Chance and Interactivity Stimulate Creativity David Kirsh (kirsh@ucsd.edu) Cognitive Science La Jolla, CA 92093- 0515 USA Daniel Smithwick (djs2@mit.edu) Department of Architecture / International Design Center 265 Massachusetts Avenue Cambridge, MA 02139 USA Abstract they are stimulative is that they facilitate perceiving new aspects in scenes and then thinking new thoughts. In a To study the value of interactivity and chance in creative cognition we looked at the creative process in architecture, materials shed there are so many things to see and touch. choreography and word discovery. Seventeen architects and It is provocative in the extreme. Our hypothesis is that novice students were given a set of blocks and asked to even perception of the mundane, such as looking at a design their dream house. Although the blocks seem familiar setting in a slightly new way, can also offer us simple they are filled with perceptual surprises. novelty, and that this everyday activity is an important Manipulation led to seeing new things and these in turn led to constituent of creative thought systematically exploited by thinking up new structural forms. In choreography we studied the creative method of a noted choreographer and observed many creative professions and by most of us during how random objects in the environment often figure in tasks problem solving. he assigned his dancers. The dancers would look for Our idea begins with an intuition shared by interesting attributes in the objects that they then played off of phenomenologists: our phenomenal experience is far richer in interesting ways. In a word discovery task we gave than typically described. We can and do discover new subjects a string of 7 letters and asked them to call out all the properties and new possibilities in the same scene over and words of three or more letters they could make. They were over again. Often without appreciating it. These new ways tested in three conditions: static - letters are fixed, interactive- letters can be moved, and shuffle - spacebar randomly of ‘seeing’ a familiar situation are sometimes as big and reorders the letters. Subjects scored best on Shuffle followed surprising as a gestalt switch – as when an ambiguous by Interactive and then Static. The element common to all figure like the necker cube toggles shape. These sort of these creative methods is that changes to the local startling experiences we always notice. Small phenomenal environment lead subjects to notice aspects of the scene in shifts, however, are so common we ignore them as new and interesting ways. This perceptual variety facilitates noteworthy. They occur whenever we attend to some creativity. aspect of a scene we have not appreciated before or when Keywords: Creativity; interactive behavior; externally we organize component elements of a scene in a slightly supported cognition. new way. There is no limit to the number of aspects a person might perceive in a scene. But this is so much a Introduction part of everyday perception that it rarely warrants In design companies a materials shed, sometimes called an discussion or reflection. For a designer, each time such a ideas cart, is often relied on for stimulation during the shift occurs it creates an opportunity to trigger a new idea. ideation phase of design. A materials shed is a collection The shift may contain the germ of an idea, something that of artifacts, such as mechanisms, fabrics, patterns, iconic can be developed into a ‘genuine’ idea. Interactivity and designs, and toys or gadgets that a designer, in the early chance are useful ways of sparking these new ways of part of research, is encouraged to visit. Among the seeing, potentially leading to perceptual surprise and thousands of items filling the shed someone in search of creative insight. inspiration might respond to an odd metallic texture, an To investigate this idea that interactivity and chance are unusual gear configuration, a suction device, a mechanism stimulants for creativity we look at three domains: exploiting a physical principle in an uncommon way. The architecture, choreography and a word game that aspect that triggers interest need not be an element that will resembles scrabble. We do not define creativity per se, nor ever work in the final design. It need not be part of assume there is a reasonable metric for it. We work, something central to the design. Its role is to provoke an rather, on the intuition that a person who generates more idea that had not been considered; an idea that helps the live candidates in a problem such as scrabble is more designer re-align how he or she thinks, or that arouses creative than one who generates fewer (other things like consideration of candidates outside the norm. Looking, memory and expertise being held constant). This is feeling and playing with things is an important part of this because scrabble is not simply a recall task, akin to stem process. completion or fragment completion. It is constraint based. The idea explored in this paper is that interactivity and Hence it is more like memory intensive problem solving chance are powerful stimulations of creativity. The reason than recall. 