful to have students understand the concept of density in an intuitive way, as a measure of how ‘crowded’ something is, rather than as simply a formula (mass/volume). We asked the students to predict whether dense forests would be at higher or lower risk for devastating fire. Interestingly, the class was split in their predictions: About half believed that crowded trees would allow fire to spread more easily while the other half believed that the extra oxygen in a less dense forest would cause the fire to spread more. This split provided excellent motivation for the students to begin their experimentation. Overall, students were engaged in the activity and enjoyed setting fires in forests under different conditions to see the outcomes. However, we also found that the students had a difficult time designing their experiments. Rather than systematically changing one parameter at a time in order to focus on the effects of each change, they often changed several variable values at once and drew their conclusions based on limited and scattered samples. Thus, some groups lost entire forests with conservative parameters while others saw dangerously crowded forests saved by sheer luck. The resulting conversation at the end of the activity was therefore highly engaged. In the end, however, students were indeed convinced that, all other things being equal, denser forests were at higher risk for devastating fire. In retrospect, we decided that students needed more guidance in experimental design and data interpretation; we incorporated this lesson learned into the subsequent implementations of our activities. Forest fire activity 2 – firescaping. For the inquiry cycle of this activity, we provided slightly more structure to support students’ experiments, including a data table containing cells for relevant data and a set of experiment instructions to get them started. Additionally, we brought the group back together to share data after the initial round of experiments, which allowed them to see that multiple trials were necessary to have confidence in their results. Students were successful at using aggregate data to generate good game strategies. After comparing results of their firebreak radius experiments, they settled on a radius between two and three, depending on how much risk they were willing to tolerate. Students also modified the radius parameter and the tree-planting parameter as part of their strategizing, transferring what they learned in the first part of the activity. Their scores also improved over time as they continued to play, observe, and form new strategies. Moreover, they loved the game cycle of the activity. Exclamations such as ‘We did it!’ and ‘Awww’ were heard as the students either saved a forest or watched the fire jump their firebreak. The students were pumping their fists and raising their hands in excitement, and were constantly cheering and yelling for their fires to slow down and not spread to the other side. One student had to be reminded not to jump on his chair. Forest fire activity 3 – Prescribed burn. This second inquiry cycle provided excellent opportunities for discussion and learning. The data table on the activity sheet, which explicitly had space for recording the results of repeated trials, helped guide students’ experimental procedure better than previous activities. During the debrief, for example, one pair of students recommended a percentage value because it had worked four out of five times in their studies. When the rest of the students shared their data using the same percentage value, however, it became clear that the percentage value would fail more than half the time. Such generative discussions shed light on the importance of repeated trials and sharing data. Students were very enthusiastic about the activity and busily engaged each other in conversation, with challenges and collaborative comments such as ‘It’s gonna be worse than you think’, and ‘obviously it’s going to kill all of them’, and ‘TEN PERCENT? You’d feel comfortable THERE?!’ 81


Eric Klopfer et al The programming portion of the activity was less successful, however. Confusion about which file to open left most of the groups struggling to find blocks that did not exist, leading to frustration. Additionally, being asked to listen to a demonstration while having worksheets in front of them was a difficult challenge for those who were eager to race ahead. By the end of the demonstration, many had lost interest and were no longer paying attention, which only confounded their confusion about which file they needed to access. For the handful of students who were successful in it, the programming activity did seem to be enjoyable and rewarding, however. One pair decided to change the shape of their firefighter breed, opting to have them look like giant carrots going out to save the forest. The game portion of the activity, on the other hand, was an unmistakable success. Students competed to see who could spend the least money (i.e. send the fewest firefighters) and yet save the forest. While one group got very lucky and was able to send no firefighters at all (a loophole we intend to close in future iterations of the activity), most of the groups used strategies based on the results of their earlier data collection, sending firefighters out at a pace sufficient to keep the percentage of dead trees in the forest below the threshold of danger that they had agreed upon earlier. Figure 7 shows one example of what most of the final results looked like. Notice that the percentage of dead trees is below 5% and you can see firefighters standing in the forest where they have recently removed dead trees.

Figure 7. Screenshot showing a typical result from the ‘prescribed burn’ challenge. Students used firefighters to keep the percentage of dead trees below 5%.

Forest fire activity 4 – Save the bunny! The fourth and final activity was an encouraging close to the series. Not only did the students really enjoy playing the game but they were able to successfully integrate their knowledge from previous activities into their designs, modifying variables such as forest density, wind, and percentage of dead trees to create levels of varying degrees of difficulty. 82


The Simulation Cycle Because students spent time designing, testing, and redesigning their games, they made discoveries about what makes games fun. At first, many students set out to make levels as hard as possible. One student triumphantly announced, ‘No one can beat my game!’ Another was thrilled to have created a level that even the facilitators could not beat. However, over time they began to make levels that were difficult but not impossible. As one student said, ‘I like it better when I have a chance to win because I’ll keep trying’. Even more exciting than creating a level the facilitators couldn’t beat was beating a level the facilitators couldn’t beat. As the activity continued, the room was filled with laughter, screaming, and constant movement from one computer to another to see what others did. Video recordings of the activity show students giving each other high-fives, patting each other on the back, and jumping out of their seats when they beat a particularly difficult level. Significantly, many of the girls in the room, although less vocal than the boys, also seemed to enjoy the game. One group was excited to have the facilitators play their difficult level. Another experimented not only with the level design but also with character creation, changing the main character to several other creatures before changing it back to a rabbit because ‘it looked nicer’. Post-activity assessment. Much of what was learned throughout the course of the activities was demonstrated in the final creation of the game (i.e. the fourth activity), but we also administered a post-test to see how students’ understanding of simulations changed over the course of the activities. The majority of student responses continued to be about appearance and plot. However, a few post-test responses were unlike any on the pre-test and showed that some students had gained an important insight into the difference between a simulation and an animation –chiefly, that simulations allow the user to change system inputs and get different potential outcomes as a result. As one student responded: On the first demonstration everything was the same. It [the fire] killed every tree. On the second demonstration it showed different – like, the first time she showed us, it burned down four trees [and] on the second time, it burned down most of it.

