This is a game design case study of my MFA thesis project at University of Southern California. You can read the thesis paper on Medium at https://medium.com/@leighgranergames/leafcutters-life-simulation-gameplay-designed-to-evoke-engagement-with-real-world-subject-matter-4dfa4bce68ec, or download it at http://digitallibrary.usc.edu/cdm/ref/collection/p15799coll127/id/469526
For my thesis project at USC’s Interactive Media Division, I was given a broad and challenging task, to develop an original work and accompanying design theory that would expand the field of Interactive Media.
I began by distilling this prompt into a more specific subject area and design challenge. My thesis would be a two year project, so I sought a subject with significant depth, but that I could significantly explore during that timeline.
My interests in Interactive Media include experiences based on real-world systems such as ecosystems, urban design, and animal behaviors. From that starting point, I established a subject at the intersection of real-world biology and emergent systems, which also had a significant history in Interactive Media: The subject of ants and ant colonies.
I recognized that ant colonies’ emergent behavior could provide deep core game mechanics, while the colorful traits of various ant species inspired a breadth of game features. I also saw that subject also piqued the interest of others who were fascinated by world of ants, which is so alien but also so near to all of us.
Research: Review of Literature
Next, I investigated areas of prior work that my subject touched on:
Artificial Life and Virtual Ants
In the 1980s, Christopher Langton drew on von Neumann and Conway’s earlier works in cellular automata to attempt to create digital “vant” creatures were “not only lifelike, but truly alive.”¹ But ultimately, Langton evaluated his creatures by the intricacies of the patterns they created, rather than by their own sense of aliveness. I realized that it would be key to my project that my creatures felt alive, and established that as a core metric of success.
Life Simulation Games
The game SimLife set out to “break the barriers between games and simulations, playing and learning — even between machines and living beings.”² From my research on Sim Life and other life simulation games, I noted that the player’s agency in a simulation game is always in tension with the accuracy of the simulation, since actions the player takes disrupt the simulated creatures acting on their own.
Creature Management Games
Next, I examined games in which the player manages many autonomous actors, such as Populous, Dungeon Keeper, Dwarf Fortress, SimAnt, and Craft the World. In these games, the player indirectly controls creatures by placing waypoints and spatial markers. However, some of these games allow the player to temporarily take control one of the creatures directly, which unfortunately disrupts the accuracy of the simulation. I therefore established a core design constraint: the player must not be able to directly control the creatures, including telling them where to go, or how to navigate particular spaces of the simulation.
I also explored games featuring AI creatures that learned via in-game conditioning or evolution. In the game Black and White, the player uses treats and punishments to reinforce behaviors in a virtual character. Meanwhile, in Sebastian von Mammen and Christian Jacob’s swarm ‘boid’ games, the artificial creatures automatically evolve to aid the player in an otherwise prohibitively difficult game.³ I noted this gameplay feature, of creatures evolving and learning to take on some of the strategic work of the game.
Virtual pets such as Tomagotchi, Furby, Nintendogs, and Pleo, evoke empathy by putting the player in the role of care-taker. To make my ants feel alive, I decided to implicitly position them as pets in need of the player’s care.
Ants in Interactive Media
I broke down the long history of ants in Interactive Media into a few categories: children’s educational games, bug-squashing games, and ant colony simulations. Existing ant colony simulations tended to share certain simplifications. Rather than specify ant species, they only depict friendly black ants and hostile red ants. They tend to feature abstract art rather than representative depictions of creatures and environments. I decided that my project must authentically represent a unique species of ant, and each ant must appear lifelike in order to inspire empathy.
The three core principles of Leafcutters design:
After reviewing prior works and literature, I was ready to establish a design challenge to expand the field of Interactive Media: How could I create an experience of play with a real-world system that would bring understanding and connection with it, in as real a way as possible?
To address this challenge, I established three core design principles:
- In order to evoke meaningful connection to and engagement with the real subject matter, the game must serve as a window to a living world, rather than an abstracted simulation.
- While abstraction will be required in adapting the subject matter, the game must be as consistently accurate to the real world as possible, in order to bridge as seamlessly into a player’s subsequent further exploration of the subject matter.
- The ants must have agency and make their own decisions. The player must never directly control them.
