Behavioral Modeling and Simulation: From Individuals to Societies (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 Games T his chapter deals with massively multiplayer online games (MMOGs) as a tool for social and organizational modeling. An MMOG is a type of computer game that enables hundreds or thousands of p ­ layers to simultaneously interact in a game world to which they are con- nected via the Internet. Typically this kind of game is played in an online, multiplayer-only persistent world (Wikipedia, 2007). These games are a different kind of animal from the models and modeling approaches previ- ously discussed. MMOGs are simultaneously tools that allow players to interact with behavioral models, frameworks for building such models, and laboratories in which these models can be tested. What are Massively Multiplayer Online Games? Games, particularly videogames, are a recent addition to the modeling and simulation (M&S) tool suite. A videogame is defined as a mental contest,   n I this chapter, the committee makes references to online communities and game manufac- turers. Because of the nature of the games world, scholarly references are not often available, nor are they the most up-to-date or accurate sources of information.   hey are also not to be confused with “game theory,” a mature research area with strong T mathematical foundations established by von Neumann and Morgenstern (1944), but, as noted in Chapter 5, having significant constraints for application to real-world problems, the most notable for this context being the massiveness of the multiplayer community-inhabiting MMOGs; few game theory studies consider more than a handful of independent players or “agents” (Moss, 2001). There are a number of other limitations to the game theory approach, discussed extensively in Chapter 5. 261 

262 BEHAVIORAL MODELING AND SIMULATION according to certain rules, played with a computer, for entertainment. In the Department of Defense (DoD), the term used is “serious game,” which we define in this report as a mental contest, according to certain rules, played with a computer, that uses entertainment to further government or corporate training, education, health, public policy, and strategic communication objec- tives. An early examination of the potential for using games for modeling, simulation, and analysis originates in the report Modeling and Simulation: Linking Entertainment and Defense (National Research Council, 1997). Games are an interaction medium, a set of engaging and immersive models, and an interactive laboratory with which models and simulations can engage. As an interaction medium, games provide a way for humans to provide input and receive feedback in real time, participating in a running simulation. If the game is immersive enough, this running simulation will fully engage the attention of the game player, and that player will focus on game play to the neglect of the external world. The most commercially successful interactive games cyclically increase the adrenaline levels of the player, while demanding little in the way of mental focus. Games that demand great mental focus do poorly in the market and typically lose player interest. M&S systems that are embedded in such an interaction paradigm need to take this into account if the expectation is to make the M&S system as engaging as a commercial game. The desired outcome for this paradigm is that the M&S system will be so engaging that soldiers will continue to work with the simulation during personal time. Many train- ing simulations derived from the game America’s Army (described in more detail below) and similar games belong to this category (Zyda, Mayberry, McCree, and Davis, 2005). Games also contain a set of engaging and immersive models, models that look very interesting from the perspective of DoD. For example, a large number of meetings start out with the phrase “if only we could build an engaging game like SimCity”—that is, SimNavy for the Navy, SimAir for the Air Force, etc. (Zyda et al., 1998). There are many problems with such statements. The purpose for which the personal computer (PC) game SimCity  and The Sims, its direct descendent, was written to entertain   imCity S is a PC game in which the user controls several elements of managing a city, such as allocation of funding, distribution of community resources (police and fire stations, schools, etc.), and community layout. Maxis (now Electronic Arts Inc.) released the first version of SimCity in 1989. SimCity was the first game in the Sims franchise, and was the inspiration for other nonviolent open-ended games, such as Sid Meier’s Civilization (Electronic Arts Inc., 2006b). SimCity is partly based on Jay Forrester’s urban planning model, which is described in Chapter 4 (Electronic Arts Inc., 2007).   The Sims is a PC game in which the user controls individual characters (Sims) in a “virtual dollhouse.” The user is responsible for managing day-to-day needs of the Sims, such as their need for fun, hygiene, food, rest, and social activity. 

