Case Study: Worldmaking

In this case study, I describe Worldmaking, a software prototype for 3d modeling I designed and developed, and a series of design activities (framed as ‘games’) conducted with pairs of designers using the software. The 3d modeling environment, while poorly suited to architectural design or building modeling, serves as a platform to explore a given possibility space generated by the interface, affordances, and constraints of the program. In conducting the activities specifically as games to be played by a pair of designers, extra rules were introduced that further refined and shifted the space of possibilities and how the designers explored that space — not only individually, but in a dialectic process of conversation with each other and the technology.


I began working on this software as a final project for a course in the second semester of my Master of Architecture program on architectural drawing and representation, and continued developing it after leaving that program and joining my current program. I further developed (and renamed) it as a Teaching Assistant for a computational design seminar course, Inquiry into Computation, Architecture and Design, taught by my advisor, Daniel Cardoso Llach. The software, Worldmaking, is a 3d modeling environment with a limited vocabulary of geometric shapes — voxels (cubes), spheres, and beams (lines) to be placed on and around a 2-dimensional plane at the center of the environment. A designer-user can add these shapes, in a color of their choosing, and can also delete them. No other design actions (such as moving, rotating, scaling) are implemented. While at first blush this might appear overly limiting (and it certainly would be as an architectural design or building modeling program), one precept that became clear to me through a generative approach was the productive application of constraints and rules in the design process. The three shapes in the environment inscribe an infinite possibility space, a world for the designer to create and explore.

An example ‘world’ in the 3d modeling environment
Figure: An example ‘world’ in the 3d modeling environment

The name of the software, shared with the name of a two-week module in the Inquiry course, comes from the opening chapter of philosopher Nelson Goodman’s 1978 book Ways of Worldmaking.1 While Goodman’s focus is mainly linguistic — how different methods and styles of speech and text can form different abstract ‘worlds’ — his ideas are readily applicable to formal design. In particular, he describes how different worlds are shaped by unique logics and truths that constitute that world’s ‘reality,’ a parallel notion to how generative design rules shape a possibility space. Design researcher Donald Schön has also invoked a notion of ‘design worlds’ as formal environments constituted by (and in dialogue with) types and rules.2 The software Worldmaking permits a narrow range of types (objects) and formal rules, but includes other features to grant further agency to designer-users and shape the possibility space within the constraints of the technology.

A turning point in the development of this project came when I implemented a real-time interface allowing multiple designers to work in the same 3d environment simultaneously. While not as direct or embodied as sketching on the same paper,3 this new affordance does make it possible for two or more designers to engage in a conversation implicitly, through the negotiation of forms in space, and also explicitly, through a built-in chat window.4 It would have been possible to investigate a generative approach to design through this software without having real-time interaction between remote designers, but the possibility of communication and miscommunication, of cooperation and conflict, makes this a much richer study. A frame for communication between designers comes from linguist Michael Reddy’s critical notion of the ‘conduit metaphor’5 — the implicit belief that it is possible to directly and seamlessly translate one’s thoughts into verbal or written communication that will then be interpreted unambiguously by others.6 In both the chat interface and through implicit communication through the interplay of forms in the modeling environment, it is impossible to definitively communicate one’s design intent to a partner, who reconstructs the meaning behind communications based on their own worldview, experience, and intent. However, the potential misunderstandings between designers represent not a failure to communicate, but a space rife with generative possibilities. The two games conducted in this study each represent one of the possible communication paradigms, implicit and explicit. The games themselves are adaptations and further explorations of work by design researchers N. John Habraken and Mark Gross: the “Silent Game” and the “Reference Game.”7 In their original work, as in the Inquiry class, the games are played with physical pieces (such as LEGO bricks); in the Worldmaking software the rules are identical but take place in digital space. The games are structures introduced not as design tools, to produce aesthetic or functional forms, but as aides for research into the design process, for “demonstrating and testing design concepts.”


I conducted the pair of games with two different pairs of designers. They are current architecture and design students at Carnegie Mellon University, and one is a recently graduated design student. All the participants have prior experience with design and/or 3d modeling software, and prior to the games, were given a ‘sandbox’ environment to familiarize themselves with the Worldmaking interface. They were randomly paired with each other based on availability and assigned specific roles (A and B) for the games.

