Shawn Graham Carleton University

Archaeologists have been simulating past societies via computation for decades (cf Wurzer et al 2015; Costopoulos and Lake 2010 for recent overviews). It is nothing new for us to perform a kind of practical necromancy to raise the dead to see what they can tell us. Archaeogaming introduces a new actor into these artificial societies: living humans. There are dangers to guard against, and opportunities to seize, when we co-write the past with our digital homunculi. In this chapter I draw on some of my own experiences to suggest a path forward on this quest.


Consider life in a small society run along patriarchal lines. The head of household’s word is law; perhaps even your household looks up to Him as well, in a chain of lesser families connected by kith and kin. All depends on your relationship with Him. Now consider a situation where He is suddenly removed – perhaps He has died suddenly. Your world wobbles a little bit, but succession rules quickly allow us to figure out who is now in charge. It is a rigid structure, yet it works. Most of the time.

For now.

But what would happen if many Heads rolled, all at once? If the Heads died in infamy and shame? How much damage can such a social world sustain before it collapses, recovers, or transforms? I am thinking now of the social world of the Romans in the late Republic or the early days of Empire, a world self-consciously rigid in the way I described, but yet, one that manages to carry on regardless. Let us, then, kill some Romans. It is perhaps one of the best way to understand the ways in which Roman society was resilient to the frequent pogroms and proscriptions of the late Republic and other eras because we can see what happens next.

Romans must die for me to explore the ways Rome’s social network reacted under stress. This does of course present some obvious practical issues, but through simulation, one that is tractable. The kind of simulation I used was an ‘agent based simulation’ (ABM) (Graham, 2009 for the actual publication of this particular simulation). Think of an ABM as a kind of giant self-running, self-organizing petri dish. Each ‘agent’ is its own program, coded to react to its environment and/or the presence of other ‘agents’ (Lake 2015). Agent based modeling allows me to raise Romans from the dead over and over, and give them patterns of interaction known from the archaeology (for instance the stamps on Roman bricks in and around Rome fossilize nodes of social interaction, as it happens). I raise these digital Romans up; I give them artificial life; and then I kill them. Since the agents are programmed to interact based on social networks known to have existed in the past, aspects of their emergent behaviour are necessarily tied to that past (cf Epstein 2006: 31-33). Thus, since I wish to know under what circumstances this society might collapse, I have them interact in an economic and social world as known from the scholarly literature. My agents harbour grudges; they nurse wounds and social slights. Their primary motivation is to find chains of patrons and clients to whom they can attach in order to obtain resources; that is, a classic rich-get-richer effect is in play. Those who have not, get shut out. They take their revenge. And then I start to put this world under stress to see what happens next.

There is something mesmerizing as I watch this artificial life creep and fight its way across the screen. As described, it is a giant petri dish, where my intervention is limited to setting up the pieces, writing the rules, and flipping the ‘on’ switch. But… wouldn’t you want to play this game? Climb the social ladder in Rome! Help your clients and find yourself better patrons – but make sure you don’t make too many enemies along the way or you too will lose… Game of Togas. It doesn’t take much to flip an agent model into a video game; it simply is a matter of whether or not the player/researcher has any active agency in what happens on the screen. In this regard then, archaeologists are already gamers. They use ABM to explore the past, but remove themselves from the action: thus an ABM is just a species of video game that plays itself. In which case, there is little reason why games-qua-games should not also be another kind of experimental petri dish for archaeologists to write the past.


As I watch the screen and tell the story of what my digital Romans are up to, as they live and die, it becomes easier and easier to believe that I’m actually watching something true about the past…

In Foucault’s Pendulum, Umberto Eco tells a story where the protagonists feed a computer with vast amounts of information, to devise a conspiracy theory, for their own entertainment, to determine a ‘truer’ story of European history. Things take a turn for the worse when the men begin to believe that the simulation is mapping out an actual ‘real’ truth – and even more dangerously, others come to believe in it too.