Implications. The forest fire unit shows three ways of connecting the simulation and game loops (see Figure 6): The first is to use the scientific inquiry process to generate useful information (e.g. width of an effective firebreak) which can then inform game play. The second is to have students program parts of the simulation model to turn it into a game (e.g. adding a protagonist agent and first-person keyboard controls). The third is to have students apply what they learned about variables in a complex system by designing difficulty levels for a game that is based on the system. Although the Simulation Æ Game paradigm engages students and teachers in more inquiry and design cycles across both halves of the simulation/game diagram (Figure 6), compared to the Simulation = Game paradigm, it also takes more classroom time and requires greater facility from teachers in enabling constructivist, hands-on learning. There are several compelling reasons that teachers and administrators might choose to make such an investment. First, teachers can use simulation models to introduce students to new ways of doing science while at the same time continuing to teach traditional experimentation skills (collecting and organizing data, interpreting graphs, designing controlled experiments, using repeated trials and averages, and critically interpreting data and drawing conclusions). Second, new thinking processes can be taught with simulation/game models, such as handling randomness in data, thinking probabilistically, and understanding complex systems. Almost all activities that involve programming to modify or build models require extra time and programming expertise –conditions that are not necessarily common among middle school science classrooms. However, as the forest fire curriculum shows, there are ways to give students design experience that incur relatively low costs in terms of class time and teacher knowledge by offloading some of the burden onto the preparation of activity worksheets and models: 1. Designing game levels instead of programming entire games can save time but still give students the opportunity to modify a model in a significant way. 2. Easy areas to modify an existing game/simulation include: keyboard controls, movement procedure, choosing breed shapes, changing agent attributes like color and size.

83


Eric Klopfer et al 3. Students can program more complicated code if they are given step-by-step instructions such as adding a new breed with associated rules of behavior, or programming a new variable and actions to add to the behavior of an existing breed; they may not have as deep an understanding of the logical sense of the code but at least students learn that the model can be shaped by the user. In other words, modifying (‘modding’) parts of a game or simulation can engage students in some programming in order to understand how the models work, even when programming is not the focus of activity per se. As curriculum developers, we spent hours discussing the phenomena that we were modeling and going through the design–build–test cycles ourselves. Often, we discovered that (1) we really had not thought so deeply about a particular scientific process or concept as we were forced to do when we attempted to model it, and (2) we unexpectedly and naturally entered into the scientific inquiry process, making new discoveries as we tested our models even though we created the models in the first place and, therefore, in some sense, should have known what the outcomes would be. These processes are a standard part of simulation development. It is natural to cross back and forth between engineering the simulation and scientifically investigating how the thing works. Many of us walk away from these sessions convinced of the powerful learning opportunities that designing and building models – not just treating them as black boxes – affords. Yet, such opportunities come only at a price: a commitment to learn programming, at least to some rudimentary extent. Perhaps schools should create more opportunities (science/technology electives or after school programs) to allow students to foster such skills. As our work demonstrates, StarLogo TNG’s ‘programming block’ interface and immediate feedback make it a serous candidate for contexts designed to foster such skills. It is also worth noting that, in our original design of the forest fire unit, we did not include an assessment but rather allowed the final game level design activity to function as a performance assessment of the unit’s learning goals. As is typical of a community contributing user generated content, students are quite motivated to complete such assessments because they are personally invested in the outcome. Students know that their peers will play and evaluate their levels, and strive to produce quality work that the community will value. The assessment is perceived as authentic by the student, since their work is not done solely for a grade. More educational research and design might productively focus on similar performance assessments that students perceive as having value to them. Finally, it is important for teachers to remember that randomness is an essential feature of many simulation models. Students often struggled with how to interpret data that contained randomness, such as mistaking it for the kind of ‘error’ they might encounter when doing traditional lab experiments. Teachers can use such opportunities to raise questions like: ‘How many trials is enough to establish confidence in one’s findings?’ ‘If a certain firebreak radius works nine out of ten times, would you feel safe protecting your home with this radius value?’ Students can gradually learn to think probabilistically about scientific processes, but it is a habit of mind that is difficult for many people, including adults. Physics Curriculum Simulations and games can motivate learners and provide ways for students to develop intuitive understandings of projectile motion (Jimoyiannis & Komis, 2003) and other abstract physics phenomena (Squire et al, 2004). However, because science practice involves the construction and validation as well as the application of scientific models, in order for classroom practice to better model scientific practice, science instruction should include the making as well as the using of models (Hestenes, 2007). Agent-based computer models are especially well suited for student inquiry and physics learning. The algorithmic thinking involved in programming such models emphasizes processes rather than facts (Cohen & Kanim, 2007). Programming provides students with an alternative means of expression that is precise and compact (Sharin & diSessa, 1993). Programming and algorithmic thinking offer alternative descriptions to complex phenomena that may be more accessible than algebraic descriptions to many students (diSessa, 2000) but questions remain as to how (a) to get students motivated to learn through programming, and (b) to make the 84