Exploration of Media Platforms
I examined a variety of potential platforms for my project:
Toward my goal of verisimilitude, I explored designing an experience featuring real ants, but I found that potential forms of interactions would be relatively limited. Biologists have interacted with ants in various ways, including placing ants into artificial environments, placing pheromone trails or other communication chemicals, or physically altering the ants themselves. I decided that such forms of interaction would change the creatures’ behavior in a way that was antithetical to the project.
Big Game: Human Ants
I considered the possibility of a multiplayer Big Game, in which each player would play the role of a single ant in a human-sized ant colony. As each player followed a set of simple rules, the mass of players would demonstrate higher level emergent behaviors. I decided against this format because of the scope and difficulty of iteration, as well as the relative simplicity of a game system which the format could support.
I experimented with depicting digital ants in real space by projecting them onto walls and floors. This was interesting, but I was not satisfied with the aesthetic effect that this achieved, and I was concerned about losing the potential for fine-grained interactive input that a more traditional monitor and mouse would offer.
TV and Mouse
In the end, I found that the television and mouse format provided several benefits important to the project. It allowed for fine-grained input, as well as a user interface that supported a deeper game system. In its gallery installation format, the project was displayed on a large television in front of a couch, with the mouse on a floating side table, to encourage players to settle into a longer experience, and to collaborate.
The Mode of Interaction: Influencing Ant Behaviors
I determined that allowing the player to directly control the ants or spatially affect the environment would negate the sense of the ants as living creatures. Also, if the player could complete a game objective through direct control of an ant, the simulation would no longer be relevant to gameplay.
After exploring a variety of schemes of indirect control, I established the core mode of interactivity: The player encourages and discourages the ants from taking certain actions when the ant encounters certain conditions. Crucially, the player’s actions affect the entire colony or caste, rather than one particular ant. While this would allow the player to make the ants act in inaccurate ways, the game objectives and rules were designed to guide the player toward essentially reconstructing the actual behavior of leaf cutter ants. For the player, this poses a challenge of problem-solving through the manipulation of behaviors of each and all ants simultaneously. This forces the player to think in terms of multiplicity and emergence rather than as a single protagonist.
Prototyping and Playtesting Interactivity and UI
I rapidly iterated on the game design and user experience with prototypes, design reviews, and playtests. I asked the playtesters the following questions:
- Describe the ants in the game. Did they seem like robots? Did they seem lifelike?
- What, if anything, about the game’s subject matter stood out to you?
- To the best of your ability, describe how a leafcutting ant colony functions.
- Would you recommend this game to a friend?
- Question 1 was designed to gauge the player’s emotional experience with the virtual ants.
- Question 2 gauged the player’s general interest and curiosity in the subject matter.
- Question 3 revealed the player’s mental model of the game system and the player’s willingness to apply that model to the real-world subject matter.
- Question 4 was used to gauge the player’s enjoyment and perceived worth of the game.
My paper prototyping showed that players could achieve game objectives by pairing conditions and actions to form simple rules with emergent behaviors. I found that it would be key to prototype gameplay with many ants, so I began digital prototyping relatively early.
Digital Prototype: Behavior Sentences
In the first digital prototype, players constructed behavior rules out of conditions and actions, in a process similar to computer programming. One problem with this prototype was the amount of “debugging” required of the player, that was neither fun nor in keeping with the design goals. Relatedly, players reported that the ants felt like robots.
Digital Prototype: Behavior Flowchart
Next, I built a prototype featuring a flowchart style behavioral interface, with the ants’ actions represented as nodes, and arrows representing conditions that would affect them. While this iteration solved the debugging problem by providing a clearer view of the ants’ internal logic, it did not make the ants seem any more lifelike. To address this in the next iteration, I partially obscured the AI system and introduced weighted randomness to simulate autonomous decision making.
Digital Prototype: Urges and Triggers
In this prototype, I required the player to interact with one ant at a time by encouraging or discouraging behaviors given its immediate external conditions. Those influences then propagated to all ants of the same caste. Player reported a stronger sense that the ants were alive. But, they also reported confusion that interacting with one ant affected other ants as well. This interface also turned out to be prohibitively difficult to manage with large swarms of ants.
