Should Planners Start Playing Computer Games? Arguments from SimCity and Second Life

Abstract

In his book Self-Organization and the City, Juval Portugali describes how, in the past sixty years, our conceptualization of cities has shifted from a portrayal as isolated, stable and transparent systems, into open, self-organizing and complex systems, and how the role of the planner has turned from that of an engineer who masters these systems into that of, at best, simply observers and participants. Though Portugali's vision of the role of the planner might be too polemic, what is indeed an issue is that planners are in need of new tools that will enable them to be more process orientated, more geared towards communication, and encouraging of the involvement of multiple stakeholders, etc. Computer simulation models meet these requirements, but although they have indeed been adopted in planning practice, they have not become an indispensable piece of equipment. A very different story emerges from the gaming industry where, for more than twenty years, computer games have continued to attract millions of players, and have even entered classrooms and planning practices. Not only do these games look increasingly realistic but the behaviour of the characters within them has grown ever closer to resembling actual behaviour, to the extent that some commentators have even started to propose that planners should turn into gamers. Does this make sense, or is it still a bridge too far? In the search for an answer, this paper will critically analyse two computer games: SimCity, which is a classic strategy game, and Second Life, an increasingly popular social virtual world. I will make a number of suggestions as to how these games could be upgraded into planning simulation models.

Keywords:

• Planning simulation models
• strategy games
• social virtual worlds
• self-organizing systems
• gaming/simulations

Spatial Decision Making and Simulation Models

In his book Self-Organization and the City, Juval Portugali (2000) sketches a condensed history of what he refers to as prototype urbanisms, namely metaphors that capture how cities were perceived and imagined over the last century. He starts with Von Thünen's “isolated city state”, and ends with his own “hypermodern self-organizing city”. In the sixty years between these two prototypes, the conceptualization of the city has shifted from that of an isolated, stable and transparent system, into an open, self-organizing and complex system. What Portugali claims and illustrates is that, in spite of being open and complex, cities do exhibit essential features of self-organization (e.g. social segregation, congestion) or, as Jane Jacobs (1961) argued half a century earlier, cities are problems of organized complexity, systems in which the parts follow specific rules and, through their various interactions, create a distinct macro-behaviour, generating recognizable patterns or shapes (Johnson, 2001). Self-organization is one of the key concepts within complexity theory, which is the theory of complex systems, ranging from social insects, to brains, traffic jams, and cities. As the width of this range suggests, complexity theory is not so much a uniform theory as it is an enumeration of the features that all these systems have in common, such as a large number of interactions, simple behavioural rules, self-organization, co-evolution, emergent properties, etc. (Portugali, 2000).

The change in perception from an isolated city state to a hypermodern self-organizing city required a radically different approach to planning. To use the words of Portugali: “Cities are chaotic and unpredictable and they self-organize themselves independently of our scientific predictions and planning rules. All that is left for us to do, as scientists and planners, is to sit and watch, or at best become participants in this huge self-organizing process” (2000, p. 46); a shift from planners as technocrats to planners as spectators that has taken place in a time span of less than thirty years. But is the role of the planner really limited to observing and participating? Can processes not be steered even though they are chaotic and self-organizing? Perhaps it is not so much the role of the planner that needs to be reconsidered, but rather the tools that a planner relies on? Such tools should allow for a more process orientated type of planning, involving a variety of stakeholders, that is geared towards communication as well as experimentation, etc.

In this regard, Batty (2005) proposes reliance on simulation models, not as reproductions of physical systems (as in a scale model), but as artificial worlds that exhibit self-organizing features similar to those observed in real life. Similarly, Epstein and Axtell (1996, p. 4) advocate the development of artificial societies, “laboratories, where we attempt to ‘grow’ certain social structures in the computer, or in silicon, the aim being to discover fundamental local or micro mechanisms that are sufficient to generate the macroscopic social structures and collective behaviour of interest.” Such tools support an iterative, project-based type of planning, where a decision maker can, in an interactive fashion, assess the impact of a project by instantly observing the reaction of a virtual population to this project. This type of planning is iterative in the sense that, in principle, each reaction of the virtual population brings the project closer to what the planner has in mind. Batty (1976), paraphrasing Harris, sees a simulation model as an experimental design based on a theory, thereby implying that the development of a model is research in itself but, at the same time, it is nothing more than that. It is not an objective expert, generating indisputable solutions, but is just another decision support tool, engendering and structuring discussion and debate. Or, as Mitchel Resnick (1994, p. 50) pointed out: “The goal is not to simulate particular systems and processes in the world. The goal is to probe, challenge, and disrupt the way people think about systems and processes in general.” As such, simulation models invite the planner to again take up a central role in the planning process, and grow from being a mere spectator into a mediator, who no longer only observes or simply takes part, but actually becomes able to steer spatial processes in a particular direction.

Judging from the number of recent publications, conferences and research projects, the most popular techniques for the development of simulation models are cellular automata, agent-based systems and combinations of both (Batty, 2005). A cellular automaton consists of a grid of cells, with each cell representing an areal unit. Whether a cell changes state is defined by transition rules, the definitions of which are typically dependent on the state of adjacent or near-by cells. By adding random noise to the rules, surprisingly complex patterns, which closely resemble real cities, can be generated (Wegener, 2001). In the context of planning, an agent-based system can be said to be an artificial society that is inhabited by agents who represent individuals, make autonomous decisions guided by perceptions, preferences and habits, and who are constrained by a spatial setting, financial and temporal resources and limited knowledge (Devisch et al., 2005). From the interaction of these agents, real-life phenomena then emerge.

Cellular automata have been widely adopted within planning practice to model a variety of phenomena, such as: fire spreading (Ohgai et al., 2004), gentrification (Nara & Torrens, 2005), urban growth (Batty, 2005) and sprawl (Fang et al., 2005). Because of their straightforward implementation, cellular automata are very useful communication tools. By changing the grid lay-out, or tweaking the transition rules, a range of “what if” scenarios can be visualized and discussed.