639 In architecture and dance making, we again work study initially was to observe how architects, architectural without a definition of creativity. Audience and critics, students and non-architects use physical materials to working without a formal definition of creativity, routinely develop design concepts--how they think with physical evaluate the creativity of these products. Their assessment things. But we were struck by the interactive behavior we is notoriously subjective. To avoid hard to defend value observed and the comments made by the participants. We judgments we rely instead on the intuition that weren’t surprised to see and hear that in the early stages of professionals are trained to be creative. It is the job of the 15-minute task we assigned them, participants worked architects and choreographers to produce novel outcomes. with the blocks in an exploratory manner: fishing for Accordingly, for these domains our approach is more design ideas by adding, removing and manipulating blocks ethnographic: we examine the practices of members of frequently. With 51 pieces, the appearance of interesting each community and assume that they use methods that assemblages is far too hard to simulate without physical tend on average to lead to creative outcomes. This means exploration. The comments that startled us were those that we do not try to judge how creative an architectural or suggested that chance ‘noticings’, perceptual surprises, are choreographic product is, we just scrutinize the activities a vital element in what drives their ideation. Through displayed by participants. In architecture, we expect interaction the participants appear to be looking for ideas – experts to have methods and habits that lead to more trying to see new things. And they reported that thought creative output than the methods and habits of novices. too. The benefit of expertise shows up in the way experts The importance of such methods is most certainly the differ from novices in their interactive strategies: the way case in choreography where our subject of study is the they move head, hands and blocks to help stimulate the dance making process of Wayne McGregor, noted British ‘noticing’ process. They seemed to be more skilled in choreographer and practitioner of methods strongly based strategies that lead to architecturally fruitful ideas. on interactivity and chance. Similar approaches are widely Our third study, ‘scrabble’ is the only classical practiced by other choreographers making contemporary experiment we performed and, as mentioned, it differs dance, for instance, William Forsythe (The Forsythe from the other two in not being a large-scale activity with Company) in Frankfurt, and Garry Stewart of the established methods for being creative. We compare word- Australian Dance Theatre to name two companies well finding performance in three conditions, static, interactive known to the first author. For McGregor, chance is not a and chance (shuffle). Our objective was to observe the component in the final performance as it is in aleatoric role that both chance and interaction plays in stimulating performance. Unlike John Cage and others who idea formation – in this case word discovery. Our main incorporate chance events as part of live performances, conjecture is that the chance condition produces the most McGregor’s dance compositions once made are performed words with interaction being second best and looking at the same on each occasion. In Random Dance tiles lying stationary on a tray the worst. This is indeed (McGregor’s company), chance or something close to it, what we found and we talk briefly about the relevance of serves as a mechanism during the ‘making phase’ to draw the scrabble problem and our findings. attention to features or structural elements that the dancers and choreographer then use as seeds or hints. These hints How interactivity helps architectural creativity help them think up movement forms that depart from To study the cognitive role that interacting with tangible previous ones they have seen or performed. Both dancers objects plays in design thinking, we gave 17 architects and and choreographer must still work hard to transform their novice students a set of blocks and asked them to design seed ideas into something more polished and useful. But their dream house on the site shown in figure 1. There chance and the interactive task that provided the starting were ten kinds of blocks available for use, a total of 51 point for the eventual idea are the key factor in stimulating individual blocks. See figure 1. There was no requirement new movement ideas. to use all the blocks, and there was no indication of scale. Interactivity and chance work in a different way in the Nothing specific was said about what would constitute a architectural activity we studied. We created a quasi- completed structure, or how the product would be judged. experimental situation to observe the creative process of Participants: 6 experts, 6 novice students, and 5 non- expert and novice subjects working with physical blocks to architecture students were selected for participation. come up with design concepts for a dream home. The experts (mean age 36, 2F, 