This, it strikes me is a problem common both to gaming and to simulation. It is all too easy to succumb to the beauty of the digital landscape, a world that turns around me the player, me the creator. In both video games and agent simulations, we have a kind of control, an agency, we do not have in other aspects of our lives. This seductive power blinds us. When we are very good at a game, when we can anticipate what happens next and hit that state where the game is just challenging enough to keep us pushing forward, we have internalized the rules that govern the game and its story. To be ‘good’ at a game is to perform (uncritically) the vision of the world that its creators have encode in the rules, in the mechanics. When we are very good at simulation, we similarly have internalized the ways in which code can be used to tell stories of the world. In which case, if we are interested in archaeogaming, it might be worth thinking about the methods that have evolved to guard against this tendency in modelers. If we are interested in mere simulation, it might be worth thinking about the methods used to understand games to guard against this tendency in gamers.

In which case it is useful to expand therefore on some of the ways agent modeling and video gaming might intersect, particularly in terms of how we evaluate the success or failure of both to ‘do good history’, as a contribution towards the methods of archaeogaming. After all, archaeology has always been concerned with understanding virtual worlds, whether those worlds are built from stone, wood, or concrete; it’s just that now we must understand the worlds built from sand and electricity as well.


The difference between games and agent simulations is not so vast. An agent based model in fact is a special class of video game where the player, does not, in fact, play. She sets it all up, and she watches to see how that world reacts. She’s interested in the whole-world interrelationships; the player of games on the other hand is necessarily interested in the reactions to his own actions. In some respects, one could make the analogy to social network analysis: simulation is to whole-network analysis as a video game is to ego-analysis (cf Weingart 2011-2012 on network analysis). That is to say, the difference is one of perspective.

If archaeogaming is going to be a serious pursuit, then the first lesson we can take from agent modeling concerns time and space. The way that time is treated in agent models is critical; time is malleable so that there is time for something to take place. It makes a difference to your model whether or not your agents update themselves one-at-a-time, each one running its procedures sequentially, versus in parallel. Emergent effects that can seem profound or meaningful might only be an artefact of how ‘time’ is imagined. Then there is the time within which something might take place. Terry Pratchett’s ‘Thief of Time’ (2001) calls this the ‘universal tick’, or the time it takes for now to become then. Agent models tick in time with the computer’s clock: does processor-clock time have any meaningful analogy to ‘historical time’? Similarly, agent models happen in a kind of space. This space can be a flat two dimensional world subject to edge effects that can muddy the waters; in some models, the left hand side of the world connects to the right hand side, and the top connects to the bottom, which gives us a torus. In others, the space is the gaps between social actors, that is, a network. How does space work in the games we are analyzing from an archaeogamer perspective?

Aarseth et al. years ago devised a typology for video games that depended upon considering several axes of analysis – space, time, player-structure, control, and rules (2003). As we begin to devise the methods for archaeogaming, I want to suggest that we pay attention to space and time in their formulation: Space contains ‘perspective, topography, and environment’; Time contains ‘pace, representation’ and ‘teleology’. Whether the virtual world we are analyzing is in ‘meatspace’ or cyberspace, these categories usefully force us to concentrate on what space and time are doing in the game/simulation in meaningful ways. Consider my simulation of Roman social life where the Romans must die. In terms of ‘space’, the simulation has an omni-present perspective: I see all, I can peer into each agent’s life at will. The environment is geometric; the world in which these Romans move is static, the conditions do not change during the run. In terms of ‘time’, the pace is turn-based (each Roman updates in turn), time is mimetic (it takes time for the Roman to achieve something), time is teleological in that the Romans have clear goals and ambitions in mind. As table 1 shows, my simulation occupies an interesting space between ‘Caesar IV’ ref,link and ‘Civilization IV’ ref, link , two games that also contain useful simulations of Roman society.

                              Caesar IV     Civilization IV     RmD
Space           Perspective     Omni-Present    Vagrant             Omni-Present
              Topography      Topological     Geometrical         Geometric
              Environment       Dynamic       Dynamic             Static
Time            Pace              Real-Time     Turn-Based          Turn-Based
              Representation    Arbitrary       Mimetic             Mimetic
              Teleology       Finite          Finite              Finite Table 1 Comparing time and space in games and an agent based model. An expansion of Kee and Graham (2014) Table 13.2.