Agent-based systems are slowly entering planning practice, and are addressing phenomena such as pedestrian flows (Batty et al., 2003), traffic congestion (Vogel & Nagel, 2005), shopping behaviour (Ali & Moulin, 2005), and location choice (Manson, 2005). Agent-based systems differ from cellular automata in that they explicitly model individual human behaviour, and address behavioural concepts such as pro-activeness, learning, joint decision making, negotiation and imitative behaviour. For this reason, agent-based systems have the potential to be more than just communication tools because, by modelling individual decision makers, they may help planners to understand (and thus potentially steer) the chaotic and self-organizing city envisaged by Portugali. In spite of their potential, most operational agent-based systems are, to date, still mainly developed as communication tools. They do model individual behaviour, but this behaviour bears hardly any resemblance to the highly complicated behaviour of actual citizens. Overall, although simulation models have entered planning practice, they have not become an indispensable piece of equipment.

Yet what seems to be a difficult “scientific” endeavour in planning appears to be child's play in the entertainment industry. Here, computer games make up a successful billion-dollar industry. A “complex” computer game like the Sims, which simulates the day-to-day activities of one or more virtual people (dubbed “Sims”) has sold more than 16 million copies since its release in 2000, making it the best-selling PC game in history (TMC Net, 2005). With its focus on individual behaviour, it can even be argued that the Sims is an agent-based system, albeit not one that is directly geared towards planners, as there is no explicit spatial component. Accordingly, what we will do in this paper is introduce two computer games in which this spatial component is present, namely SimCity and Second Life, and measure the extent to which these games could contribute to the actual development of planning simulation models.

In the next section, we will briefly introduce the reader to SimCity and Second Life and position them within sketched planning simulation models. In the third and fourth sections we will critically analyse both games and make suggestions about how to turn them into tools that could help planners to operate in Portugali's self-organizing city. In the concluding section, we will try to answer the question raised in the title as to whether or not planners should become full-time gamers.

Introducing SimCity and Second Life

In the search for how computer games could foster the development of planning simulation models, we will select two highly popular games with an explicit spatial component: SimCity and Second Life. SimCity is a so-called strategy game, wherein winning the game is a matter of skill. The archetype of a strategy game is chess. Computer simulation strategy games can range from war-games, to economic simulations, role-playing games, or, as is the case with SimCity, city-building games. Crucially, there is more than one way to play and win the game, and in order to improve the player has to “learn” the rules of the game. SimCity is predominantly a single-player game. By taking up the role of a mayor, for example, a player has to either build a new city from scratch, or has to manage an existing city through events such as a natural disaster or a nuclear power plant meltdown. To succeed, he can build homes, set local tax rates, construct a power grid, provide public transportation systems and so on. Ever since its first release in 1989 SimCity has been extremely popular, having sold more than 2 million copies in the US alone, “and has probably introduced more people to urban planning than any book ever has” (Starr, 1994, p. 1). Since the release of version four, SimCity has been populated by Sims. One Sim is equal to one individual, with his/her own features, skills and moods. Sims behave autonomously and react to their environment. Every simulated morning, for instance, they get in their cars and drive to work. When the traffic is too heavy, without requiring instructions from the mayor, they will try to find an alternative route. Since the introduction of Sims into SimCity, the game can be categorized as an agent-based system. The behaviour of the Sims is one of the indicators that the mayor can utilize as a way of gauging the success of his governance strategy. As an example, when a Sim has to spend too much time in traffic jams on his way to work, he may decide to move out of the city. When others follow his example, the mayor has a problem. As well as having to satisfy a population, the mayor may also face natural disasters such as: tornadoes, fires, or earthquakes; or super-natural attacks by monsters or extra-terrestrials. Because there is no preset goal or contest in SimCity, the developers refer to it not as a game but rather as a “toy”. It is the player who decides what kind of city to build, and whether to emphasize its size, wealth, beauty, or its harmony with the environment (Starr, 1994).

Our second computer game, Second Life, is a so-called social virtual world. It differs from a strategy game like SimCity in that its aim is not necessarily to win or even to play, but rather to socialize. Such a virtual world really functions more as large-scale online community that uses elements of gaming in the service of its more significant goal of developing a society (Book, 2004). For this reason, social virtual worlds are typically based on everyday environments, and contain cafés, galleries, movie theatres, concert halls, as well as tropical islands, tourist attractions, theme parks, shopping malls, flea markets, etc. Many social virtual worlds include commercial activities: residents own property and run business ventures that generate virtual or, in some cases, even real income for them (Book, 2004). Second Life was developed by the Linden Lab and made public in 2003. Since late 2006, it has become immensely popular and reached a total population of 3,289,433 registered residents on 4 February 2007. Second Life differs from other virtual worlds, such as role-playing games like World of Warcraft, in that it is completely reliant on player-generated content. Initially, Second Life was empty because all content comes from the residents. A resident of Second Life is known as an avatar, which is a three-dimensional figure visible to the player. Each avatar can construct objects, ranging from shoes, to buildings, to even whole islands. These objects then form the setting for the meeting of other residents, socializing, participating in or organizing events, and shopping. Second Life's virtual currency is known as Linden Dollars, and is exchangeable for US Dollars. As of 4 February 2007, around one million US dollars were spent every 24 hours. What makes Second Life even more unique in game land is the lack of game rules. Avatars simply meet, form communities, and when necessary, draw-up their own “community contracts”. The success of Second Life also brought its first critique, which concentrated mainly on the number of residents: around 85% of the registered users would only visit the virtual world once (Shirky, 2007): one try, then bye-bye. However, since January 2007, the number of avatars running around at any particular time seems to follow a regular daily pattern, reaching an approximate minimum of 12,000 (around 11 a.m. Universal Time Coordinated (UTC)) and maximum of 25,000 (around 10 p.m. UTC).