GAMES 263 and engage the game player. No real attempt was made in that game or its successors to accurately model the real world, nor was there any attempt to verify, validate, or accredit (VV&A) the results of that game—it is pure entertainment. It does, however, suggest a way in which one can develop potential outcomes or possibility spaces that can then be considered for further analysis. Probably the most interesting aspect of games like SimCity (http://simcity3000unlimited.ea.com/us/guide/), The Sims (http://thesims. ea.com/), and Civilization IV (http://www.2kgames.com/civ4/home.htm) is that these games were built for relatively small amounts of money and on schedule, and they still perform as extremely successful entertainment. DoD M&S programs with budgets two orders of magnitude larger have failed to deliver even a tenth of the capability to create a space of potential outcomes for consideration (Bennington, 1995). Games are also an interactive laboratory with which models and simu- lations can engage. They can play the role in social and organizational model­ing that linear accelerators play in particle physics—testbeds built and used to perform experiments and analyze results (Carley, Moon, Schneider, and Shigiltchoff, 2005). Like linear accelerators, MMOGs are expensive to build. The costs of successful immersive game development run from $8 million for the first two years of game development for a Spartan effort like America’s Army to more than $100 million to develop an MMOG and its infrastructure. To use a game as an interactive laboratory, it must be built or acquired before experiments can be performed with it. If the intention is to connect a social model to an MMOG for validation or improvement of the social model, money is needed either to build the MMOG or to acquire the use of it and the tools and permissions that allow its modification from a willing game development partner. Note, however, that the entire FY 2008 esti- mated budget of the Defense Advanced Research Projects Agency, $3.085 billion (see http://www.darpa.mil/body/budg.html [accessed July 2007]), is comparable to the current cash assets of a gaming giant like Electronic Arts (see http://finance.yahoo.com/q/bs?s=ERTSandannual [accessed July 2007]), plus the expected revenue of $1 to $1.5 billion from an operating MMOG. The size of the financial stakes for MMOG game companies means that getting the attention of a game development partner may rely more on personal connections or a fully funded joint basic research agenda than on any financial incentives that DoD could offer.   onsider C the single MMOG World of Warcraft with an estimated 8 million+ players paying $12.99 a month, for annual revenues of approximately $1.2 billion. 

264 BEHAVIORAL MODELING AND SIMULATION State of the Art Our review of the state of the art in MMOGs considers the three roles of games separately—games as an interaction medium, games as a set of models, and games as interactive laboratories. Games as an Interaction Medium Games as an interactive medium are always changing and improving. The drivers for innovation in the game industry are new technology for making the games ever more immersive and interactive, as well as industry competition and emulation. The driver for many years has been the increas- ing graphics speeds for PCs and consoles. That drive has made photorealism one of the major pushes for interactive games. Soundscape complexity has also made games more immersive as PCs and consoles have improved their sound support capabilities. The first Dolby 5.1 certified game, America’s Army, was developed in 2002, and now this feature is included in almost all games. We are near the point of diminishing return for graphics improve- ments, and people are now focusing on “fully interactive worlds.” The best example of the fully interactive world style is Rockstar’s Grand Theft Auto: San Andreas (GTA-SA). While the story line may lack redeeming social value, the game is so popular because the game player can interact with everything in the game’s world in a nonlinear fashion. This means that the player can navigate the world and do whatever comes to mind, without being constrained to a single story path, as in many games. The fact that there are missions to complete in GTA-SA is perhaps unimportant. It is the journey and the accompanying interaction that immerse and retain the player. If one wanted to have one game as representative of state of the art, then GTA-SA is that game with its fully interactive world paradigm, but the number of games attempting to copy that paradigm is quite large; the most notable is the Godfather game of Electronic Arts. Games as a Set of Engaging and Immersive Models The games Sims 2, by Electronic Arts Inc., and Civilization IV, by Firaxis, represent the state of the art with respect to games as a set of e ­ ngaging and immersive models. Sims 2 is a game that allows the player to create virtual characters, or Sims, and then direct them over a virtual lifetime. ­Settable parameters include gene mix across generations, life goals, popularity, fortune, family, romance, knowledge, financial status, and life- style. Sims can be pushed to extremes “from getting busted to seeing a ghost, from ­ marrying an alien to writing a great novel” (Electronic Arts 