The Simulation Cycle technical aspects of programming accessible to a wide array of students and teachers. This is where the intersection of games and simulations comes into play: Learning scientists are increasingly turning to video games as tools for learning (Gee, 2007). Games can provide an ideal pathway into simulations for students, not only providing motivation and context for those who have grown up playing games (which is approaching 100% of students) but also leveraging valuable learning processes that are welcome in the classroom. Learning Physics through Programming. As diSessa’s (2000) work shows, fundamental physics concepts can be made accessible to students as early as sixth grade by using simple programming activities. We hypothesized that, at the high school level, programming can help students build a deeper understanding of traditionally difficult physics concepts. Thus, we introduced StarLogo TNG programming activities into five physics classes at a private school in the Boston metropolitan area over a two-year period, motivated by the belief that programming can foster learning in a way that ‘transforms the experience of students substantially from doing what adults say in semicomprehension into a really rich and appropriate kidlike experience, more like what they want to do and can do without adults intruding awkwardly’ (diSessa, 2000, p. 38). To this end, we implemented a short ‘Introduction to Programming through Games’ unit along with other introductory physics activities in physics classes in the fall of 2008. Over a twoweek period, students were introduced to and given practice with tools they would be using throughout the year, including Loggerpro (Vernier Software) for data collection and Excel (Microsoft) for data analysis. Equation solving and textbook reading skills were also practiced. The first programming activity introduced the concept of computer code as a set of instructions and StarLogo TNG basics in a game-like setting. Students learned programming fundamentals as they designed a set of instructions for their agents to follow. The agents simply had to move around within Spaceland and score as many points as possible by moving to higher ground. In the game, agents (shaped as ducks) move about the terrain, gaining points based on the elevation of the land that they are on; they gain points on the red and yellow (above ground) and lose points on blue (below ground) (see Figure 8).

Figure 8. The terrain in Spaceland used for the first physics programming activity. Different colored areas are different heights.

Given the code-structure, illustrated in Table II, students chose from a subset of movement blocks to try to maximize the ducks’ time on red and yellow. They kept a record of the changes in code and corresponding changes in score over time. Subsequent discussions of programming strategies brought each group’s discoveries to the attention of everyone in the class. Students then completed a follow-up practice activity in which they used the same programming blocks in a different game requiring a different strategy and an on-computer programming practical as part of the unit assessment. Both the practice activity and assessment components were critical in establishing games and programming as a serious part of

85


Eric Klopfer et al the curriculum in the minds of the students, and gave the teacher important feedback into the nature of the students’ learning.

Table II. An excerpt from the instructions given to students in their physics programming activity.

Assessment Results. In addition to assessment scores, feedback came from responses on blogs that students kept during the class. Students responded to several reflective questions on their blog such as, ‘What did you learn about science and/or science simulations from the programming of games you did?’ and ‘What did you perceive as being game-like about the activities that you did? What wasn’t game-like?’ Four important themes emerged in their reflections: an improved understanding of input and output, a better grasp of the role of random fluctuations in physical systems, understanding scientific methodology, and the creative nature of such problem-solving endeavors. The connection between input and output is an important systems concept, one seen throughout physics and engineering. However, students are not always given opportunities to observe and understand such connections in the everyday classroom. In this activity, one student not only made a surprising connection between input and output but connected their observation on the computer to previous experiences in their physics lab. I learned a lot about input and output. If you input something you are going get some sort of output. Like in this game, you input code and you get an output as movement of the agent. The same thing goes for the spring lab we did. We input mass and we got a certain amount of stretch as our output.

The ability of this student to transfer the understanding from the game activity to a physics lab in a different sub-domain demonstrates the potential transferability of the knowledge gained through the programming activities to knowledge of the physical world. A second important issue that arose in the blogs was regarding another important systems concept: the role of random fluctuations in physical systems. This concept is often difficult for students to grasp, as they chalk up randomness to ‘experimental error’. Students often struggle with randomness and how one deals with the random fluctuations of data in real experiments. They want to know whether or not they got the right answer in a math problem, physics problem, or science lab. Many are confused when told that, due to randomness in the system, a range of results is acceptable. In contrast, after our unit, one student made the following observation: 86


The Simulation Cycle In reality, randomness is everywhere. However, in the program, randomness was just as inclusive in obtaining results. Therefore, I had to expect varied results every time we collected a score. There is virtually no way to completely eliminate randomness. (Emphasis added)

When this student saw that repeated runs of her game gave different scores, she was taking an important step toward understanding the need to repeat experiments and replicate results. Such insights can go a long way toward helping students understand randomness and varied results in other environments. Other students commented on the scientific method. One physics student wrote a response in her blog evidencing a greater understanding of scientific methodology as a result of the programming activities, showing an understanding of the connection between the two cycles within the simulation paradigm outlined here (see Figure 1), recognizing the need for iteration between programming and experimentation: From the programming games I learned that there is a different way to solve every problem. The method my partner and I used was trial and error that seemed to be the most popular method in our class. By doing this continuously we became frustrated easily but we learned that we had to move forward. This relates to science in a way that when conducting an experiment, the scientist gets frustrated but pushes through to get an end result.