Although SimCity and Second Life are intrinsically different (i.e. SimCity is a strategy game whereas Second Life is a social virtual world), they do have a number of features in common, two of which are especially relevant to this paper. Firstly, both could be categorized as being artificial societies, similar to those developed by Epstein and Axtell (1996). This is particularly evident in the case of Second Life, in that everything in this virtual world has to emerge from the (inter)actions of individual avatars. Likewise, as we have argued previously, since the release of version four, SimCity is in fact an agent-based system, within which Sims interact and react autonomously to the interventions of the mayor, thereby generating a range of interdependent phenomena, and as such making up an artificial society. It is important to mention here that the behaviour of the Sims and the impact of the mayoral interventions are largely programmed into the game-code, so there is a limit to the emerging complexity. A mayor can actually learn the underlying rules, and master this complexity. This is a possibility that the developers deny, claiming that the incorporated variety guarantees that a player will never be a mayor of the same city twice (Glean, 2005).

A second trait that both games have in common is their educational appeal. In a recent paper, Daniel Lobo (2005) emphasized that SimCity is “not just a game”, and pointed out how it is used, for example, to generate interest in geographical information software, or to teach twentieth-century local government. He noted that the game's developers explicitly consider the educational market to be one of their key audiences, resulting in teachers' Sim guides, Sim school licences, and special Sim kids' products. As Nicholas Johnson (2006) pointed out, Second Life is also promoting itself as being educational by providing, amongst other things, free trials for educators in which they are offered a virtual plot of land on which students can build something for one semester, free of charge. Educational institutions do seem to have taken up the challenge. According to Second Life's developers (2007), around 85 universities, colleges and schools worldwide have reconstructed part of their campuses in Second Life. The Harvard Extension School, for example, offers a distance education course in Law at the Court of Public Opinion, where students' avatars meet once a week to engage in group-work (CyberOne, 2006). Moreover, fourth-year architecture students at the Royal Institute of Technology in Stockholm have opened what they claim to be the world's largest virtual architects' office “LOL architects”, as part of a course, “Production of architecture”, which examines the history, theory and practice of representation and production of architecture (Lindstrand, 2006).

Because of this educational appeal, both games have even been suggested as planning support tools, and more precisely as communication instruments. SimCity, for instance, has been employed to generate interest in planning activities, and to raise awareness about how planning decisions are interconnected (Lobo, 2005). Accordingly, in the rest of this paper, we will take this suggestion seriously and try to imagine how we could turn computer games into an indispensable piece of equipment in planning practices.

It is important to note here that the idea of using these games in planning was clearly not in the minds of the developers. In an interview with Starr (1994), to the question of how he decides what to incorporate in SimCity and what to leave out, the developer of the game, Will Wright, replied, “We go for game play. Whatever is most fun.” Such a response caused Starr to categorize these games as “edutainment”; entertainment designed to educate as well as to amuse. Or, as Daniel Lobo (2005, p.15) indicated “Forgetting the playfulness of SimCity in the classroom would be equal to teaching civic behaviour with fighting games such as Mortal Combat, Quake or Tekten IV.”

Attempts to convert planners into gamers have been made before (Starr, 1994). However, what we are proposing here is different in that we would like to plead for a more systematic and critical evaluation of computer games. To this end, we will adopt an evaluation method that was proposed by Clarke (2003), who claimed that planning models in particular, and computer models in general, can be said to follow a distinctive format, consisting of four generally repeating model components:

(1) input, both of data and parameters, often forming initial conditions; (2) algorithms, usually formulas, heuristics, or programs that operate on the data, apply rules, enforce limits and conditions, etc.; (3) assumptions, representing constraints placed on the data and algorithms or simplifications of the conditions under which the algorithms operate; and (4) outputs, both of data (the results or forecasts) and of model performance such as goodness of fit (Clarke 2003, p. 2).

In the sections that follow we will analyse both SimCity and Second Life against these four components, each time concluding with a number of suggestions about how to upgrade the game into a planning simulation model. In fact, Clarke also mentioned a fifth component, namely that of the modeller including “their knowledge, specific purpose, level of use, sophistication, and ethics.” This component refers to the act of modelling itself and, for this reason, is not addressed.

How to Turn SimCity into a Planning Simulation Model

Input. SimCity grows by patches, which are plots of land of about 15 by 15 m². One patch generally houses one program that is detailed to the level of containing lamp-posts, outdoor swimming pools, garbage cans, and even graffiti. During the game, the player can only choose from a database of ready-made patches, and is unable to add or remove detail. What the player can do, however, is design new patches which he or she can later import into his/her city via the database (see for instance the SimCity 4 fan site: Simtropolis). The accumulation of patches gives the SimCities a collage-like feel, wherein each individual clipping might resemble existing landmarks, or parts of real cities, but where the final result more often than not is a fictional city. This was not the case in the original releases, in which the purpose was to re-create the general layout of historical cities, often in trying circumstances such as having experienced a major disaster. Examples are the city of Bern, which in 1965 suffered from extreme traffic congestion, Detroit in 1972 was wrecked by crime and depressed industry, and San Francisco in 1906 was completely destroyed by an earthquake.

SimCity is populated by Sims. In contrast to the actual Sims game, SimCity's Sims are merely in the background, and are only there to provide feedback to the mayor. SimCity 4 comes with 21 default Sims. The player only has an indirect say in the composition of his population: constructing more single-family houses will, for instance, attract more middle-class Sims.