4M) were professionally Architects do not typically play with blocks in their studio trained practicing architects with 4-8 years of teaching as a generative process. They work with diagrams and experience at the college and graduate level. sketches, 3D models (digital and physical), and they The novices (mean age 20, 5F, 1M) were second or third review hundreds of images of related projects. While year undergraduates all enrolled in the same architectural working with physical materials and prototypes is a large design department. part of modern architectural practice and an important The non-architectural design students (mean age 20, 2F, component of architecture school curricula, little is 3M) were all undergraduates enrolled at the same understood about how tangible interaction facilitates university, but from different departments, such as creativity (Maher et al 2014). The primary objective in our mechanical engineering. 640 As can be seen in figure 2 the blocks may seem simple cognition. Architects, like artists, benefit by wandering at first, but being parallelepipeds, based on parallelograms, through perceptual richness, especially when that richness they have greater complexity and visual interest than has structural implications. cubes, all of whose faces are squares. When a parallelepiped is rotated 90 degrees, or when a subject repositions him or herself, the shapes that are in view are all different. As a result of this high structural complexity there are many 2D shapes that can be generated by rotating a single block. Completely unexpected attributes might be noticed for the first time. These shapes might be seen in projected shadows, or in the angle of an occluded surface, or even in the texture and shading of a block viewed from a non standard perspective. Figure 2. Both shading and the shape of faces change unpredictably when a parallelepiped is rotated. Both remain constant for cubes under the same rotation. We found that our subjects have dozens of interactive strategies for moving themselves and blocks so as to see novel possibilities. In figure 3 we show a few of the interactive strategies that we observed. There were substantial differences in the way the groups behaved. Experts, compared to novices, spent far more time looking at the site and in particular at how their constructions interact with the site. They also spent more time looking at their constructions from multiple perspectives. Figure 3. Interaction differences between novices and Figure 1: The top figure shows the design task site and block experts. Percents do not sum to 100 because not all type counts. The bottom three images show participants actions are shown. working with blocks to design their dream house. If our conjecture is right then perceiving surprising We also found significant differences in the way experts aspects in a setting, regardless of whether those aspects are and non-experts discussed their designs. Not surprisingly easy or hard to create, lies at the heart of much creative the experts used more technical language. But they also 641 concerned themselves far less with the functions of WM makes a new dance.1 WM relies on giving dancers individual rooms or spaces and instead focused on problems or tasks that involve using arbitrarily selected developing a design concept. For instance, one expert said ‘seeds’, typically derived from materials found in the local “I really like the ability to make a single figure that has a environment, as prompts or stimuli to generate movement. lot of interior/exterior spaces. I can’t go through it and say For example, dancers might be told to look around the what’s my bedroom, what’s my living room…but as a dance studio and pick out six arbitrary curves, surfaces or plain figure I like it and can imagine it occupied in many structures. They then are asked to choose six arbitrary parts ways.” This contrasts with a typical remark from a non- of their body – their left hip, perhaps, or their navel, right expert “I’m using this [area] as the entrance…and I’m ear, left shoulder, mid thigh, and tailbone – and then to going to use these [green blocks, not shown] as the grass sequentially ‘sketch’ their chosen curves and structures and flowers…and these [the wireframe blocks] will be using a random assignment of body part to curve. A windows. These [the blue blocks] look heavier so those dancer might sketch the elliptical curve of a ceiling lamp will be the big walls…like the frame of the house…These using her tailbone, then using her left shoulder she might are my flowers, oh…and a fountain now too.” sketch the joint of a folding chair, then pick up the end of Interactive strategies work because they reveal novel that chair’s curve with her right ear and continue on to shapes and possible configurations. Our inspiration here sketch the pedals of the piano with her navel before using stems from an observation by George Stiny (2006) on the her hip and so on. The result is an odd mixture of smooth relentless ambiguity of structure. For any 3D or even 2D and irregular movements that may contain forms that shape there are an indefinite number of ways of looking at rarely arise. No one expects the