Initially, students did not know how to move forward when trial and error failed. In part, this activity provided them an opportunity to make analysis and reasoning across trials part of their personal repertoire of problem-solving strategies. Finally, students commented about the creativity involved in the problem-solving activities they engaged in. For example, one student wrote: Problem solving is the best way to learn because we are forced to understand every angle. The program also triggers my creativity, which I believe is not as important in school. Instead of knowing a certain way to figure something out like math problems, we have to discover different ways to reach the quickest time possible. This type of learning is out of the ordinary and I believe it is a necessity! (Emphasis added)

As this example illustrates, the programming activities fostered valuable skills in problem solving and creativity that are often (as the student points out) ignored in school. The nature of the programming task we designed, though certainly analytical at heart, gave students an opportunity to discover different ways to reach a goal rather than just ‘knowing a certain way’ determined by the teacher or the closed nature of a problem. Such flexibility and problem solving are part and parcel of what it means to take a more constructivist approach to teaching science. As one student boldly stated, ‘This game has truly taught me not only the idea of programming but also some ideas about life’. Throughout these activities, students were primarily engaged in the Scientific Method, as shown in the first loop of Figure 6. They tinkered, tested and observed the outcome of many code changes. In best-case scenarios, their observations informed the changes they made in code, using the second loop of Figure 6. More often, however, they struggled with using their observations in productive ways. When code changes didn’t improve their score, students were quickly frustrated. When they didn’t receive output that was commensurate with their changes, they resorted to blind trial and error, which led to more failure and frustration. However, given the ‘game’ context of the lesson, this frustration provided an opportunity to deeply engage students in discussions of problem-solving theory and practice. Students were open to suggestions that feeling frustration was a signal for the need to step back, slow down, observe carefully, and engage the intellect in determining a next step. Despite their periodic frustrations, students’ programming skills did develop throughout the unit and they began to demonstrate that they understood the many ways in which they could explore and learn from simulations and games. The confidence and abilities they gained throughout the game unit helped build a foundation of experience that would allow StarLogo TNG to serve them as a learning tool for more traditional physics material later in the year. Kinematics. That opportunity came soon after, during a kinematics unit in the topics of acceleration and two-dimensional motion. The acceleration section of the kinematics unit included not only labs, text reading, algebraic equations and problem solving, but also a programming component 87


Eric Klopfer et al where students followed a structured worksheet to build StarLogo TNG code for acceleration (see Figure 9). After completing the acceleration activity, students used keyboard controls to control a virtual car within the StarLogo TNG environment, and then used the terrain editor to design a 3D landscape for driving. These worlds took many forms. There were tracks for racing cars, fantasy worlds for exploration, mazes to explore and treasures to collect. In this activity, students participated in both the first and second loops of Figure 6. They were in the upper cycle when they built their game and fantasy worlds but moved into the lower cycle to test their changes in landscape and acceleration code. Acceleration and turn parameters had to be adapted to fit each unique world, but the desire to build a smoothly functioning fantasy world motivated students to unpack the details of their acceleration code. Thus, though explicitly engaged in the design and construction of simulations and games (the second loop of Figure 6), students were tacitly building their skills with the acceleration algorithms as well. Motivation, both at the computers during class and outside class time, was high. As one student in academic trouble responded when asked if he did his physics homework, ‘We do most of our work in class, but at night I plan my strategies for the next day’.

Figure 9. Excerpt from one of the student worksheets in kinematics explaining how to program acceleration in StarLogo TNG.

Our belief was that the code and game-building experience would leverage a deeper understanding of the concept of acceleration than many students were able to get from traditional algebraic representations alone. Preliminary data suggests our belief was well founded. Assessment for the acceleration unit included not only the standard fare of multiple-choice questions and problems but also a comparison of programming code and algebraic representation of motion for students to analyze. In one question, students were asked whether the given code or equation represented constant velocity or accelerated motion. In Figure 10, assessment questions are followed by representative responses of students who answered the multiple-choice parts of the questions correctly. Though the responses of the students who answered correctly show that they recognized that the equations represented accelerated motion, they did not connect the algebraic equations to the actual motion of real objects. They simply responded with definitions. On the other hand, the code told them exactly what the object did as it accelerated, as evidenced by the specificity of their answers. This preliminary evidence supports diSessa’s (2000) claim that programming is better suited to learning about motion than is algebra. Programming is oriented to process, embodies an appropriate level of abstractness and is directed locally and specifically to a particular motion. At the same time, this evidence also suggests that, although programming gave students an abstract and quantitative understanding of acceleration, understanding the algorithm may not help them understand the algebra. 88


The Simulation Cycle

Figure 10. Part of acceleration unit assessment with student responses.

In addition to helping the students learn physics, the game-design experience in the acceleration unit spawned a burst of creativity and deep involvement in the design and construction process. Students’ responses to follow-up questions – asking them (a) to detail the code and concept of which they were most proud and (b) what they felt they learned about programming, physics, and their own learning and creativity, throughout the process – evidences their understanding of that process. One group, for example, who built a second set of acceleration code blocks because they wanted a second car in their game, responded: J and I got the idea of racing each other randomly. One day J said, ‘We should race each other’ and that we did. First it was difficult to understand how to create a second car in which to control but after talking to [the teacher] about how to create it, it was not as difficult as we had previously thought. We created each speedup/slowdown/turn block into a ‘jspeedup’, ‘jslowdown’, ‘jturn’ block [variations on the procedures that they had first created] and changed the controls on the keyboard to something different. We changed the forward button to ‘w’ and turn [buttons] to ‘a’ and ‘d’. After completing this, it didn’t work for some reason. This puzzled J and I. After questioning [the teacher], he told us that we had to put extra stuff into the ‘race’ section of the programming. That we did: The process of creating this new car was a bit aggravating at times but at the end of completing this task, it was well worth it. Having a second car to race against was really a great time. We were very pleased with our results. (Emphasis added)