Both the patches and the Sims react to the interventions, or the lack thereof, by the mayor. Interventions can range from the very general, such as setting the tax rates, to very detailed, such as providing an extra zebra crossing, or regulations that set a curfew for youngsters, or legalize gambling. Dreesen (2006) describes an experiment in which she, as mayor, wanted to measure the effect of a series of regulations on the amount of traffic in her SimCity. She decided to subsidize carpooling, set up a shuttle service for commuters, and taxed excessive production of exhaust fumes. After a simulation period of ten years, Dreesen concluded from her SimCity graphs that the regulations had no impact on traffic volume as such, but that they did result in decreased air pollution, albeit at an additional cost to the government.

In order for SimCity to move from being a game to a planning simulation model, the input needs to completely change from (collections of) patches into individual parcels, and from a quasi-homogenous background population into individual autonomous actors. So far as the power of the mayor is concerned, the variety of traffic-related interventions that are included in the SimCity Rush Hour package should be extended to other domains, such as housing, greenery, etc. As previously mentioned, the developers of SimCity see the game first and foremost as entertainment. Given that in reality planning is not exactly interactive, in that it generally takes years before one is able to assess the impact of a decision, the developers of the game have dramatized the behaviour of the Sims, and deliberately exaggerated the effects of activities, in order to provide feedback to the mayor. There is a less artful, and altogether more practical explanation for this exaggerated behaviour, which is coarse and involves simplified actions. More subtle behaviour would need the Sims to be able to perceive, memorize and react to small changes in their environment, which would require an enormous modelling effort and computation capacity. This is a path that the developers chose not to follow. As an illustration of the exaggerated (i.e. simplified) behaviour referred to, Diane Carr (2004) points at the reductive and low resolution maps that are included to measure the impact of a policy that a mayor chooses to follow: maps of aura, of density, crime or property value. By being selective about what to include, games evoke a suspension of disbelief that can enrich the pedagogical value of the experience (Carr, 2004). Similar considerations have been made about simulation models in general. Swartout and Lindheim (2002), for instance, examined the use of flight simulators in military training. Such simulators are typically almost exact replicas of the real thing, complete with all the necessary switches, knobs, gauges and controls. What Swartout and Lindheim found is that, because of this great amount of detail, even experienced pilots are no longer able to differentiate between the real and the virtual object. This led to the authors concluding that simulation can be too detailed, to the point that it becomes counterproductive. They give the example of an air speed indicator in an aircraft simulator which has a slightly different speed indication than in the real aircraft. Due to this slight difference, pilots seem to continuously overspeed when flying the real aircraft, while not being aware of it (i.e. the speed indicator has exactly the same position as in the simulator but in fact indicates another speed).

Algorithms

SimCity is designed as a cellular automaton, populated with agents. As mentioned earlier, a cellular automaton is basically a grid of cells (or patches) wherein the state of a cell changes depending on the states of the neighbouring cells. This dependency is defined in transition rules. Being a game, these rules are hidden from the player, and the game's purpose is for the player to unravel them through endless game-play, or in the words of Will Wright: “Playing the game is the process of discovering how the model works” (Starr, 1994, p. 7). From the moment that the game's rules are known, the game becomes predictable. At this point, the player cannot get any better, thereby removing the reason for playing.

Transition rules are deterministic: an increase in the value of a plot, for example, will always raise the value of neighbouring plots. The behaviour of the Sims is also deterministic. For instance, on their way to work, Sims will always take the shortest route, until they encounter a traffic jam. Then, the next day they will try out an alternative route (Peck, 2007). All this is programmed behaviour, which is scripted into the game-code. To take another example, when the Sims are not working, they might decide to just go for a drive, with absolutely no idea of where they are going (Jordan, 2007). At such a time, they are in the power of so-called attractors and repulsors. Attractors are objects that seduce the Sim to come closer, whereas repulsors make the Sim continue in the opposite direction. Just about any object in the game can have a set of attraction and repulsion properties attached to it. Schools, for instance, attract children in the morning, and repulse them when the school day is over. Landmark buildings make tourists line up, newly erected buildings will attract passers-by, and disasters and fires will make them run away, screaming. Or, as Jordan summarized, the Sims will have a reaction to everything you do.

In order for SimCity to move from being just a game to a planning simulation model, the behaviour of the Sims needs to become more realistic, and include behavioural concepts ranging from pro-activeness, joint decision making, bargaining and imitating, to more abstract concepts such as trust, intuition, commitment, friendship, and so on. According to Starr (1994), the purpose of SimCity is not accuracy or prediction but communication. If the game did indeed become more behaviourally realistic (and thus unpredictable), its development would seem random, and players would not be able to learn, and thus improve at playing it. In this context, one recalls Jane Jacobs referring to cities as problems of organized complexity, implying that although the behaviour of a single individual might appear random, the aggregate behaviour of thousands of individuals might indeed display regularities. Households are an example in that they mainly seem to move within the current housing market (Dieleman, 2001). But such regularities are generally too abstract to be of interest to a player. It is for this reason that a Sim can never behave impulsively, but is bound to execute scripts.

Another requirement is transparency. In order to engender debate, one needs to know what is being debated. So, rather than investing in the development of exact replicas, a modeller should instead invest in the clarification of the assumptions and algorithms underlying his model. As argued earlier, SimCity is deliberately designed as a black box, although the developers are planning to publish an open-source version that will allow players to not only design patches, but to also tweak the code, thereby manipulating the rules of the game.

Assumptions

As has already been pointed out, players accept the fictitious nature of SimCity because it increases their playing experiences. For these players, whether the assumptions regarding city growth, spatial behaviour, interaction between Sims, etc. are realistic or not is not really relevant. The point of playing is to get to know the game, whatever the assumptions. The degree to which SimCity is representative does however become an issue when the game is being used in an educational context. Although the game's assumptions are never referred to explicitly, they can be reconstructed on the basis of interviews and hours of game-play. In one interview, Will Wright refers to the basic conceptual framework of SimCity as a “capitalistic, land-value ecology” and argues, that while it fits the development of American cities in the twentieth century, it would not account for the development of St Petersburg. According to Starr (1994), SimCity is somewhat more constrained, in that the game seems to require a particular type of American city built on an industrial base. Lobo (2004) claims that the assumptions have nothing to do with a particular conceptual framework, but rather with aesthetics. A city composed of segregated patches is simply more pleasing to the eye than a city composed of mixed-use patches. In the same way, homogenous class-segregated neighbourhoods are favoured over socially mixed neighbourhoods, and grid structures are favoured over organic development.