The Silent Game

In the Silent Game, written communication between participants through the chat interface is discouraged (except for saying, “I’m done,” or asking clarifying questions). Player A establishes a design intent by placing up to 5 shapes in one corner. Player B places up to 5 shapes in the opposite corner, in order to demonstrate that they understand Player A’s intent. They then repeat this process with free range over the environment, further elaborating and exploring the design intent. There are no goals such as filling the available space or making a specific pattern; the game is open-ended and simply ends after 30 minutes, leaving time for discussion with the participants.

Screenshots taken at the end of both Silent Games
Figure: Screenshots taken at the end of both Silent Games

In the first pair, Player A interpreted ‘5 shapes’ to mean collections of voxels and spheres, as opposed to the individual objects. They embodied their design intent with adjacent groups of objects resembling 3-dimensional Tetris pieces. Player B followed this intent without copying it exactly, modifying some pieces while respecting the ‘language’ established by Player A. After the initial stage, both players kept to the edges of the building plane in the environment. While they switched shapes often, they each kept their own color throughout (gray for Player A, white for Player B).

In the second pair, Player A drew five turquoise spheres in one corner, and Player B replicated the arrangement, rotated 180 degrees, in the opposite corner. Player A built a second level onto this structure, and Player B copied it again. At this point I intervened, worried that Player B would continue to simply imitate Player A’s forms, and I asked Player B to establish a new design intent in one of the remaining corners for Player A to follow. B drew a green sphere, red voxel, and blue triad of beams, and Player A followed this with three distinct voxels in the opposite corner, forming the corners of a right triangle.

The 1st design intent in the lower-left and upper-right corners, the 2nd design intent in the upper-left followed by the lower-right corner
Figure: The 1st design intent in the lower-left and upper-right corners, the 2nd design intent in the upper-left followed by the lower-right corner

Finally, in the free drawing stage, both players focused their efforts around this latest development, Player A’s three voxels, connecting them with spheres, and building a two-height voxel form with beams as a ‘cornice,’ as well as on the original three voxels.

The Reference Game

In the Reference Game, with the same players as A and B, Player A becomes the ‘doer’ and Player B becomes the ‘talker.’ This means that Player B can no longer add or delete shapes in the environment, and instead gives written instructions to Player A via the chat interface, who attempts to follow Player B’s instructions and interpret their design intent. Player B may then clarify or provide further instructions to Player A, who continues drawing in the environment.

Screenshots taken at the end of both Reference Games
Figure: Screenshots taken at the end of both Reference Games

In the first pair, Player B asked Player A, Can you drawing something “democratic”? [sic].8 In response, Player A drew two differently sized rectangles using gray voxels, an arrow out of spheres, and a larger rectangle, symbolically imply a causal relationship between the two smaller rectangles and the larger one.

Player A’s response to Player B’s first instruction
Figure: Player A’s response to Player B’s first instruction

On observing this, Player B was not satisfied, and said, The 3d model is quite different from what I have imagined. They clarified that it should be more scattered. Player A deleted the large rectangle and arrow and added individual voxels, dispersed across the surface. At this, Player B further requested, Can you mix different shapes? looks too generalized individuals. Player A then deleted some of the voxels and replaced them with spheres and scattered beams. At this, Player B was excited, and encouraged more conceptual exploration: it would be also great if the model shows how individuals reach a consensus! Player A responded by reincorporating some larger groups of shapes, but this time using combinations of voxels, spheres, and beams. For unknown reasons, Player A also built a voxel border around the edges, and Player B hinted, if i draw it, I would make the border a bit loose, to which Player A deleted some of the shapes to dematerialize the boundary.

Final state of first reference game
Figure: Final state of first reference game

In the second pair, Player B gave a starting instruction that appeared to offer little room for creative interpretation: Form a line of alternating voxel in rainbow color gradient starting from the one corner. After Player A built a diagonal line of alternating voxels and spheres from a corner to the center, B added: Please continue on with the alternating voxel until the opposing edge is reached… In a muted reverse rainbow color… muted = Darker. After this, B gave more ambiguous instructions: Please draw something "dark" on the darker spectrum of the rainbow… “dark" = "sad"... divide up the space into positive and negative spaces. In response, Player A drew a symbolic ‘sad’ face (in dark purple) and plus and minus signs, all out of spheres.