89


Eric Klopfer et al In this excerpt, two students, motivated by the capability to compete, engaged in an iterative design cycle to create a game that would allow them to achieve their goal. The value of and pride in their product made the sometimes aggravating attention to detail, which is, after all, part of science, worth the effort. When asked to reflect on what they learned throughout the process, they responded: • Programming – ‘Programming is a very intricate process that takes a while for a complete project to be completed. However, the results always prove to be very pleasing.’ • Physics – ‘We learned about the restrictions that an object had in its environment and how it reacted to them. We also learned about accelerated motion and how an object handles different speeds, especially with our steering setting. The faster we made the vehicle move, the harder it became to control its movement.’ • Your own learning – ‘We learned that electronic experiments are very helpful in learning about physics.’ • Your creativity – ‘The world in the beginning of the process was essentially a blank canvas, and we were able to construct an environment based on our own ideas. We had no difficulty doing this because StarLogo TNG provides all the tools needed to ‘paint’ with no limits.’ A second example comes from a student group that was frustrated when, after a change in code, their car would not move. In response to ‘their biggest pride’ question, they wrote: After we changed our agents from people to animals, we figured that a forest setting would fit the game better than the urban setting that we had previously. In the process, we changed the color of the road from black to brown, which conflicted with the code that was set. Oblivious to the specific problem, we failed to solve it. Later, [my partner] figured out the problem and was able to configure the code according to our new setting. When tested, the vehicle was then able to move forward again and the problem was resolved. We were extremely excited to see the vehicle operating again because we are very proud of the environment we created and we wanted to see it being actually used. (Emphasis added)

Here, students show ownership of their project and their own learning through the pride in their game – and how that pride plays out in terms of the tenacity in fixing the problem and, eventually, bringing their game to life. One group member sent an email to the teacher hours after class, stating ‘I figured out the reason the car wouldn’t move forward in the new environment. It was because the road was brown and there was an if/then statement underneath the slowdown code that kept it from moving ... I’ll show you it tomorrow. =)’. It is often difficult to get students to spend the time and get the experience necessary to learn fundamental physics concepts. Here, motivated by the positive outcomes from their game designs more than anything else, students invested time and in return had the kind of first-hand experiences required to internalize such concepts adequately. Projectile Motion. The focus of the next unit was motion in two dimensions. Many strategies, including simulations, have been designed to help students understand that the vertical and horizontal motions of a projectile are independent of one another, but the concept of simultaneous but independent change remains difficult and frustrating to teach and learn. We hypothesized that by adding the programming of two-dimensional (2D) motion in a virtual world to the usual mathematical analysis of that motion in the physical world, we could enhance understanding of this difficult concept. The unit was designed to challenge students’ pre-existing ideas about independent motion, by starting with the principle of independence more generally. Students began the unit by building separate simple programming procedures that changed an agent’s attribute like color, size or location. Procedures could be run separately or simultaneously, producing many humorous combinations of change, such as creatures that independently changed shape and altitude. Students had little difficulty seeing that their agent’s color change was independent of changes in its shape or its movement in the x or y direction. After this initial activity on independence but before beginning a more formal study of 2D motion, students were given a simple treasure hunt game to explore and modify. It was used to assess their understanding of the relationship between motion in the x and y dimensions with 90


The Simulation Cycle diagonal motion. All movement was directed by the arrow keys, which gave only 0, 90, 180 and 270 degree motion. Treasures were scattered all over a world and opponents moved to draw paths that prevented one another from reaching them (see Figure 11). Drawing upon previous experience from playing games, the students’ fingers had an ‘understanding’ of the relationship between the x and y components of diagonal motion before they formally learned about it or could verbalize the relationship. As is common in many games, they could move diagonally by simultaneously moving in the x and y directions. When students were asked to write about how they came up with their strategy for movement, however, they found it difficult to put the process into words. The following student response is a case in point: To move the elephant around the game map we used the arrow keys. The up arrow would move the elephant up, the down moved it down the left moved it left and the right moved it right. If you wanted to move at an upward 45 degree angle in either direction then you used the up and either left or right arrow key. If you wanted to move the elephant in a downward 45 degree angle then we used the down arrow key and either the right or left arrow key. If you just held the two keys together you were stuck going at an angle but if you let up on one of the keys or were to continue tapping it the elephant would be changing direction. For example if you were holding the up and left arrow key and you wanted the elephant to go more up then you would stop pushing the left arrow key and if you wanted the elephant to go more left then you would stop pressing the up arrow key.

Although the student has grasped the basic concept that pressing the keys together allows the character to move at a 45-degree angle, he stops short of a mechanistic explanation of why. Another student offers an answer that shows slightly greater understanding: We moved our agent by using the arrow keys. If we wanted the agent to move right, then we pressed the right key. This pattern was followed if we wanted to move the agent in any direction. To move the agent upwards and diagonally to the right, then we pressed the up and right keys. This pattern could be followed to make the agent move in any diagonal direction. The agent turns by using different amounts of X and Y, for example, we can press a lot of X and then do a few taps upwards and it will move in a different direction as opposed to just holding down X. (Emphasis added)

Here, the student displays the additional understanding that one can change the angle by varying the ratio of x to y movement.

Figure 11. The treasure hunt game used as a part of the projectile motion unit. Paths drawn by opponents stop the player from reaching treasures on the other side.

As the unit progressed, it delved deeper into independent motion but continued the theme around game design. To get realistic projectile motion, students added a procedure for the negatively directed acceleration of free-fall to their independent x and y motions. To get an agent to jump over an obstacle, students had to separately build – yet simultaneously execute – a constant 91


Eric Klopfer et al velocity, move-forward procedure and a vertical-jump procedure that employed acceleration. Preliminary evidence did show that students transferred knowledge from their programming experience to new situations. For example, in a lab report where students had to calculate the range of a projectile, one freshman wrote: Programmers program motions independently in the virtual world; on the other hand physicists see the vertical and horizontal motion as independent of one another in the physical world. … I believe when an object is moving vertically it is independent and not interfered by horizontal motions. Our ball fell within our predicted value. This confirms our assumptions of vertical and horizontal motions being independent.