In order for SimCity to move from being a game to a planning simulation model, assumptions must obviously correspond to real processes and relationships. Carr (2004) points out that SimCity is not a simulation of a city, but rather a simulation of a fiction of a city. One could, therefore, claim that SimCity in its current version is in fact modelling the world from the viewpoint of the planner, whereas it should instead simulate the spatial behaviour of the true actors who drive urban processes, namely the citizens. But even then, Presky (2002, p. 3) argues that “all of the calculation models underlying simulations reflect totally the assumptions and biases of the designer(s), in terms of how they choose, relate and weight the various inputs.”

Output

Since the release of SimCity 4, a player can not only observe the city from a helicopter view, but by changing the game-mode, he or she can also roam through the streets, and experience everything through the eyes of a Sim. In this way, the mayor-player can keep an eye on everything that is happening within his or her realm. Furthermore, the mayor can rely on a wealth of constantly changing data, in the form of maps and graphs, that show the city's population growth and density, the demand for residential, commercial and industrial land, unemployment levels, information about the power and water supplies, crime, traffic congestion, pollution, and various other aspects of the city's development. The same sources also report changes in interest rates and the growth of the national economy and neighbouring cities. Newspapers periodically deliver reports of local sentiment, including the latest public opinion polls and inane, jumbled stories about local and made-up international events (Starr, 1994).

In order for SimCity to turn from a game into a planning simulation model, an information filter needs to be provided. The wealth of available data is, in the main, deceptive, for there is a lack of hierarchy in that all buildings are equally detailed, and all graphs seem to be equally important. This wrongly gives the data an aura of objectivity, making it tempting for the user to employ the game as a neutral validation instrument, rather than as a (heavily biased) experimentation tool.

Overall, SimCity has the potential to evolve into a relevant planning simulation model. It is operational, comprehensive, and evocative. At the same time, as the review above indicates, it lacks a number of essential features: it is not completely disaggregated, behaviour is deterministic, the number of incorporated behavioural concepts is limited, and the code is a black box. As already hinted at, the open-source version of SimCity might address some of these issues, potentially increasing the number of incorporated behavioural concepts, thereby allowing for more realistic Sim-behaviour. Such a model would allow the planner, in the role of mayor, to select virtually any city and reproduce it as a SimCity, and to select virtually any population and reproduce it as Sims. In such circumstances, the purpose of the game would no longer be to deal with UFOs or to recover from earthquakes, but to take an actual city with an actual population, and assess the impact of actual planning interventions, ranging from alternative traffic schemes, to re-development projects and building legislation.

What would, however, continue to be problematic is that behaviour would still be scripted. In other words, in spite of the increased number of incorporated behavioural concepts, Sims would still be incapable of unprepared strategies. Perhaps they can take a shorter route to work if there is congestion, but they would never buy a run-down house in an up-and-coming area, decorate it and sell at a profit. Indeed, although major advances have been made in the modelling of behavioural concepts such as learning, joint decision making, bargaining and pro-activeness, hardly any advances have been made when it comes to modelling more abstract concepts such as trust and friendship.

Finally, SimCity also exists as a multi-player online game, during which a city can be shared online by passing it from mayor to mayor, for a limited period of time. That is, you borrow a city, see where it is up to, push its development in a direction of your choosing, and then leave it when your term expires. In considering the idea of employing SimCity as a planning tool, Daniel Lobo (2005, p. 16) proposed letting the players work together, managing the city as a team, rather than making them compete as rival mayors. “This development, in addition to an access to the black box, would open new ways to explore conceptual city scenarios that, with the right framework, could start at SimCity helping us to understand urban environments better.”

How to Turn Second Life into a Planning Simulation Model

Two things should be noted before undertaking our evaluation. As is the case with SimCity, Second Life is not in fact a true game because a player cannot win or lose. Furthermore, Second Life is not actually even a model because it is basically a piece of communication software, and nothing more. Yet, this software does allow players to replicate themselves as three-dimensional avatars, construct imaginary worlds, and synchronize interactions with fellow avatars. It could be argued that the emerging world is a simulation, albeit one that is continuously under construction and authored by more than a million developers. Because it is not a true model, not all of the components proposed by Clarke are relevant. Second Life, for instance, does not rely on assumptions about spatial behaviour, as these all come from the players. Given that our objective is to compare two computer games, and to suggest game modifications, we will attempt to address all components in any event.

At the time of writing, Second Life is being fantasized about in almost any newspaper or magazine, wherein the endless possibilities of this virtual world are debated. The tech magazine Wired, for example, reported as long ago as October 2006 that “an increasing number of residents are ditching their jobs back on Earth to make their living entirely within Second Life's economy.” Catherine Winters (2007), on the Social Signal Blog, predicted that the population of Second Life could well reach up to thirty million in a year from now. A more critical evaluation is only just beginning to take shape (for an online debate, see Coleman et al., 2007). This current enthusiasm may well have coloured the following arguments.