GAMES 265 Inc., 2006a). The game allows the player to fulfill dreams, to try extremes, and to basically explore potential outcomes and possibility spaces. Civilization IV is a game that allows the player to create a civilization from its inception to its pinnacle and eventual demise. Players can choose peace and growth or choose a war footing, all from an easy-to-use interface. Civilization IV comes with a stream of easy-to-use modification tools that allow players to create and integrate their own interests into the game. As in Sims 2, Civilization IV allows the player to explore possibility spaces and potential outcomes. It is that capability that makes the modeling and simulation of these games very interesting to DoD and to the Department of Homeland Security. The question is often asked, “How do we connect these games, with near-zero modification, to real news feeds so that we can compare their ‘predictions’ against what subsequently happens in the real world?” Of course, these games are written to explore potential outcomes and not to be predictive, but there is a continual quest to achieve predic- tions, as in the film “Minority Report.” Games as an Interactive Laboratory MMOGs as interactive laboratories provide a state-of-the-art capability with respect to DoD goals. The idea is, for example, that if it were pos- sible to test models of what causes insurgencies against large groups of real online people, one could then understand and run the models backward to change the conditions so that the insurgencies do not happen. This is a tall order, built on several premises. The first premise is that models exist of the cause of insurgencies, and there is no way to test those models in real life. An additional premise is that if one could test and prove those models against real people in MMOGS, one would then have greater confidence in deploying the ideas embodied in the models in real life. The interesting point of such discussions is the desire to test social models using existing MMOGs rather than having DoD create its own testbed, thereby saving $100 million of testbed development costs that could be used to create models. To consider the top 10 MMOGs, how relevant or close to the problem are they? World of Warcraft, City of Heroes, City of Villains, Final Fantasy XI, Eve Online, Guild Wars, RuneScape, Everquest 2, Maple Story, Dark Age of Camelot, and Lineage 2 were the top 10 MMOGs listed by one site as of July 2007 (see http://www.the-top-tens.com/lists/top-ten-mmorpg-games. asp [accessed July 2007]). Although their visuals are far from realistic, and their stories are mostly about worlds that don’t exist and quests not linked to real life, the stories are all about the fights between good and evil, not unlike today’s global war on terrorism. So the thought is to take one or more of these MMOGs, modify the story a bit, put in links to the 

266 BEHAVIORAL MODELING AND SIMULATION predictive models to be tested, and then see if one can begin the process of predicting player behavior in the MMOG with connected systems. If that can be done, then perhaps it will be possible to run the models backward to stop insurgencies before they form or interdict them earlier before they gain strength. Relevance, Limitations, and Future Directions This section explores how MMOGs can be used to address DoD prob- lems, the limitations on that use, and the next steps needed to address those limitations. The discussion is organized around the three major capabili- ties offered by MMOGs: an interaction medium, a set of models, and an interactive laboratory. Games as an Interaction Medium Interactive games are great interfaces to models and simulations, because designers have created an interface typically more intuitive than that in comparable DoD-developed M&S systems. Interactive games typi- cally require no reading of manuals and have the player up and running in three minutes or less. The corresponding time is typically months for com- parable DoD M&S systems. So if the goal is to put models and simulations into the largest number of hands possible, then an interactive game interface is the right way to go. An additional advantage of interactive games is that their development and modification tools are easier to use than the simula- tion setup tools used by DoD. If defense simulations were as easy to set up as games, modelers’ ability to explore possibility spaces and potential outcomes would be dramatically increased. Games as an interaction medium are limited, at the moment, to games designed and implemented by the game development industry for entertain- ment purposes. For DoD use, those games must either be used as they are or modified with available tools. An additional limiting factor is that DoD does not typically have access to personnel skilled in game development. Interactive games, their supporting hardware infrastructures, their sup- porting software, and their input devices are under constant pressure to innovate and evolve. The biggest change coming in the next few years will be in the underlying models of human and organizational behavior, par- ticularly with respect to the modeling, display, and input of human emotion into the interactive game. Think of this as adding to the communication modalities already employed in games: visual display, auditory display, haptic display, and (coming soon) two-way emotional communication and display. Low-cost sensors that read parts of the human emotional state have already been designed for use as game input devices. These sensors provide 