As a final example showing the potential of transfer and the extent of student ownership, consider the responses of juniors in a physics class who built swimmer-in-river simulations as part of a similar unit on vectors. The simulations have a swimmer with velocity s swimming at a given selected across a river that has current velocity, r. After building and playing with the simulation, students were assigned two-dimensional motion problems for homework. During a class discussion over the answer to a contested problem, one student went, unprompted, to the computer and opened his swimmer model. He plugged the variables of the problem into the code of the model and ran the model. As the swimmer reached the opposite shore he exclaimed, ‘I told you I was right!’ Later, when asked in an interview why he chose to use the computer rather than mathematical analysis to prove his point, he said. ‘This way you could see I was right’. These early studies demonstrate that student-programmed, game-like simulations are well suited for student inquiry and physics learning. The algorithmic thinking involved in programming emphasizes processes and the programming experience leads to a quantitative as well as qualitative understanding of physics concepts. Students extend their intuitions and use programming and game-design as a means of expression that is both precise and creative. Evidence supports the idea that algorithmic thinking may be more accessible than algebraic descriptions of physics concepts to many students. As our studies show, StarLogo TNG is a robust, stable, classroom-ready platform that can make the technical aspects of programming games and simulations accessible and motivating to students. Conclusions The curriculum and tools used as a part of these classroom simulation activities are significant in their more holistic representation of the simulation process and the way they are designed at the intersection of games and simulations. Recognizing that the role of simulations in the scientific enterprise goes well beyond the use of ‘black box’ simulations, we sought to design activities that combine both development and use of simulations at different stages in the curriculum. These activities provide multiple points of access into the simulation cycle that includes recognized cycles in engineering design and the scientific method. Acknowledging the commonalities between games and simulations, our activities use games to motivate the construction and use of simulations in content-heavy science classes. Our approach can foster significant classroom use of simulations that are relevant to the curriculum and to the students – but not without some significant work and a serious change in mindset. Our major findings and caveats include the following. The intersection of games and simulations is a large, rich space. There are many approaches to working at the intersection of games and simulations. As shown in the case studies detailed here, such approaches can range from having students making games out of simulations to having them develop games that incorporate simulations at their core to simply using games in instruction based on simulation principles. Such combinations are motivating and empowering to students and can be tied to the standard curriculum to work even in traditional classrooms. The entire simulation cycle can happen in real classrooms, but it takes time. Both the physics and middle school science curricula, described above, as a whole covered every part of the simulation cycle (Figure 6), although few activities incorporated the entire cycle. Choosing and using a simulation activity that supports a given curricular goal will bring simulations into the classroom in a manageable, sustainable and time-effective manner. The blocks-based programming of Starlogo 92


The Simulation Cycle TNG puts programming aspects of the design and construction cycle into reach for most students and teachers. But this more integrated approach takes more time (at least at first) when compared to didactic approaches (or even more limited simulation-based approaches). Such time commitments may be justified as preparation for future learning, but how far that preparation goes, so to speak, has yet to really be measured. Curriculum development that incorporates the whole simulation cycle can be done, but it takes time. Not only does classroom use of a more complete simulation cycle take more time, but the development of such curricular materials also takes time. Similar lessons must provide sufficient scaffolding to insure student success while also providing the flexibility for creativity and innovation. Encouraged by the responses of teachers and students and the need to bring meaningful simulation learning into middle and high school classes, we continue to develop and improve curricular materials that will make game-based simulation learning materials available to all interested teachers and students. We have clear indications of learning (and even transfer) but how far it goes remains unclear. The results reported here indicate student learning of science concepts and process through simulations. Some middle school students showed gains in understanding even after short exposures to the StarLogo TNG based games. Many physics students showed substantial gains in understanding of fundamental concepts critical for later lessons. Some students showed the transfer of their specific conceptual understanding or simulation skills to other domains. However, student understanding of algorithmic principles of physics did not seem to help them on their algebra-based physics problems. One needs to consider why and determine if simulations can help students make that leap. Student engagement is universal (or nearly so). Skeptics often think of the use of games in the classroom as a way of softening the curriculum, particularly in difficult subjects like physics. The students in this physics class show that not to be the case. Surveys showed that 75% of the students who used StarLogo TNG agreed with the statement that the StarLogo TNG unit was more difficult than other units, while 100% of the students felt the unit was more rewarding. This is clearly an example of what Papert (cited in Caperton, 2005) has called ‘Hard Fun’ – learning that is fun not in spite of it being hard but rather because it is hard. The use of games, simulations and programming provides the potential for promoting student understanding, modeling scientific practices, increasing rigor, and engaging students in science, technology, engineering and mathematics content – all of which have obvious value in preparing students to be scientists, think scientifically, or just be scientifically literate citizens. The exact form this innovation takes in classes is still to be determined, but we believe there is an entry point for teachers and students at all levels, from dabbling with simulation use (in the lower cycle) to projects that take students through multiple iterations. Approaches that facilitate significant classroom use of simulations that remains relevant to the curriculum and to the students promise worthwhile gains and, perhaps even, a change in classroom culture. As Polman & Pea (2001) have found, such activities often require some adjustments of teaching style. For example, familiar ‘cultural tools’ such as Initiation–Reply–Evaluation (IRE) sequences simply no longer work because they are premised on known answers and teacher-driven activity. In contrast, StarLogo TNG simulations, because they are transparent, promote both strong student and teacher input into the learning process. As such, they have the potential to stimulate and improve science education. References American Association for the Advancement of Science (1993) Benchmarks for Science Literacy. New York: Oxford. Begel, A. (1996) LogoBlocks: a graphical programming language for interacting with the world. Thesis, Massachusetts Institute of Technology. Belcher, J., Murray, J. & Zahn, M. (1999) Force Field: using animation in teaching electromagnetism. http://web.mit.edu/ jbelcher/www/NSF.html (accessed November 23, 2008).