Input

Second Life started off as an empty world, nothing more than one gigantic ocean. Avatars buy islands and develop them. Currently, islands are priced at US$255 per hectare. Avatars move around in this archipelago by flying from island to island, or by teleporting directly to their favourite event. Although there are no limitations to what avatars can do, most of the activities seem to revolve around reconstructing real-life cultural forms and re-enacting real-life events. One of the most popular things to do is to attend live music concerts. A quick look at the activities organized in Second Life on 9 February 2007 reveals that there was a big sale for ladies at 8.00 a.m., French elegance coaching at 10.00 a.m., a concert by Louis Landon at 10.30 a.m., an aliens' exhibition at 11.30 a.m., and so on. As the developers of Second Life love to proclaim, everything is possible, the only limit is your own imagination and, of course, the number on your credit card because hardly anything in this virtual world is free. In case your technical skills slow down your imagination, companies actually exist that are dedicated to designing experiences and delivering add-on software for 3D virtual worlds. The Electric Sheep Company helped Reuters to open up a branch in Second Life, worked with Major League Baseball to bring the 2006 Home Run Derby to avatar baseball fans, and combined forces with Starwood Hotels to virtually promote their new hotel line, which will not open in the offline world until 2008.

In order for Second Life to turn from a game into a planning simulation model, the necessary requirements seem to be met. Second Life is completely disaggregated, and the amount of detail is dependent only on the players themselves. It would, for example, be technically possible to reconstruct a complete city in great detail in Second Life, and experiment, for example, with alternative street lay-outs. Technically, this works against the philosophy of Second Life, which revolves around communication requiring places, not complete cities, and requires single events, not complete 24-hour cycles. Avatars do not take their cars to work, but teleport from, for example, an exclusive shoe sale to a meeting commemorating the life of Anna Nicole Smith. Apart from it being a fragmented environment, both spatially and temporally, it can be argued that unlike SimCity, Second Life is an abstraction of reality rather than a simplification of it. This implies that although Second Life is obviously a low-resolution (i.e. simplified) representation of the offline world, it does respect the multiplicity of both objects and object interactions in reality. This is not the case with SimCity, in which not only the resolution of the offline objects is reduced, but likewise the number of object interactions, thereby simulating a fiction of reality, rather than reality itself.

Algorithms

In principle, avatar behaviour is directed by algorithms, but also by real people interacting with a keyboard. SimCity, on the other hand, depends on nothing but algorithms: objects deteriorate, attract or repulse Sims, catch fire, or become radioactive, and Sims go to work, move house, or go shopping, all executing hundreds of scripts that are inaccessible to the player. Not so in Second Life, although some players do equip their objects with a certain level of intelligence, such as developing a room that increases in size as the number of guests increases (Maher & Merrick, 2005), or a statue that plays a message as you get closer. Yet the main point remains the stimulation of interaction, not the simulation (and thus scripting) of behaviour.

Players are represented in Second Life as three-dimensional avatars. A player can personalize his or her avatar by tweaking up to two hundred settings (for instance you can even change the size, shape or pattern of your avatar's iris and pupil) and you can decorate him or her with attributes ranging from clothes and jewellery to tattoos, and even horns, tails or wings. As Book (2004, p. 8) observed, “Many events and activities in social worlds revolve specifically around avatars, in the form of avatar customization classes, clothing sales, costume contests, modelling contests, and fashion shows.”

Although most players visit Second Life to meet soul mates, the virtual world is also a place of experimentation; “A place where commercial, educational, non-profit, governmental, and amateur groups co-exist and interact. It is a playground where we can try on new identities; test new products and practices; explore new ways that core institutions might operate” (Jenkins, 2007). All of this operates within certain boundaries, because as the developers quickly realized, even a virtual world is unable to function without agreements, resulting in the production of community standards and behavioural guidelines. Guideline number six, for example, “Disturbing the peace” determines that avatars cannot disrupt scheduled events, repeatedly transmit undesired advertising content, use repetitive sounds, generate self-spawning items, etc. Misbehaving avatars can be suspended, or even excluded from the Second Life Community. As well as general guidelines, individual avatars can also draw up rules when they secure the objects that they have created, or the land that they own. So, an avatar can forbid others to trespass on his property (Johnson, 2006). It should however be noted that although these rules guide, they do not define behaviour.

In order for Second Life to become a planning simulation model, the necessary requirements again seem to be met. Avatar behaviour is guided via the keyboard and is not (only) scripted into rules. Accordingly, avatar behaviour is as non-deterministic and transparent as the behaviour of the individual manipulating the keyboard.

Assumptions

The main ambition of Second Life's developers is to provide optimal settings for synchronized communication. No claim is made about the simulation of real world phenomena. It is therefore more relevant to speak of constraints rather than assumptions. Constraints in Second Life are mainly due to technological limitations rather than deliberate choices. Land, for instance, is scarce, amounting only to the storage capacity allowed by the server. In theory, however, there is no end to Second Life: there is always more land, and there are always more resources. Another constraint is that currently, behaviour is limited to a number of scripted gestures, such as smiling, laughing, winking, shrugging and sticking out your tongue. Yet it can surely be anticipated that with the ongoing evolution in human-computer interaction, this will not be an issue in the near future. “It's not like real life. Not by a long shot. One is animating a proxy through multi-layered terrains of information … but the procession towards ever more complex simulation in computing is there. Not every user can code, but certainly more users will learn to script (or edit video or stream media) as Flickr and Youtube have made clear” (Coleman, in Jenkins, 2007).

Yet, in order for Second Life to turn into a planning simulation model, solving these technological constraints might not be enough. Even though individual virtual behaviour may start to represent real-life behaviour, the overall virtual world will remain a caricature. De Nood and Attema (2006) point out that the avatar population in Second Life is not representative of the offline population because people with IT-related jobs are over-represented. A similar observation no doubt also holds for age, gender and race. This online offline discrepancy is even more obvious if the roles that avatars adopt are looked at. There are no avatars with Tourette's syndrome to give just one example. What is, however, remarkable is that this lack of resemblance to real life does not tend to apply to individual avatars. It might be expected that the weirdest creatures would be walking around, but there seems to be an unwritten law in social virtual worlds, that avatars should remain at least somewhat faithful to their owners' offline appearances. Of course, one never knows whether an online (romantic) relationship might at some point extend to the offline world (Book, 2004). In this context, the term somewhat faithful has to be understood as being faithful to the owner's appearance, albeit with more exaggerated gender-related features.