Final Interface Design
The final game interface maintains the organic feel of the Urge/Trigger prototype, while adding controls for manipulation of large groups of ants. This interface includes selectable icons corresponding to castes and conditions, and shows how likely an ant is to take various actions in those situations.
As in previous prototypes, the player cannot change an individual ant’s urges, but changes those of the entire ant caste. The forces the player to think at the collective colony level, and allows each ant to retain its autonomy.
The Triggers and Urges System
When the player encourages or discourages an action, that feedback becomes mapped to the conditions that the ant is currently experiencing (and this influence is applied to all ants of that caste).
This figure shows a simple example. When a trigger on the left is true, the ant will experience all of the urges connected to that trigger.
The figure below shows what will happen when the ant smells food while outside. Both triggers are activated, and the urge to “Pick up Food” occurs at its full weight of 3.
In the following figure, only one of the triggers associated with the “Pick up Food” urge is true, and therefore the urge occurs, but at a lesser weight. Since this partial weighting reduces exponentially rather than linearly, the resulting weight is approximately 1, less than half the full weight of the urge.
Note that this might cause the ants to pick up food inside the nest, which could be counterproductive if, for example, the player intended the ants to gather into the nest from outside.
A negative urge can alleviate this problem, as shown in the figure below. If the player discourages the ant from picking up food when that ant is inside the nest and smells food, the ant will gain a negative urge in that situation. As shown below, the negative urge will prevent the ant from picking up food inside the nest by reducing the urge’s weight to a negative number. However, it will not prevent the ant from experiencing the urge to pick up food outside (although it will slightly dampen it).
This system of triggers and overlapping positive and negative urges allows the player to quickly establish a robust system of urges based on many triggers at once. Since clicking on an ant will select all of that ant’s active triggers, the player is able to create basic patterns of AI in relatively few clicks, and refine them later using the more precise icon interface.
Adaptation from Biology
I designed the game system by drawing directly on biology.
Simple rules and emergent behavior
The system of castes, urges, and triggers is inspired by Wilson and Hölldobler’s descriptions of ant castes, tasks, and roles, with workers making “elementary decisions based on local stimuli that contain relatively small amounts of information.”⁴ For example:
…continue hunting for a certain foodstuff if the present foraging load is accepted by nestmates; follow a trail if sufficient pheromone is present; feed the queen more if final-instar larvae are present; and attend the larvae and other immature stages if regular nurse workers are absent. Each of these rules is easily handled by the individual worker, even when we allow for brains as small as a tenth of a cubic millimeter. Each action is also performed in a probabilistic manner with limited precision. Yet when the actions are put together in the form of dense hierarchies involving large numbers of workers, the whole pattern that emerges is strikingly different and more complicated in form, as well as more precise in execution.⁵
The game features a system of leaf cutting and fungus gardening, including leaf pulp, leaf waste, and food production, as well as the “sickening” of the fungus, which is based on the description of fungus gardening and escovopsis microfungus infestation as described by Wirth, Herz, Ryel, Reyschlag, and Hölldobler.⁶
During feature exploration, I implemented the infamous cordyceps fungus, infamous for turning infected ants into “zombies” which then spread spores to more and more ants. But, it destroyed the game balance because it quickly killed the entire ant colony. In researching it, I found this to be fairly accurate! For the sake of gameplay, I had to remove it.
During most skilled playthroughs of the game, the nest becomes cluttered with leaves, food, and larvae. I initially thought this was a failure of my simulation. But after observing a display of leaf cutting ants at the California Academy of Sciences, I noted that their nest was similarly cluttered. So, I learned that my simulation was on the right path.
Evaluating Leafcutters as an expansion of the field of Interactive Media
As described above, my task was to create my own thesis project that “expanded the field of Interactive Media.” Within that broader task, I established a design challenge to myself to create an experience of play with a real-world system that would bring understanding and connection with it, in as real a way as possible.
Through continuous prototyping and user testing, I consistently improved and honed:
- The players’ experience of the ants as lifelike.
- The players’ general interest and curiosity in the subject matter
- The accuracy of the players’ mental model of the game system
- The players’ willingness to apply knowledge of the game system to the real-world subject matter.
- The players’ enjoyment and perceived worth of the game.
Describing Leafcutters in lenses of Interactive Media
I evaluated and considered Leafcutters through a variety of lenses from theories of Interactive Media.