dancer to stop there. it. See figure 4 where Y might be seen as (A) three X’s on These movements themselves are ideas that dancers can their side; (B) three clipped hexagons; or (C) one triangle build on to make more integrated, interesting phrases. and three small V’s. Sketching tasks, such as the one just described, are for beginners. WM has extended this sort of method in many ways. Here is a brief transcript of an actual task he gave his (super-expert) dancers when making his 2009 piece “Dyad 1909”. The group was working in a modern studio with a 30ft high ceiling and diverse electronic equipment suspended. WM asked the dancers to lie down and after a Figure 4: Shapes are relentlessly ambiguous. Figure few moments said: adapted from Stiny (ibid). Look Up. Pick boundaries in your field of vision to The idea that at any moment a new aspect of a scene pay attention to. Describe the boundary in which to might catch one’s eye was discussed in post-task work in detail (while on the floor). Pick one detail or interviews. Experienced architects spoke of looking for aspect within the boundary and describe it. Where is guiding design principles that serve to circumscribe their it most black? Assign that blackest thing to your design space. They might work with a subset of blocks, body as a weight. Use that concept of weight as an turning them this way or that, changing their viewing anchor for what you are exploring. Describe angle, hoping to find a simple relation that can form the something that is outside your boundary. Move to a basis for variation or elaboration. They might focus on big new location and repeat these steps. Do the task ideas such as spatial proportions, lighting, and massing. about 5 times total. Each time an architect changes position or stance – as The term ‘describe’ in dance often means to describe when leaning forward, moving sideways, or handling a through body motion. In this case, WM wanted the dancers block – s/he constructs new the opportunities to see things. to characterize the detail or aspect they found interesting in When probed most found serendipity to be important. their self-chosen boundary using internal language or imagery before using that description as inspiration for The use of Randomness in Choreography movement. They spent the next 55 mins on the floor, From the early to mid-20th century performance artists, moving their bodies in original and often striking ways as painters, musicians, writers, and poets often wove random they worked on this task. They came back to this task a elements into their artistic performance. They used “free few times in later sessions. association and accidental movements to generate The output from this task, selected and refined by the randomness” (Leong et al, 2008) at the time of dancers, then selected and modified by WM, goes through performance. By contrast, chance and interactivity is used many more transformations. Two dancers, each with their by the choreographer Wayne McGregor, (henceforth WM), own independent movements in pace and shape might be to stimulate and hint at movement ideas in the ideation told to work together to make a duet. Dancer B might be phase of creation. told to take part of A’s movements and incorporate them For several years the first author has been studying how 1 For description of this project see (Kirsh, 2013) and (Kirsh et al 2009). 642 somehow, or use them as counterpoint movements to help The results as shown in table 1 confirmed our B modify his own. The seed ideas that the task generated conjecture: Subjects produced the most words in Shuffle are edited, cherry picked by WM, added to, altered, followed by Interactive then Static with the main effect combined, stretched, reworked, used as seeds for new being significant F(2,50) = 11.28, p <.001. Data were random functions. At no time is the process mechanical or analyzed using a repeated measures Anova. out of human intervention, especially by WM who personally appropriates the movements himself to ensure Condition Mean # of words SD consistency in feel. Shuffle 18.88 4.14 Nonetheless, over the next three or four weeks it remains Interactive 17.67 4.31 open to novel influences while at the same time being Static 16.56 5.35 refined and thematically shaped. In choreography as in Shuffle > Interactive (t=2.49, p=.02); other areas of design a germ idea is subject to hundreds of Interactive > Static (t=2.20, p=.038); constraints and adaptive forces. Steve Jobs put it like this: Shuffle > Static (t=4.90, p<.001). Table 1. Results of scrabble experiment show that shuffling there is a tremendous amount of craftmanship in helps subjects find more words, with interaction also helpful. between a great idea and a great product. And as you evolve that great idea, it changes and grows. It never Discussion: in scrabble it is easier to evaluate ideas than comes out like it starts because you learn a lot more to generate them, so any process that helps to generate as you get into the subtleties of it and you also find good candidates ought to improve performance. The same that there are tremendous trade-offs that you have to often applies to problem solving and creative thought more