Jenkins (2007) makes a similar observation when it comes to the formation of social networks. “If you look at the rise of social tech amongst young people, it's not about divorcing the physical to live digitally. MySpace has more to do with offline structures of sociality than it has to do with virtuality. People are modelling their offline social network; the digital is complementing (and complicating) the physical. In an environment where anyone could socialize with anyone, they don't. They socialize with the people who validate them in meatspace [sic].” As in real life, there seems to be little cross-group interaction. Having your avatar wear a more muscular body does not seem to be enough.

A final point at which the virtual seems to come close to the reality is the demand for stability. Coleman (2007) claims that a civilizing mission is going on in Second Life that is not directed by the developers but by the avatars themselves. De Nood and Attema (2006) point out that one of the recurring complaints among avatars is the lack of a regulating authority to penalize negative acts by locking up fellow avatars, for things like stalking, gossiping, and insulting others. They also refer to a call for monetary stability in the form of a financial institution, independent from the Linden Lab developers, that would control and regulate all financial transactions. This stability is necessary, they claim, to attract more businesses and banks to Second Life.

In spite of these similarities, it is important to note that it is not the ambition of the developers of Second Life to even come close to the offline world: “The goal of a virtual world is not perfection of simulation but augmenting our channels of communication” (Coleman et al., 2007). And this is where it differs from a computer game like SimCity. Both might be interpreted as simulations of real-world phenomena but in Second Life phenomena emerge, whereas in SimCity they are programmed into the virtual world. It is for this reason that it is also safe to say that Second Life does not favour a particular planning approach over another, nor would it be one that is shared by all avatars.

Output

As already argued, Second Life is basically just communication software. As with any messaging program, you type in your line, and wait for another to reply. Depending on the speed of your computer and your network connection, reactions are instant. The difference to common chat programs is that in Second Life you can see your conversation partner, and you can comment on what is going on around both of you. When your avatar is in walk mode, your view is limited to the direct environment. By switching to fly mode, your view increases, even though it is at the cost of a loss in detail. As well as walking and flying, you can also consult maps and lists of events and popular places. In contrast to the mayor in SimCity, you will never get to know, or be able to observe, the total virtual world because not all events are advertised, because most events require an entrance fee, or because you do not have permission to enter a place. So if someone wants to operate as a planner in Second Life, they would have to convince other avatars of their mission, and persuade each of them to follow their advice.

The hierarchy requirement that is necessary for Second Life to turn from a game into a planning simulation model is met. The criterion is not to create a complete and consistent virtual world, but rather a collection of singular places, each developed with a particular purpose in mind. As a result, Second Life will not be perceived as a neutral validation tool, but indeed as an experimentation tool. And a valuable one, considering the fact that everything that takes place in this virtual world is digitally recorded and stored.

Overall, as with SimCity, Second Life has the potential to evolve into a relevant planning simulation model. Avatars represent offline decision makers, and as first observations indicate, behave “realistically”. Moreover, there are no hidden assumptions related to planning and there is no black box. As pointed out, current media attention makes it difficult to come to any discussion regarding the planning potential of Second Life with a fully considered point of view. So, despite the risk of stating the obvious, we will consider two very different future scenarios for Second Life: one in which we give in to the worship and are indeed convinced that the virtual population will reach thirty million in a year from now, and another, more pessimistic scenario, where we proclaim that Second Life will remain a niche application.

Scholars devoted to the first scenario claim that virtual worlds like Second Life will, in time, take over the Internet, remodelling it into a three-dimensional world in which we shop, meet people and search for information as in the offline world. Coleman et al. (2007) predicts, in this respect, that virtual shopping will in the long run outdo electronic shopping. From this viewpoint it is not difficult to imagine a planner or an architect realizing his projects first in this virtual world, then inviting future residents for a virtual “tour around the house”, addressing remarks and suggestions as they come up. In this scenario, virtual worlds embody the ideal simulation model: a parallel world that is identical to the real world. This scenario is not that far-fetched, since the first attempts to do so are being undertaken. Residents of the borough of Queens, NY, for instance, are being invited in Second Life to communally design their new neighbourhood's park (Pfaffendorf, 2006).

In the second scenario, social virtual worlds will remain a niche application, with a stable population well below a hundred thousand, about the size of a small town. In such a scenario, replicating whole projects in silico as a crash-test for the real thing does not seem to make much sense. The value of such a “world of geeks” lies more in small-scale experiments. This potential is, as De Nood and Attema (2006) point out, already being exploited by companies such as Coca Cola, Toyota and Adidas, who are expanding their publicity campaigns in Second Life. They argue that for some designers, virtual worlds offer a way to get feedback, or to promote their latest compositions.

Authors like Jenkins (2007) point more to the potential to conduct social experiments, claiming that the dependence on user-generated content turns this world into one continuing experiment in participatory culture. “What's striking to me is … that a significant percentage of people are willing to generate and share content they produced themselves.” Content here not only refers to objects, but also to all kinds of social forms that are emerging out of so-called community contracts. What is particularly interesting to planners about such virtual communities is the value of concepts like public or private, identity, authenticity, representation, etc. In this respect, it is interesting that part of Second Life's code has been made public as open source. This has spawned a range of applications, one of which allows the user to automatically conduct surveys in Second Life (Menti, 2007). Since visitors usually go there to chat, a request to complete a questionnaire is hardly ever turned down. Regardless of the scenario, Second Life will, due to its nature, remain a fragmented world that people will only visit with a particular agenda, namely one that is typically devoted to socializing, shopping, entertainment, and, to a lesser extent, education. The Guardian (Jeffries, 2006) puts it more crudely: one third of the avatars go to Second Life to do business, the other two thirds go there for sex. Since any experiment depends on the seriousness of the attitude of those taking part, the employment of Second Life as a planning support tool will therefore mainly be relevant in small-scale, well-defined planning contexts, such as the communal design of a park, the testing of a new shopping concept and intervention in an urban block. In such circumstances, people are invited to explicitly engage in the experiment rather than there being a reliance on random avatars that are, most likely, just looking for their next date.