93


Eric Klopfer et al Buckley, B.C., Gobert, J., Kindfield, A.C.H., et al (2004) Model-Based Teaching and Learning with Hypermodels: what do they learn? How do they learn? How do we know? Journal of Science, Education and Technology, 13(1), 23-41. http://dx.doi.org/10.1023/B:JOST.0000019636.06814.e3 Caperton, I. (2005) For Seymour Papert ‘Hard Fun’ is the Essence of Good Games and Good Education, Telemedium: the journal of media literacy, 52(1 & 2),16-19. Cohen, E.W. & Kanim, S.E. (2007) Algebraic Difficulties in Physics. http://spacegrant.nmsu.edu/NMSU/2004/cohen.pdf (accessed November 16, 2008). diSessa, A. (2000) Changing Minds: computers, learning and literacy. Cambridge MA: MIT Press. Gee, P. (2007) What Video Games have to Teach Us about Learning and Literacy. New York: Palgrave Macmillan. Hestenes, D. (2007) Modeling Instruction in High School Physics, Chemistry, and Physical Science. http://modeling.asu.edu/modeling-HS.html (accessed November 16, 2008). Jacobson, M. & Wilensky, U. (2006) Complex Systems in Education: scientific and educational importance and implications for the learning sciences, Journal of the Learning Sciences, 15(1), 11-34. http://dx.doi.org/10.1207/s15327809jls1501_4 Jimoyiannis, A. & Komis, V. (2003) Computer Simulations in Physics Teaching and Learning: a case study on students’ understanding of trajectory motion, Computers & Education, 36, 183-204. http://dx.doi.org/10.1016/S0360-1315(00)00059-2 Keeley, P., Eberle, F. & Farrin, L. (2005) Uncovering Student Ideas in Science, vol. 1. Arlington: NSTA Press. Kuch, B.B. (2007) Developing and Implementing a High School Simulation Course to Provide Rigor and Relevance to the Curriculum, in Proceedings of the 39th Conference on Winter Simulation: 40 years! The Best Is Yet To Come, 2344-2352. Piscataway, NJ: IEEE Press. Massachusetts Department of Education (2006) Massachusetts Science and Technology/Engineering Curriculum Framework. Malden, MA: Massachusetts Department of Education. Meir, E., Perry, J., Stal, D., Maruca, S. & Klopfer, E. (2005) How Effective are Simulated Molecular-Level Experiments for Teaching Diffusion and Osmosis? Cell Biology Education, 4, 235-248. http://dx.doi.org/10.1187/cbe.04-09-0049 Moodley, S. (2004) The Effects of Computer-Based Dynamic Visualization Simulations on Student Learning in High School Science. PhD thesis, Boston University, USA. National Research Council (1996) National Science Education Standards. Washington, DC: National Academy Press. Pallant, A. & Tinker, R. (2004) Reasoning with Atomic-Scale Molecular Dynamic Models, Journal of Science Education and Technology, 13(1), 51-66. http://dx.doi.org/10.1023/B:JOST.0000019638.01800.d0 Polman, J.L. & Pea, R.D. (2001) Transformative Communication as a Cultural Tool for Guiding Inquiry Science, Science Education, 85, 223-238. http://dx.doi.org/10.1002/sce.1007 Scheintaub, H., Klopfer, E., Scheintaub, M. & Rosenbaum, E. (in press) Complexity and Biology – bringing quantitative science to life in science classrooms, in F. Roberts (Ed.) Linking Mathematics and Biology in High School Classrooms. New York: Springer. Sharin, B. & diSessa A.A. (1993) Dynaturtle Revisited: learning physics through collaborative design of a computer model, Interactive Learning Environments, 3, 91-118. http://dx.doi.org/10.1080/1049482930030201 Sherrell, L.B., Robertson, J.J. & Sellers, T.W. (2005) Using Software Simulations as an Aide in Teaching Combinatorics to High School Students, Journal of Computing in Small Colleges, 20(6),108-117. Squire, K., Barnett, M., Grant, J.M. & Higginbotham, T. (2004) Electromagnetism Supercharged! Learning Physics with Digital Simulation Games, in Y. Kafai, W.A. Sandoval & N. Enyedy (Eds) Proceedings of the 6th International Conference on Learning Sciences, 513-520. Mahwah: Lawrence Erlbaum Associates. Stieff, M. & Wilensky, U. (2003) Connected Chemistry – incorporating interactive simulations into the chemistry classroom, Journal of Science Education and Technology, 12(3), 285-302. http://dx.doi.org/10.1023/A:1025085023936 Wiggins, G. & McTighe, J. (2001) Understanding by Design, 2nd edn. Upper Saddle River: Prentice Hall. Wilensky, U. (1997) StarLogoT. Evanston, IL: Center for Connected Learning and Computer-based Modeling, Northwestern University. http://ccl.northwestern.edu/cm/starlogoT/ Wilensky, U. (1999) NetLogo. Evanston, IL: Center for Connected Learning and Computer-based Modeling, Northwestern University. http://ccl.northwestern.edu/netlogo/ Wilensky, U. (2003) Statistical Mechanics for Secondary School: the GasLab modeling toolkit, International Journal of Computers for Mathematical Learning, 8(1), 1-41. http://dx.doi.org/10.1023/A:1025651502936