make. (1995) broadly. But chance alone is unlikely to deliver good ideas. The chance of shuffling directly producing a legal word is But without that great idea there would be no great so small at 60 to 1 that we almost never find usable words product. The curious thing is that the source of this idea from shuffling. And yet shuffling reliably stimulates legal may come from a seemingly arbitrary feature or detail in word finding. the environment. Creative choreographers today have There are a few reasons to think that shuffling is about dozens of techniques to help their dancers and themselves hinting and that its success is generalizable. First, random to find inspiration in small things. They have techniques generators lower the cost of hypothesis production. that start from some essentially unconnected, dance Pressing a shuffle or randomize button is fast, so fast that a irrelevant place, such as the shape of a lampshade, the button can be pressed many times in the hope that it will behavior of a moving image on a computer screen (Leach eventually surface a good search region for closer & deLahunta, 2015) and draw from that a movement idea exploration. Internal and ordinary interactive generation in that can be refined, combined with others, and structured scrabble is slower, more complex and more effortful. into a pat of a larger dance that has no narrative or People hoping to produce more ideas might want to find a semantic relation to the original seed. hint-maker that relies on chance. This seems to be McGregor’s modus operandi. An experiment on chance in scrabble Second, by using randomness in generation an entire To experimentally explore the value of chance as a search space can be probabilistically guaranteed to be creative stimulant we carried out a small study to test how covered (Rabin, 1976; Motwani & Raghavan, 1995). well subjects play a computerized scrabble-like game. Randomized algorithms break patterns we don’t know we There were three conditions: Shuffle: hitting a ‘shuffle’ are following. Random generators produce candidates with button randomly shuffles seven scrabble tiles, Interactive: larger variance than subjects can on their own because subjects can manually move tiles using a mouse, and there are no biases or filters that limit candidates based on Static: scrabble tiles are fixed in place.2 Subjects were prior expectations of what is good. Every element and given 3 minutes and asked to call out all the English 3, 4, every region in a domain has equal probability of being 5, 6 and 7 letter words they could compose from a 7-letter chosen. As long as agents can decide quickly whether a stimulus such as ‘ETCUSFA’. In each of the three candidate is interesting – once they see it – they can cash conditions, there were 4 stimuli, and each was capable of in on the variance, taking up good opportunities when they making about 80-100 three to seven letter words of arise and ignoring the rest. comparable frequency. We tested 26 subjects (11 Females, The third and perhaps most interesting reason why 15 Males, mean age 25) recruited from students at UCSD random resetting facilitates performance better than other campus and the local population. All participants were conditions is that it duplicates the effect of adding native English speakers, defined as having learned English additional people to a team. Adding cognitive diversity to a before age 5. team is known to facilitate creativity. (Kurtzberg, 2005) Each new person operates with a different cognitive outlook and method; hence biases are partly washed out. 2 This is a revision of an experiment we ran many years Where this is not viable adding chance may be the easiest ago (Matlock et al, 1999) that now includes the random way to diversify. In fact, incorporating chance may be reset condition and a new set of analytical techniques. 643 even more facilitative: there are no group dynamics that interactivity and chance (shuffling) offers is motion and hinder participants from suggesting truly wild ideas. rearrangement. These processes change letter order which These ideas, while quite unconnected to the skilled in turn leads to new noticing’s. Words that were unseen methods we observed choreographers and expert architects before are now discovered. using, are not at odds with them. Although shuffling is not Phenomenologists have pointed to the richness and much like the interactive strategies found in large-scale variety of phenomenal experience. It is precisely this creative practices, both our artistic groups made efforts to richness that artists and others use for creative stimulation. add churn to their thinking. They counted on the causal They have learned how to interact with things to increase richness of the world to throw up surprises when poked. serendipitous seeing. Interactivity that goes beyond mere externalization of what is in the head offers the possibility of this sort of surprise. References Jobs, Steve. (1995) quoted in Kasper Tidemann “Steve Conclusions Jobs on the difference between a great idea and a great We have been arguing that interactivity and chance play product.” 23 June, 2014 a special role in creative cognition. There is no one way to