Running Head: VOLUNTEER COMPUTING GAMES Volunteer Computing Games: Merging Online Casual Gaming with Volunteer Computing

We describe volunteer computing games, a new paradigm for computing which merges volunteer computing and online casual games. In volunteer computing, the main goal is to harness the computational power of many users’ computers to solve a large computational problem. Unfortunately, participants in volunteer computing come from a very limited demographic—primarily middle-aged male computer experts. Since casual games have an almost universal appeal and, consequentially, a broad player base, we argue that volunteer computing games will significantly impact volunteer computing efforts by attracting and retaining more users. We describe one framework for implementing volunteer computing games to solve problems whose solution space can be modeled as a search tree. Using our prototype game as an example, we discuss general design principles that are of particular importance with volunteer computing games. We argue that the full power of volunteer computing games occurs when human computation is added to the mix. An ideal volunteer computing game will attract men and women of all ages and harness the computing power of both the users and their computers. We conclude with a very brief discussion of volunteer computing within Massively Multiplayer Online Games (MMOGs). Volunteer Computing Games 3 Volunteer Computing Games: Merging Online Casual Gaming with Volunteer Computing Introduction As scientists have become more frequently confronted by increasingly complex problems, research teams have turned to computers to aid in the exploration of these challenges. Unfortunately, technology has simply not kept pace. A single processor is no longer adequate for the problems researchers now encounter. As a result, a research team may not receive tangible results for weeks, months, or even years. One solution is supercomputing. However, this is not a practical choice for many researchers. Although supercomputers provide improved computing power, they are both uncommon and expensive. Thus, recent efforts have started to look towards parallel processing and distributed computing. A second option is the use of a computing cluster. But clusters are also not without their limitations. Since clusters contain a predetermined number of nodes, the system’s processing power has an inherent ceiling. A third option is volunteer computing. Volunteer computing is the process that allows people from across the globe to donate their computer resources in a joint effort, effectively creating an extremely powerful supercomputer. Currently, the world’s largest contributor to volunteer computing is the Berkeley Open Infrastructure for Network Computing (BOINC). BOINC uses approximately 400,000 volunteer computers, contributing an average of over 400 TeraFLOPS of combined processing power (Anderson & McLeod, 2007). Among the forty projects that BOINC sustains, two of the most popular are Rosetta@home and SETI@home. While Rosetta@home helps determine the 3-dimensional shapes of proteins in the hopes of curing diseases, SETI@home processes digital signals from a radio telescope at the Arecibo radio observatory (Anderson, 2004). Volunteer Computing Games 4 In order to understand the full potential of volunteer computing, consider that as of 2004, approximately 150 million personal computers were connected to the internet—a number that is estimated to exceed one billion by 2015 (Anderson, 2003). Of these machines, consider further that many are idle for nearly ninety percent of each day. Even active computer users typically use less than ten percent of their machine’s CPU (Justiniano, 2005). Accordingly, it is only logical for the research community to take advantage of this enormous source of unused computational potential. Volunteer computing does have its fair share of obstacles, particularly in regards to participation. Theoretically, the number of volunteer computing participants should be limited only by the number of computers online. However, the 400,000 active BOINC users currently compose less than 0.3 percent of the total number of computers connected to the internet. Therefore, taking into account that the average broadband internet user spends 48 percent of their leisure time online (Marketing Charts, 2007), we are exploring combining the accessibility and entertainment of online casual games with the effectiveness of volunteer computing. Casual games are “games that generally involve less complicated game controls and overall complexity in terms of gameplay or investment required to get through the game” (Wallace & Robbins, 2006). The IGDA reports that “the core and casual downloadable games sector grew over 100 percent from 2003-2004” (Wallace & Robbins, 2006). In fact, Nickelodeon recently announced that it will invest 100 million dollars in casual games over the next few years (Androvich, 2007). As casual gaming has become an increasingly popular leisure activity, the innovative