GAMES 267 virtual sensors indicating mental focus, adrenaline, surprise/response, and relaxation, along with physiological measures of heart rate, blink rate, breathing rate, and oxygen level in the blood. Software using these mea- sures is already under development for use in evaluating games before they are shipped to determine what does or does not work in the produced game. Experiments are under way to determine how to use emotional inputs in games, including how to display appropriate emotions back to the player based on his/her personal state. Games as a Set of Engaging and Immersive Models The set of models inside engaging and immersive commercial games are proprietary and somewhat of a black box. We (DoD and its modeling researchers) cannot look at those models or modify them, other than the parameters exposed from the game’s interface or provided via modification tools. We cannot VV&A those models—but we probably haven’t really been able to achieve real VV&A with defense models and simulations, either (see Validation, Chapter 8). What we do know is that games like The Sims 2 and Civilization IV look quite capable for use in defense prob- lems, if only we could modify them, even slightly, for defense purposes. It would be interesting to know whether one could explore more of the space of potential insurgency outcomes with Civilization IV, developed at a cost of some $20 million, than with JSIMS, developed at a cost of $1.8 billion. Could modelers do that exploration with just the available Civilization IV modification tools? Likewise it should be possible to run experiments in virtual worlds simi- lar to Second Life (http://www.secondlife.com), perhaps with a somewhat less benign set of rules, which would have military and strategic applica- tions. For example, imagine Second Life with sovereign state entities, some of which were motivated to expand and dominate other regions of the game space. What would be the behavioral/organizational reactions of the other players? It is likely that genuine social experiments could be undertaken in settings like this, at a cost far below JSIMS. As another example, the popular board game Diplomacy is already available for online play (see http://www. diplom.org/index.py); it ought to be possible to modify it to bring it up to date in terms of state actors and allow for multiplayer states with their own internal decision-making processes, political parties, cultures, etc. Of course, there would be issues that would need to be thought through, such as access to the online games by hostiles, the potential for abuse of human subjects by traumatizing their avatars, and how to make the costs and benefits “real” (so that the players are not casual about starting virtual wars, for example). However, it seems clear that the potential gains are large enough to warrant some real effort devoted to overcoming these obstacles. 

268 BEHAVIORAL MODELING AND SIMULATION Since the models inside games are typically proprietary and not pre- cisely what DoD requires, this mismatch makes it hard for DoD to accept and utilize such models. If DoD were to establish its own serious game development studio, this limitation could be overcome. Also, since games are typically designed for entertainment, they often provide a misrepresentation of reality—for example, compressed time, inaccurate social networks, and missing cultural factors. If games are to be used for training, then greater attention needs to be paid to which aspects of reality need to be more carefully characterized in the games. This requires basic research on what factors are needed for what purpose, the inclusion of facts about social and cultural behavior, and the inclusion of social and organizational scientists as members of the game development team. Games as an Interactive Laboratory MMOGs are proprietary and written for a particular entertainment purpose, with rules very much unlike those found in real life. Players get to be heroes, villains, and superheroes in MMOGs and are often able to transport their virtual characters across large terrains without apparent cost in time or physics. So while it looks as if modelers might be able to do some experiments with MMOGs relevant to DoD concerns, there are definitely issues in the details. MMOGs as interactive laboratories are limited in their use for DoD purposes because they were built for entertainment. For MMOGs to become widely used in DoD, DoD may need to establish its own studio to build such an MMOG, using a mix of game industry veterans and defense M&S personnel. A vision for what this might look like is the collection of art resources and animations from the America’s Army game ported to a larger, more open platform (U.S. Army, 2007). Right now, America’s Army is built on the Epicgames Unreal-2 game engine, an engine limited to small squad-on-squad play (32 players total) and small areas of ­terrain (1 km × 1 km). In addition, access to the art resources and code from America’s Army has been restricted to DoD due to proprietary game engine license issues and close control of those resources by the Army project management team. Good small training systems have been built using the America’s Army material (see http://info.americasarmy.com), but in general, the close hold of the source code and game resources has made it difficult for DoD scientists desiring to use that material to be able to build the additions and extensions necessary to carry out their research. As DoD moves to the larger realm of MMOGs for social model experimentation and to a game engine capable of handling much larger terrains of concern, the openness and accessibility challenge needs to be solved for the greater DoD good. 