Let’s put the shoe on the other foot. How do Caesar IV and CiV hold up against the standards used to understand agent based models? Let us use Iza Romanowska’s framework (2014; see also her longer discussion 2015) for evaluating agent models. For Romanowska, the key elements to usefully evaluating the success of an agent based model are:

scope appropriateness resolution how complicated is it? parsimonious parameters utility

Scope and appropriateness deal with research questions. Are we building a model to explore a hypothetical or to understand patterns in the data? Caesar IV clearly has a research question at its heart: how does experience in the provinces make a man fit to govern in Rome? When I reframed ‘Romans must die’ as a game in my opening, – help your clients succeed, find a patron to help you succeed – I was framing a question about the role of patronage in generating social structure in Rome. Appropriateness: is a resource management sim an appropriate tool for answering the question about governance? Resolution: Caesar IV shows me individual Romans, with whom I have to interact. That may be too low a level given the scope & appropriateness. Parsimonious parameters – how many knobs and dials can I twiddle? What and where are the feedback loops? Complexity theory here teaches that simpler is better (Romanowska, 2015). Utility: not, is this a ‘fun’ game, but rather, ‘what have we learned?’ How are we changed?


Finally, I put it to you that one of the most powerful ways that archaeogaming could intersect with a digital public archaeology becomes evident if we consider the original purpose of the Netlogo agent based modeling environment (currently, the most widely used ABM tool in archaeology). Netlogo was originally designed to teach students about complex phenomena by getting them to observe the ways small parameter changes could affect the global behaviour of the complex system (Wilensky, 1999). The ‘Evolving Planet’ archaeogame (Rubio et al, this volume) takes this approach. But we could go further. What if we put the players into our agent-based models? Not just tweaking the global, but engaging the first-person? Holistic and Ego-centric at the same time. Netlogo comes with an extension called ‘hubnet’, which allows individuals to take on the role of a single agent within an otherwise fully digital simulation. The Netlogo developers call this ‘participatory’ simulation’. Is there room in archaeogaming to merge humans and machine-made societies? That tools change us and what it means to be human is a truism of archaeology: archaegaming perhaps is a way to understand what this digital moment is doing to our humanity.


Archaeologists are naturally gamers already. Archaeologists have been building virtual worlds long before video games emerged. We have already developed methods and techniques for understanding the virtual worlds of the past; the things we see as archaeologists in the virtual worlds of the present accordingly can be grounded in the methods and techniques of archaeology. This small essay has suggested a framework based on typologies of time and space coupled with perspectives on agent modeling validation techniques to help guard against the seductive lure of the digital, so that when Romans must die, the die usefully.


Costopoulos, A., M.W. Lake (editors) 2010 Archaeological Simulation: Into the 21st Century. University of Utah Press, Salt Lake City.

Epstein, J. 2006 Agent-based computational models and generative social science. In Generative Social Science: Studies in Agent-Based Computational Modeling, edited by J. M. Epstein, pp. 4-46. Princeton UP, Princeton, NJ.

Espen Aarseth, Solveig M. Smedstad, and Lise Sunnana 2003 A Multi-Dimensional Typology of Games._ Proceedings of the Level Up Games Conference_. pp 48-53. Netherlands Digital Games Research Association, Utrecht.

Graham, Shawn 2009 Behaviour Space: Simulating Roman Social Life and Civil Violence. Digital Studies / Le champ numérique 1.2

Kee, Kevin and Shawn Graham 2014 Teaching history in an age of pervasive computing: the case for games in the high school and undergraduate classroom in Pastplay: Teaching and Learning History with Technology, edited by Kevin Kee, pp337-366. University of Michigan Press, Ann Arbor MI.

Lake, M.W. 2015 Explaining the past with ABM: On modelling philosophy. In Agent-based modeling and simulation in archaeology, edited by G. Wurzer, Kerstin Kowarik, and Hans Reschreiter, pp. 3-35. Springer, Vienna.

Romanowska, Iza 2014 How to evaluate a simulation: a quick guide for non-modellers Simulating Complexity

Romanowska, Iza 2015 So You Think You Can Model? A Guide to Building and Evaluating Archaeological Simulation Models of Dispersals Human Biology 87(3):169-192.

Weingart, Scott. 2011-12 Demystifying Networks, Parts I & II. Journal of Digital Humanities

Wurzer, G, Kerstin Kowarik, and Hans Reschreiter (editors) 2015 Agent-based modeling and simulation in archaeology. Springer, Vienna.