Second reference game after a few instructions
Figure: Second reference game after a few instructions

Ambitiously, Player B continued: put another set of alternating block on the top of the current block with a different type of voxel. if the bottom is sphere, a cube should be placed… repeat this until two or one block remains on the current row. Player A gamely agreed, building a large, multicolored, triangular wall that spanned the environment’s modeling surface diagonally, and writing afterward: whew done.

Final state of second reference game
Figure: Final state of second reference game


Habraken and Gross, reflecting on their research with the Silent Game and Reference Game (using physical game pieces such as nails and washers), write that “Players must cooperate and try to understand each other’s intentions.” This is the case whether both players are ‘silent’ and communicate only through form, or when one is allowed to speak/write. While the ‘talkers’ in the Reference Game appear to have more agency than the ‘doers,’ who follow their instructions, in both games the instruction-givers engaged in a negotiation with their partner, clarifying and furthering certain points. However, it is necessary to clearly define the relationship between the players. In the first pair, after finishing both activities, Player B described mixed feelings on the roles of the Reference Game: I feel a bit bad about that I kind of order him/her. I wanted to accept their imagination. I think that is what collaboration is. In future iterations of the Reference Game, it might be helpful to reframe the role of the instruction-giver to encourage exploration of emergent forms that Player A designs in the modeling environment.

There are trade-offs in following rules literally or interpretively, and in specifying narrowly-defined or loosely-defined rules. In the second pair’s Silent Game, Player B copied Player A’s forms exactly; in some contexts understanding rules is synonymous with following those rules precisely. When the roles were reversed, Player A took significant creative liberties in interpreting Player B’s intent. At first it was unclear why Player A was adding these particular shapes, but their later explanation demonstrated an abstracted interpretation of B’s intent. As a result, the game entered a new, more dynamic phase, with both players creating novel patterns in response to each other’s forms. In a context of cooperative designing, divergent interpretations can be used generatively. In the first pair’s Reference Game, Player B’s prompt to draw something “democratic” — an abstract notion with no obvious embodiment — served as a simple rule that (through Player A’s work and the resulting interplay) generated unexpected possibilities. This case study uses a digital platform to explore this mode of communication, an openness to possibility and willingness to explore. However, a notion of ‘generative dialogue/conversation’ has a significant history in the (non-computational) worlds of group facilitation, decision-making and participatory design.9 How does this mode apply to computational design specifically?

In the context of automated computational technologies, one could imagine versions of the Silent Game and the Reference Game where one ‘player’ is a set of algorithms that can ‘learn’ through playing the games. Importantly, as the games have no winners or losers, a computer player’s goal would be two-fold: To ‘understand’ the human player’s design intent (encoding it in some digital representation) and to act dialectically alongside the human partner, sparking new ideas from unexpected interpretations and actions. How these new ideas would be qualitatively different from those that arise through human-to-human communication is an open question. In the next chapter, Simulating, I will make the argument that computer simulations, specifically, are useful in producing unexpected or emergent properties to be considered and acted on by a human designer.

1. Goodman, Nelson. Ways of worldmaking. Hackett Publishing, 1978.
2. Schön, Donald A. "Designing: Rules, types and words." In Design studies 9, no. 3 (1988): 181-190.
3. But certainly more egalitarian than a ‘collaborative’ model where one designer looks over the shoulder of another, working at the computer.
4. A setting allows the designer-user to specify their name, a nickname, or to remain anonymous, identified only by a randomly generated ID.
5. Reddy, Michael. "The conduit metaphor." Metaphor and thought 2 (1979).
6. There are strong parallels between the conduit metaphor and the notion of technological neutrality described in the introduction to this thesis — both understate how difficult the process of embodying one’s thoughts is (which anyone taking a drawing class for the first time will agree is a constant struggle).
7. Habraken, N. John, and Mark D. Gross. "Concept design games." Design Studies 9, no. 3 (1988): 150-158.
8. All quotes taken from the chat interface are in italics, with typos/grammatical choices left intact.
9. See the work of David Bohm, William Isaacs, and Otto Scharmer’s Theory U.

results matching ""

    No results matching ""