possibilities to use them have similarly increased. Luis von Ahn’s research group at Carnegie Mellon is already taking advantage of casual gaming to assist in various human computing tasks, including labeling images on the web with Volunteer Computing Games 5 the ESP Game (von Ahn & Dabbish, 2004) and Phetch (von Ahn, Ginosar, & Kedia, 2006; vonAhn, Ginosar, Kedia, Liu, & Blum, 2007), locating objects in images with PeekABoom (von Ahn, Liu, & Blum, 2006), collecting common sense facts with Verbosity (von Ahn, Kedia, & Blum, 2006), and annotating sounds and music with Tag a Tune (Law, von Ahn, Dannenberg & Crawford, 2007). Recently, Von Ahn introduced several of his games on GWAP (http://gwap,com), a site devoted to what he calls “games with a purpose.” As von Ahn (2006) puts it, “Each year, people around the world spend billions of hours playing computer games. What if all this time and energy could be channeled into useful work?” It is important to note that von Ahn is interested in the computational power of the user, not the computer: “Rather than designing a computer vision algorithm that generates natural language descriptions for arbitrary images (a feat still far from attainable), we opt for harnessing humans” (von Ahn, Ginosar, & Kedia, 2006). Thus, von Ahn’s games work as alternatives to computer algorithms (Thompson, 2007). Currently, researchers at the University of Washington are developing a game called Fold It! (http://fold.it) that allows players to assist in predicting protein structures, an important area of biochemistry that seeks to find cures for diseases. The game presents visual representations of proteins to the player and allows them to use several key algorithms from the Rosetta@home code to find the optimal folding structure of the protein. By tracking the strategies of the best players, they hope to develop new heuristics to be incorporated into Rosetta@home as well as “go well beyond established limitations of all purely computational methods...eventually we want to leverage the human ability to think abstractly, and visually compose 3d structures... effectively tapping into a problem solving part of the brain instead of visual recognition part of the brain.” (Z. Popović, personal communication, February 14, 2008). As with von Ahn’s Volunteer Computing Games 6 research, the idea here is to harness the insights of the players, not necessarily their computers. They believe that by presenting geometrical optimization problems to many users, they can solve them better than by only employing traditional algorithms, since the insights of a few users may provide a heuristic that was previously unknown, or a clever user might stumble upon a better solution than what is already known. As far as we can tell, there are not similar efforts to utilize the popularity of online casual gaming to assist in non-human computation. Although the Fold It! Team may use their game to encourage participation in Rosetta@home, it appears von Ahn’s games are not being used to encourage participation in volunteer computing. Thus, we present Wildfire Wally, a prototype of a volunteer computing game—a casual game which is an implementation of a volunteer algorithm. In the game, players play as Wally, a red-bearded wilderness personality who is desperately trying to protect his forest from a raging fire. Wally extinguishes the flames by either dousing blazing trees with water, or by creating fire-lines that isolate a burning area of the forest. The player simply clicks on a cell to perform the appropriate action. If the cell contains fire, the fire is partially quenched. If the cell does not contain fire, some of the trees are removed to begin creating a fire line. As players progress from level to level, wind gusts and dropping humidity make containing the fire increasingly difficult. If a certain number of trees are not conserved, the player loses. Each move in Wildfire Wally corresponds to a decision in a traversal of a search tree for an instance of the maximum clique problem. This allows each player to exhaust a portion of a potentially enormous search tree. In addition, players have the option of running computation in the background while they play. Thus, many players from around the world can contribute to the solution of the same problem while playing a game. Volunteer Computing Games 7 When properly designed and implemented, a volunteer computing game allows a researcher to harness the computational power of both the player and her computer, as well as introduce her to the idea of volunteer computing in general, and a specific research project. We believe all of these outcomes have the potential to change the face of volunteer computing, both by increasing participation in these projects significantly—translating into added computational power—and increasing the sophistication of the contributions by clever players. Motivation Volunteer Computing Limitations While volunteer computing has proven to be useful to the research community, we have previously expressed