GAMES 269 Funding a specialized MMOG game studio outside the government would be one approach to the challenge, perhaps within the environment afforded by a specialized cross-departmental university center, such as the Entertainment Technology Center at Carnegie Mellon University (see http://www.etc.cmu.edu/) or as a university-affiliated research center but with more internal development capability than seen at labs like the highly successful Institute for Creative Technologies at the University of Southern California. There are many applications for which this MMOG will be of value. A particular MMOG that would have great applicability is one that implemented the various alternative futures as described in the report Map- ping the Global Future (National Intelligence Council, 2004). Understand- ing those potential outcomes and being able to roll back to a state in which the potential outcome does not happen would be a great tool for designing better national policies. References Bennington, R.W. (1995, May). Joint simulation system (JSIMS): An overview. In Proceedings of the IEEE 1995 National Aerospace and Electronics Conference (NAECON 1995) (pp. 804–809). Available: http://ieeexplore.ieee.org/iel3/3912/11342/00522029.pdf ­[accessed Feb. 2008]. Carley, K.M., Moon, I.-C., Schneider, M., and Shigiltchoff, O. (2005). Detailed analysis of factors affecting team success and failure in the America’s Army game: CASOS technical report. (Report #CMU-ISRI-05-120.) Pittsburgh, PA: Carnegie Mellon University. Available: http:// stinet.dtic.mil/cgi-bin/GetTRDoc?AD=ADA456079&Location=U2&doc=GetTRDoc.pdf [accessed Feb. 2008]. Electronic Arts Inc. (2006a) About the Sims 2. Available: http://thesims2.ea.com/ [accessed Feb. 2008]. Electronic Arts Inc. (2006b). Maxis: A timeline. Available: http://www.maxis.com/about/ about_timeline1.php [accessed April 2007]. Electronic Arts Inc. (2007). Inside scoop: The history of SimCity. Available: http://www.maxis. com/about/about_timeline1.php [accessed April 2007]. Moss, S. (2001). Game theory: Limitations and an alternative. Journal of Artificial Societies and Social Simulation, 4(2). Available: http://jasss.soc.surrey.ac.uk/4/2/2.html [accessed April 2008]. National Intelligence Council. (2004). Mapping the global future: Report of the National Intel- ligence Council’s 2020 project. Washington, DC: U.S. Government Printing Office. National Research Council. (1997). Modeling and simulation: Linking entertainment and ­ efense. Committee on Modeling and Simulation: Opportunities for Collaboration d B ­ etween the Defense and Entertainment Research Communities, Computer Science and Telecommunications Board. Commission on Physical Sciences, Mathematics, and Applica­ tions. Washington, DC: National Academy Press. U.S. Army. (2007). America’s Army: The official U.S. Army game. Available: http://www. americasarmy.com/ [accessed Feb. 2008]. von Neumann, J., and Morgenstern, O. (1944). Theory of games and economic behavior. Princeton, NJ: Princeton University Press. Wikipedia. Massively multiplayer online game. Available: http://en.wikipedia.org/wiki/MMOG [accessed April 2007]. 

270 BEHAVIORAL MODELING AND SIMULATION Zyda, M., Hiles, J., Rosenbaum, R., Roddy, K., Gagnon, T., and Boyd, M. (1998). ­SimNavy- Phase 0 building an enterprise model of the U.S. Navy. In Proceedings of the 1998 Tech­ nology Initiatives Game. Available: http://gamepipe.usc.edu/~zyda/pubs/SimNavyPaper98. pdf [accessed April 2008]. Zyda, M., Mayberry, A., McCree, J., and Davis, M. (2005). From Viz-Sim to VR to games: How we built a hit game-based simulation. In W.B. Rouse and K.R. Boff (Eds.), Organi- zational simulation (pp. 553–590). Hoboken, NJ: John Wiley & Sons.