Michael Mateas presents Expressive AI as a concept that changes the focus from an AI system as a thing in itself (presumably demonstrating some essential feature of intelligence) to the communication between author and audience.⁷ Leafcutters can be viewed through this lens: the ants’ artificial intelligence system allows for the emergence of large-scale collective behavior, key to the user experience.
Evocative Knowledge Object
Rich Gold introduces “evocative knowledge objects” in The Plenitude, and Steve Anderson describes them as “tools, systems and architectures that allow us to think differently about the world,” or more simply, “objects we think with.”⁸ ⁹ ¹⁰ I designed Leafcutters to evoke engagement and knowledge of a natural system. The game guides the player very loosely, with optional achievement-style intermediary objectives. The game contains little didactic information, instead teaching through the player’s interaction with the system. Also, the game prompts more questions than it answers, inviting player speculation. For example, the species of the ants, fungal garden, and the natural environment are specific but not named, so the player is free to learn more about them outside of the game.
Leafcutters also functions as an educational game. James Paul Gee writes that “In good games, players feel that their actions and decisions — and not just or primarily the designers’ actions and decisions — are co-creating the world they are in and the experiences they are having.”¹¹ In Leafcutters, the player shapes the ants’ behavior, resulting in visual patterns of movement and visible change in the environment. Over the course of normal play the colony population grows, amplifying this customized visual pattern.
My contribution to Interactive Media: the lens and process of Adaptation Game Design
The final lens of considering Leafcutters is my own contribution to the field — the concept of the Adaptation Game.
Leafcutters is a game designed by a process of adaptation from a natural system. At the most visible level, the narrative and artistic elements of the game are modeled on the subject matter. The game system is also an adaptation, designed to accurately represent the natural systems of leaf cutting ants. Furthermore, the play strategies which arise in this game system mirror behaviors found among real ants. The visuals and audio are designed in a realistic aesthetic to convey the real-world subject matter as well. Leafcutters combines these multiple layers of adaptation in order to faithfully represent leafcutting ants in video game form.
In my opinion, the process of adaptation by which I designed Leafcutters is the most fundamental invention in this project. Leafcutters represents an artifact of such a process of adapting life systems into interactive media, which can be generalized and reapplied.
¹Langton, Christopher G. “Studying artificial life with cellular automata.” Physica D: Nonlinear Phenomena. 22.1–3 (1986): 120–149. Print.
² SimLife. Maxis, 1992.
³ von Mammen, Sebastian, and Christian, Jacob. “Swarming for Games: Immersion in Complex Systems.” Applications of Evolutionary Computing. Ed. Mario Giacobini, Anthony Brabazon, Stefano Cagnoni, Gianni Di Caro, Anikó Ekárt, Anna Esparcia-Alcázar, Muddassar Farooq, Andreas Fink, and Penousal Machado. Heidelberg: Springer Berlin, 2009. 293–302. Print.
⁴Hölldobler, Bert, and Wilson, Edward O. The Ants. Belknap Press, 1990. 358. Print.
⁵Hölldobler and Wilson 358
⁶Wirth, R., H. Herz, R.J. Ryel, W. Beyschlag, and B. Hölldobler. Herbivory of Leaf-Cutting Ants: A Case Study on Atta colombica in the Tropical Rainforest of Panama. Berlin, Heidelberg: Springer-Verlag, 2003. 18. Print.
⁷Mateas, M. “Expressive AI: A hybrid art and science practice.” Leonardo, Journal of the International Society for Arts, Sciences, and Technology. 34.2 (2001): 147–153. Print.
⁸Gold, Rich. The Plenitude: Creativity, Innovation, and Making Stuff. The MIT Press, 2007. Print.
⁹Brown, John Seely. Upside Magazine. Intervew by Marcia Conner. 1993. Web. 21 Mar 2011. <http://www.johnseelybrown.com/linklearn_int.html>.
¹⁰“Interactive Media Division Forum for 9/24/08.” YouTube. Web. 21 Mar 2011. <http://www.youtube.com/watch? v=3bn327W5y_g>.
¹¹Gee, James Paul. “Learning by design: Games as learning machines.” Interactive Educational Multimedia. 8. (2004): 15–23. Print.