that it has fallen far short of its full potential. Less than 0.3 percent of FIGURE 1 A mock-up of Wildfire Wally Volunteer Computing Games 8 computers participate in volunteer computing at all. Since the success of volunteer computing is entirely dependent on participation, it is a problem that deserves special attention. Currently, there are at least four distinct barriers to participation in volunteer computing. First, many people have never heard of volunteer computing and do not know that they can contribute to one of over one hundred active projects. This is particularly disappointing considering that the projects are varied enough to appeal to any given individual. If people are unaware of their existence, they cannot participate, no matter how eager they might be. Second, there is very little motivation for people to participate. Most forms of volunteer computing consist of merely running an application in the background. While individuals can be ranked by their contributions, these contributions correspond to the amount of time running the application rather than any meaningful actions completed by the users. Moreover, these rankings are dominated by teams and individuals with far superior computing resources, making the contributions of the average participant seem insignificant by comparison. Third, those who are currently active in volunteer computing come from a very limited demographic group. A survey of potential BOINC users indicates that 93.9 percent of 33,988 respondents are male. Only 4.7 percent of respondents fall in the age range of 0-19 years old (BOINC, 2007) Fourth, many people do not have the technological aptitude to participate. Contributing to volunteer computing often consists of downloading and properly configuring an application. For many users, this either exceeds their limited abilities, or is simply an extra step they are reluctant to take. In fact, about 38 percent of those participating in the survey indicated they do not run BOINC for technical reasons. Furthermore, the same BOINC survey shows that only 3.2 percent of all participants consider themselves “beginners” when it comes to using a computer. Volunteer Computing Games 9 A whopping 56.6 percent consider themselves computer experts (BOINC, 2007). Clearly, there is a divide. The average participant of BOINC projects is a middle-aged male computer expert. After considering that an estimated one billion people have access to the internet, most of whom are not middle aged male computer experts, it is apparent that the power of volunteer computing has gone largely untapped (Internet World Stats, 2007). Benefits of Casual Gaming While volunteer computing has fallen short of its potential, online casual gaming has flourished. An estimated 100 million people in the United States alone played a computer game in 2006 (Wallace & Robbins, 2006). In addition, the people who participate in casual gaming provide a refreshing divergence from the lopsided demographic involved with volunteer FIGURE 2 Participation by gender in volunteer computing using BOINC (BOINC, 2007), online gaming (ESA, 2006), and casual gaming at trygames.com (Macrovision, 2006). Volunteer Computing Games 10 computing. As can be seen in Figure 2, although only 6 percent of volunteer computing participants are female, 42 percent of online gamers are female. This number jumps to 71 percent when we consider only online casual games. In fact, the average casual gamer is a woman in her forties (Wallace & Robbins, 2006). Since different games may appeal to different demographics, it is possible for games to be designed for a specific age group or gender. Casual gaming is capable of breaking all of the barriers to volunteer computing. It has already been suggested that casual gaming can solve demographic problems, but the steadily growing popularity of casual gaming can also overcome awareness obstacles. This popularity has resulted in the emergence of numerous casual gaming websites, including AddictingGames, MiniClip, Yahoo! Games, and many more. Millions of internet users play games, and we expect that a well designed volunteer computing game would be at least as popular as other casual games. In addition, the fact that the game is contributing to important research might become an added attraction. The popularity of casual gaming also suggests that very little technical savvy is needed to participate. Many online casual games are played in a web browser and do not generally require users to install any additional software. In addition, casual games have simple rules and control mechanisms—often only requiring the use of the mouse, and maybe a few keys on the keyboard. Finally, because good games are engaging and entertaining, there is an immediate incentive to participate. In addition to the goals in the game, internet users have the pleasure of knowing that their gameplay contributes towards a larger problem. Participating in volunteer computing becomes enjoyable because it is a pleasant side-effect and not the main focus of