DATA GAMES FOR ECOLOGICAL DESIGN

. Can a complex site, such as an urban landscape, be better understood through a game? Might this playful preparation be useful for design? In response to such questions, this paper discusses a practical project that structured design-oriented site research as a develop-ment, implementation and deployment of a locative mobile game in which designers learn by racing colonies of virtual organisms. The anal-ysis of this experiment demonstrates that this approach can support cre-ativity and provide benefits compatible with goals of ecological design.


Aims and questions
Contemporary design increasingly aims to deal with whole ecosystems rather than with discrete objects. Such systems are characterised by emergent and unpredictable events. On one hand, direct, qualitative engagement with these systems is necessary for sound design actions. On the other hand, understanding, imagination and planning can be informed by quantitative data. To overcome this duality, this paper proposes that benefits of both approaches can be enhanced in an "inspirational design environment" (Binder et al., 2011:27-50) that can collate heterogenous inputs. The project discussed below examined whether and how such capabilities can be supported by location-based games.
This work extends the existing research on the use of visualisation technologies, interactivity and games 1) by using a game environment in support of ecological design and 2) by combining embodied experiences with generative simulations. The novelty of these contributions can be confirmed in reference to recent overviews of serious games (Laamarti et al., 2014;Lope & Medina-Medina, 2016) and to examples of recent work. The closest relevant category of the existing work can be defined as 'data games', or games where 116 S. ROUDAVSKI, A. HOLLAND AND J. RUTTEN play involves engagements with real-world data, cf. (Friberger et al., 2013) or chapters 2-5 in (Nijholt, 2017). Existing approaches have attempted to 1) provide interactive access to visualised data to deepen understanding (Yi et al., 2007); 2) encourage engagement by providing reward systems (Handler & Ferrer Conill, 2016); and 3) generate possible conditions by extrapolating from available data (Dickinson et al., 2015). The work presented here seeks to take such approaches further by supporting the design's job of imagining possible futures from unavoidably incomplete premises. The argument is presented in application to a specific case-study that focuses on the design of urban landscapes, a challenge that makes explicit systemic complexities that are also characteristic elsewhere. This case-study aims to: 1) deepen designers' understandings of the environment; 2) encourage creative participation; and 3) expand the repertoire of design methods.