94


The Simulation Cycle Wilensky, U. & Reisman, K. (2006) Thinking like a Wolf, a Sheep or a Firefly: learning biology through constructing and testing computational theories – an embodied modeling approach, Cognition & Instruction, 24(2), 171-209. http://dx.doi.org/10.1207/s1532690xci2402_1

ERIC KLOPFER is Associate Professor and Director of the Scheller MIT Teacher Education Program. Klopfer is a trained biologist, and former high school/middle school computer teacher and technology coordinator. Klopfer’s research focuses on the development and use of computer games and simulations for building understanding of science and complex systems. His research explores simulations and games on desktop computers as well as handhelds. He is the creator of StarLogo TNG, a new platform for helping children create 3D simulations and games using a graphical programming language. On handhelds, Klopfer’s work includes Participatory Simulations, which embed users inside of complex systems, and Augmented Reality simulations, which create a hybrid virtual/real space for exploring intricate scenarios in real time. He is the codirector of The Education Arcade, which is advancing the development and use of games in K-12 education. Klopfer’s work combines the construction of new software tools with research and development of new pedagogical supports that support the use of these tools in the classroom. He is the co-author of the book, Adventures in Modeling: exploring complex, dynamic systems with StarLogo, and author of the book, Augmented Learning: research and design of mobile educational games (MIT Press, 2008). Correspondence: Eric Klopfer, 77 Massachusetts Avenue, MIT Bldg 10-337, Cambridge, MA 02139, USA ([email protected]). HAL SCHEINTAUB received his PhD in biophysics in 1973 from SUNY Buffalo. From 1980 to the present he has been a full-time secondary school science teacher, teaching in both public and independent schools. For the last 10 years he has taught at Governor’s Academy in Byfield, Massachusetts and has collaborated with Dr Klopfer at the Scheller Teacher Education Program in MIT for eight of those years. He has worked to create ways to bring complex systems concepts and computer programming into high school science students. Correspondence: Hal Scheintaub, 1 Governor's Academy, 1 Elm Street, Byfield, MA 01922, USA ([email protected]). WENDY HUANG is the program manager for the Scheller Teacher Education Program. Her current work includes supervising undergraduate student teachers, assisting instructors of teacher education courses at MIT, and developing and implementing curricula for the StarLogo TNG project. Prior to her work at MIT, she worked for three years on the design, pilot, and publication of several middle school math, engineering, and science curricula (Building Math, Science Corps) at the Museum of Science (Boston) and Education Development Center and was a Teach for America math/science teacher for four years at a public junior high school in the South Bronx, New York. Correspondence: Wendy Huang, 77 Massachusetts Ave, MIT Bldg 10-337, Cambridge, MA 02139, USA ([email protected]). DANIEL WENDEL graduated from the Massachusetts Institute of Technology in 2006 with his Master of Engineering degree in computer science. While a student at MIT Daniel’s coursework focused on computer systems, development, and human–computer interaction. Daniel also completed his humanities concentration in education through the MIT Scheller Teacher Education Program (STEP), fueling his interest in education and educational technologies. Daniel has been a member of the StarLogo TNG development team since 2004, and he has led the team for the past year as a member of the STEP staff. Correspondence: Daniel Wendel, 77 Massachusetts Ave, MIT Bldg 10-337, Cambridge, MA 02139, USA ([email protected]). RICAROSE ROQUE managed the MIT Scheller Teacher Education Program’s collaboration with the Department of Homeland Security (DHS) to develop Starlogo TNG curriculum units inspired by DHS research in natural disaster prevention and management and epidemiology. She completed an M.Eng. in 2007 and B.S. in 2006, both in Computer Science, from the Massachusetts Institute of Technology. Ricarose conducted her M.Eng. research at STEP on StarLogo TNG, which enabled her to combine her background in engineering with her interests in education and user-centered design. Since joining STEP in 2006, she has managed the redesign of the StarLogo TNG front end 95


Eric Klopfer et al interface and implementation, led numerous workshops for students, and helped develop StarLogo TNG curriculum units for middle school teachers and students. Correspondence: Ricarose Roque, 77 Massachusetts Avenue, MIT Bldg 10-337, Cambridge, MA 02139, USA ([email protected]).

96


The Simulation Cycle: combining games, simulations

The Simulation Cycle: combining games, simulations

Suggest Documents

Simulations and games - Hal

Rhetorics, Simulations and Games

Combining Life Cycle Environmental and

Environmental Product Development Combining the Life Cycle ...

Simulation games, a learning tool - Training Games

Major Games Cycle - Google Sites

Combining Process Simulation and Agent

Metrics in Simulations and Games for Learning - Games Research Lab

Network Security Games: Combining Game ... - Semantic Scholar

Combining losing games into a winning game

Network Security Games: Combining Game ... - Semantic Scholar

Simulation games in operations management: the

Serious Games and Virtual Simulations Debriefing ...

Simulations and games - Archive ouverte HAL

Distinguishing between games and simulations: A ... - CiteSeerX

Agents for games and simulations - Springer Link

Learning Science Through Computer Games and Simulations

Adaptivity Challenges in Games and Simulations: A

Students'' Devised Classroom Games-Simulations: An ...

Serious Games, Interaction, and Simulation

Kids and Simulation Games - Itofisher.com

Physics Simulation Games - Semantic Scholar

A life cycle assessment framework combining

Investigation of PBL schemes combining the WRF model simulations ...