So, Should Planners Really Become Gamers?

As to whether or not it makes sense to persuade planners to engage in SimCity or Second Life, we conclude that it does not, at least not in their current forms. But, what if these games were upgraded along the lines we have suggested, namely by the Sims behaving more realistically (and also acting proactively, negotiating, forming social networks, etc.), and by convincing Second Life's avatars to seriously engage in (planning) experiments? In other words, what if these games could truly model cities as complex systems, exhibiting the features that were pointed to in the introduction, such as co-evolution and self-organization? Would these games then allow for a more process orientated type of planning, involving a variety of stakeholders geared towards communication as well as experimentation, thereby allowing planners to again take up a central role in the planning process, and providing them with a tool to steer spatial processes?

An attempt by Batty et al. (2004) to define the role of visualization in spatial modelling can help to answer this question. Batty et al. came up with three so-called roles for spatial (simulation) models: firstly, to inform the user about a given process; secondly, to assist the user throughout a process; and thirdly, to help the user understand a process. In order to find out how games can advance planning simulation models, let us consider how good both SimCity and Second Life (i.e. in their hypothetically upgraded versions) are at playing these three roles.

Models to inform the user in fact do nothing more than communicate a given message. This communication can be interactive and even consist of a range of media techniques, but the content of the message remains fixed. Both SimCity and Second Life obviously fulfil this role. SimCity is especially appropriate as a communication tool because of the high level of detail and because of the range of evaluation techniques such as graphs and tables that are included. What makes Second Life especially suitable is the potential to discuss included data with fellow avatars. Andy Hudson-Smith et al. from Centre of Advanced Spatial Analysis (CASA) (2007), for instance, developed a piece of software, Google Map Creator, to visualize and share geographical data. These visualizations are then exhibited in Second Life.

Models that assist users during a planning process allow, for instance, the evaluation of differing scenarios by the visualization of the impact of planning decisions. Again, both SimCity and Second Life can be employed to perform such impact analysis. It is, however, important to mention that however detailed and realistic Sim behaviour can be, simplifications and assumptions will remain, of which the planner must be aware and learn about while “playing”. When such a planner finally reaches the point of mastering the simulation, he may have also gained new insight that he wants to implement into the simulation, or he may want to run similar experiments, in other settings, as a form of validation. Yet both of these options require the development of a new model, because SimCity can currently only be employed to perform scenario evaluations within the assumptions upon which the game has been developed. So far as Second Life is concerned, we have argued that whether the running of a particular scenario makes sense or not, depends on how seriously the avatars engage with it. Directly inviting people to participate as in the case of the borough of Queens' neighbourhood park design might circumvent this issue.

Finally, models that help a user to understand (previously unknown) processes should stimulate experimentation with the system's input, in search of patterns and regularities in the system's output. Linking input to output might then give insight into the (self-organizing) processes that are taking place in the system under investigation. The usefulness of SimCity as a tool with which to understand the processes governing a system depends heavily on how realistic the behaviour of the Sims is. As pointed out earlier, significant progress has been made in the modelling of behavioural concepts such as learning, joint decision making, bargaining, and pro-activeness, but there have been very few advances when it comes to modelling more abstract concepts such as trust and friendship. The number and type of self-organizing processes that can be researched using (the upgraded version of) SimCity is therefore limited. In Second Life, avatar behaviour is, in principle, realistic since actual people direct it. As such, it should indeed be possible to experiment with Second Life and gain insight into not only individual spatial behaviour, such as how people use, perceive and conceive a particular place, but also insight into how these individual behaviours amount to (spatial) trends. In conclusion, what SimCity and Second Life cannot do in their current form, they may be able do in an upgraded form, thereby helping the planner to operate in the chaotic, unpredictable and self-organizing city of Portugali.

We would like to end with a suggestion made by Richard Duke, one of the pioneers of the development of urban planning games. In his latest book, Policy Games for Strategic Management, Duke defines a simulation as “a conscious endeavour to reproduce the central characteristics of a system in order to understand, experiment with and/or predict the behaviour of a system” (2004, p. 36). In his opinion, a game differs from a simulation in that the players are the central point in a game, whereas in a simulation, it is the computer. Since the sixties, Duke has combined games and simulations in so-called “gaming/simulations”, in which real-life role-playing games are supported by computer simulations. In doing so, his argument is that a game in itself is too abstract to address specific planning problems, and a (large scale) simulation is too integrated and data-hungry to be applied to an unfamiliar situation. Gaming/simulations, he claims, lie somewhere in between, relying on models to predicate known facts and on actor intuition to answer unexpected events. So, in adopting Duke's approach, for a computer game to be able to seduce planners to become gamers, it must at least evolve into gaming/simulation.

If we look at SimCity and Second Life from this perspective, then SimCity would clearly take up the role of the computer simulation, and Second Life would function as the role-playing game. Taken together, the result could be either a SimCity populated with and run by avatars, or a Second Life with a mayor, councillors and planners. What both games would not be able to achieve on their own, convincing planners to become gamers, they may be able to achieve in tandem.

Acknowledgements

I would like to thank the four anonymous reviewers for their helpful and sound advice. A draft version of this paper was presented at the 5th AESOP Complexity and Planning meeting in Stuttgart, Germany, March 9-10, 2007.