Design experiment
The case-study is a design experiment that focuses on a socio-ecological system of the Merri Creek parklands in Melbourne, Australia. In this experiment, the park is complemented by a virtual environment that incorporates a customwritten geo-referenced cellular-automata engine that is visualized as a navigable space (Figure 1, right) and accessed through mobile devices. This digital environment supports an ecosystem of virtual life where plant-like species spread in reference to spatially distributed affordances. These spatial distributions reference the geometries and materiality of the physical site reflecting its grassy fields, muddy river banks, asphalted vehicular roads and pedestrian paths. The result is a hybrid 'design environment' that allows participants to simultaneously interact with the physical site and the virtual datascape.
In this paper, the term "design environment" includes the totality of entities such as objects, concepts and instruments that are involved in the process of designing and are recognized as outcomes. In this context, the hypothesis of the project is that an arrangement of the design environment that aims to shorten the distance between design representations and designed phenomena can result in a deeper understanding of the challenges at hand.
For the experiment, the site was limited to a 200m 2 section of the park. As characteristic of all sites, this location has been shaped by many simultaneous processes. Examples of these processes include traffic needs resulting in a highway overpass and cycling routes, cultural provisions such as sporting facilities and scenic picnic areas as well as activities that aim to restore natural wetland systems after several decades of degrading industrial use.
Seeking to reflect these overlapping intensities, the test site is discretised as a 300x300 grid, which is populated by layers of numerical information. The experimental environment includes information on: 1) amount and health of vegetation; 2) soil moisture; 3) intensity of pedestrian traffic; and 4) site topography. These data sets have been selected to demonstrate the possible diversity of data types, sources and quality. In this case, information on the photosynthetic activity of vegetation has been obtained through aerial nearinfrared photography; site hydrology sampled with low-cost soil moisture sensors; pedestrian traffic information collected from an online fitness platform; and topography from publicly available maps. The design of the simulation allows an arbitrary number of inputs of heterogenous nature. Varying formats, extents, consistencies and resolutions can be combined to create a compound virtual environment. This capability is significant because it supports 'configurability' (Binder et al., 2011:50) -an important characteristic of all inspirational design environments. The configurability supported by this approach can be particularly powerful as it integrates intuitions of human designers with diverse numerical evidence and direct on-site experiences.
This grid becomes a virtual world when populated by cellular-automata simulations of plant-like organisms. An instance of the world is initiated with several species, distributed at random. Without external intervention by participants, individuals multiply into adjacent empty cells and establish colonies. Individual organisms within cells progress through an initial expansive period, several mature phases, and a period of decay leading to disappearance. Colonies with different behaviours, properties and preferences emerge in response to parameters controlling this life-cycle, resulting in a rich and dynamic virtual landscape. The experiment discussed here featured four distinct types of colonies tagged by colour (Figure 1, left). In response to the contextual conditions, these colonies might spread rapidly and then quickly die off, grow more slowly as clumps, exhibit observable spatial preferences, and demonstrate emergent, partially predictable behaviours.
Each species is conditioned to prefer a spatial condition defined by the site properties. When a species is in its preferred habitat, its multiplication rate is accelerated. For example; one colony grows along moisture gradients. It looks at the neighbouring empty cells and populates those with sufficient soil moisture. If all neighbouring cells are occupied, the colony has no place to expand and will soon die. These rules support self-organizing growth patterns that can react to the site conditions in nonrepeating and surprising ways, without the need for pre-arranged and pre-situated events. This cellular-automata engine does not aim to simulate the site realistically. Instead, it emphasises the dynamic and incompletely controllable character of the design challenge, inviting reflection on the role of design and deeper engagement with the design situation. In this context, the simulation acts as an interpretative device that links heterogenous data sets, human expertise and immersive learning. These capabilities contribute to greater 'creative density' (Binder et al., 2011:50), another important characteristic of inspirational design environments, by facilitating unexpected combinations of data and chance encounters at new locations, with its possible effects.
Simulations using cellular automate have been previously used to model urban or regional interactions. Even when they cannot reliably predict the future states, they are useful as devices that can map and visualise the scope of possibilities and link actions with potential consequences. The project discussed here sought to overcome the abstraction and the simplification that are unavoidable in computational models by narrowing the gap between the virtual and the physical worlds through the use of mobile devices.
This effect is achieved when the simulation is accessed and experienced on site, fulfilling an important requirement of inspirational design environment that need to encourage holistic experiences to the 'genius loci' of target locations (Binder et al., 2011:50). Prototype testing demonstrated new opportunities enabled by such embedding. For instance, this approach is useful for contextual assessment of available data. Incomplete or unevenly distributed datasets are common in practical situations. When a virtual environment informed by such data can be directly compared to on-site conditions, its quality can be more readily appraised and its import usefully reassessed.
Dissonances between different forms of representations, interpretations and experiences can be beneficial because they disrupt habitual assumptions and encourage contributions for a broader range of voices thus supporting "connectivity" -another essential component of the inspirational design environments (Binder et al., 2011:50). To emphasize the virtual environment's capacity to support participative critical engagement and creativity, the access to the simulation is structured as a game. Within it, each species is represented by a set of geometries that change with maturity. The result is a rich world where relative states of each colony can be understood quickly and intuitively.
Several players begin at the same time, in one location and are given control of one species in the simulation. The objective of the game is to make this species most populous. A round lasts ten minutes. Mobile devices show the state of the virtual ecology and the bar chart of the population numbers ( Figure  1, right). When a round begins, players quickly disperse searching for locations where their species might thrive. By physically moving across the site, players alter their location and intervene in its dynamics. Their trail leaves a residue that temporarily converts an empty cell into a habitat suitable the player's species, facilitating multiplication.
In an example of an interaction, one player might notice an emerging rival colony in the scrubby area on the far side of the creek. Confident that vegetation is the that colony's preferred habitat, the player will sprint to the bridge, cross the creek and towards the aggregation of the competing species. Her species will move along the trail and soon surrounds its rival. Without any empty cells to expand to, the rival colony grows old and disappears.
Loosely structured, the game can be played in multiple ways. A group of three might decide to work with a single mobile device. Others form looser coalitions or rivalries: they observe each other but operate separate devices and are not aware of what happens on other participants' screens unless they ask. Each phone runs an isolated instance of the simulation but species behaviour, underlying data maps and site attributes are shared across instances, enabling comparisons, competition and cooperation.
Once the ten-minute game cycle is complete the final population numbers are displayed and players return to discuss the outcomes. One person might complain that the Green easily overtakes the other species, with others nodding in agreement. Sometimes the initial seeding gives an advantage to a species. Another might pass her phone around, showing the Blue overwhelming everything else. Suggesting where they think each species grows best, some want to play another round -eager to see a new starting arrangement and test their just-acquired knowledge of the site. Everyone is allocated a different species, and another round begins. Cast as players, individuals are encouraged to take unusual actions within a shared structure that encourages them to compare and discuss their experiences and understandings.
The next section explores how such interventions can be used as methods to expand design beyond the creation of desired states and towards continuous and iterative negotiation of many forces characteristic of ecological design.

Discussion and conclusion
The project's third objective has been defined as an attempt to provide new design methods in relationship to ecological design and its core principles: 1) place-specific solutions; 2) reliance on ecological accounting; 3) designing in partnership with nature; 4) inclusive participation; and 5) foregrounding of the natural processes (Van der Ryn & Cowan, 2007).
The goal of growing designs from the specificity of concrete places is far from trivial as comprehensive understandings of what comprises such places are not readily available. The proposed approach has the potential to deepen the place-specific research that typically occurs at the beginning of design.
The experiment has also demonstrated that situated simulations can contribute to integration of ecological accounting into design by providing a platform that can combine diverse datasets, provide an interpretative layer that can be compared to site conditions and motivate on-demand data collection.
This commitment to deeper design research also contributes to the goals of design with nature and foregrounding natural processes that is further strengthened by deployment of simulation engines that can model real-world interactions. Deployment of such models can inform the initial, intensive stages of designing and be productive during the ongoing management of change. In addition, it invites design participants to reconsider their role in the design process by emphasising an understanding of designing as the continual negotiation of dynamic relationships rather than the creation of steady states.