Sketches for Design and Design of Sketches

Design consists of generating ideas and adapting them to users. Cognitive research can give insights into each of these processes. Here, we describe two projects involving sketches. Sketches are cognitive tools; they have many uses, among them: to promote inferences and new ideas; to organize and convey information. For the former, ambiguity is productive; for the latter, clarity is necessary. From ambiguous sketches, experienced designers are more adept than novices at reorganizing parts to generate new ideas. This raises the possibility of teaching this skill of constructive perception. The sketches people spontaneously produce for clear directions, such as route maps and assembly instructions, omit some information and highlight, even distort, other information. The omissions and distortions are systematic and so useful that they can serve as cognitive design principles for computer algorithms that automatically generate individualized visualizations. Research into how people use sketches to get design ideas and how they produce and comprehend sketches can inform and facilitate design. Sketches for and by Design It is said, though not without controversy, that what distinguishes design from art is function. Design is for a purpose, usually a human one. As such, design entails both generating ideas and adapting those ideas to intended uses. This occurs iteratively. Form and function. Studying how people go about both these tasks gives insights that can facilitate the design process. Two relevant projects will be described. The first investigates how designers and novices get ideas from sketches and applies those insights to suggestions for promoting generation of ideas. The second seeks to develop computer algorithms for designing individualized visualizations, algorithms that are informed by cognitive design principles. Insights from Sketches Why do designers sketch? The simple answer is that they are designing things that can be seen. But this simple answer underestimates the contributions of sketching to the cognition underlying design. After all, designers could construct things in their minds in three dimensions, and to varying extents, they do. But the mind rarely has sufficient capacity to contain an entire object of design; sketches can overcome this limitation. The mind may not notice inconsistencies or incompleteness; sketches demand some consistency and completeness. The mind may not have the capacity to construct, hold, and evaluate a design; sketches hold the constructions in view of the designer, freeing the mind to examine and evaluate. Thus, sketches, like other external representations, relieve short-term memory, demand consistency, and augment information processing. They are also public representations of thought, so they can be shown to others and reasoned on collectively. What the mind does in evaluating sketches to promote design has fascinated designers and cognitive scientists alike. Our own investigations have included experts and students of architecture and design as well as laypeople. They have included analyses of the spontaneous, detailed, step-by-step reports of the thoughts of designers as they designed a building complex as well as experimental manipulations of interpretations of sketches. We review some of those studies and their results here. Role of Sketches in Design Ideas. In contrast to other visualizations, such as diagrams and graphs, sketches, especially early ones, are replete with ambiguities. They are, after all, “sketchy;” that is, vague, committing only to minimal global arrangements and figures. Rather than inducing uncertainty or confusion, ambiguity in design sketches is a source of creativity, as it allows reperceiving and reinterpretating figures and groupings of figures. A designer may construct a sketch with one arrangement in mind, but on inspection, see another arrangement enabling a new, unintended interpretation (e. g. Goldschmidt, 1994; Schon, 1983; Suwa, Gero, & Purcell, 2000; Suwa & Tversky, 1997). Both beginning and experienced designers are facile in making new inferences from their own design sketches. However, experienced designers are more adept at making functional inferences than novices, whose inferences are primarily perceptual (Suwa & Tversky, 1997). A functional inference is seeing the flow of pedestrians in a sketch of a plan whereas a perceptual inference is seeing new spatial relations among structures. The facility of seeing function in structure is a hallmark of expertise in numerous domains from chess (Chase and Simon, ??; de Groot, 19??) to mechanical devices (Heiser and Tversky, submitted). What enables designers to see new implications in sketches, especially their own? The analysis of the protocol of one experienced architect as he designed a building complex is instructive. After perceiving new perceptual configurations in his sketch, he was more likely to get a new design idea than after interpreting the sketch in the same way. Getting a new design idea in turn led to perceiving new perceptual relations in the sketch, and so on, a productive cycle (Suwa, Gero, & Purcell, 2000). Stimulatingng New Design Ideas. Can the strategy used by the expert architect to enable new design ideas be explicitly adopted by others to same end? To see if searching for new perceptual relations could be used purposefully to enable new interpretations by a larger population, we gave undergraduates the ambiguous sketches shown in Figure 1and asked them to come up with as many interpretations as possible for each, a procedure adapted from one used by Howard-Jones (1998; Suwa, Tversky, Gero, and Purcell, 2001). Participants generated ideas for four minutes for each drawing. About two-thirds of the undergraduates, either spontaneously or by suggestion, adopted a strategy of attending to the parts of the sketch, either focusing on different parts or mentally rearranging the parts of the sketch, in order to see new interpretations. Participants attending to parts produced more interpretations, on average, 45 for the different parts group and 50 for the rearrange parts group, than the others, who did not adopt that way of interacting with the sketch and who generated on average 27 interpretations in the four minutes. Drawing 1 Drawing 2 Drawing 3 Drawing 4 Figure 1. Four Ambiguous Drawings One factor that plagues designers and problem solvers in general is fixating on old ideas. Early in the design process, designers generate a flurry of ideas, but later in the design process, they find it harder and harder to see a design differently and generate new ideas. Adopting one of the parts strategies also provides protection against fixation. Undergraduates who adopted one of the parts focus strategies for generating new ideas produced relatively more ideas in the second half of the session than those who did not adopt a parts focus strategy. Comparing Experts and Novices in Generating Interpretations of Sketches. We replicated this experiment on groups of practicing designers, design students, and laypeople (Suwa & Tversky, 2001). The practicing designers produced more interpretations and were more resistant to fixation than any of the other three groups, design students and laypeople, who did not differ from each other. Participants reported in detail the strategies they used to generate new interpretations. Primary among them were regrouping parts and changing reference frames. Participants also reported reversing figure-ground relations in the service of generating new ideas, but used this strategy less often. Both experienced designers and novices of all types used the same strategies, but the designers succeeded in generating more interpretations and resisting fixation. Perceptual reorganization is only half the process of generating ideas; the second, critical half is finding meaningful interpretations, a process which is conceptual in nature. This suggests that what separates experienced designersfrom novices and laypeople is the process of linking perceptual reorganizations to conceptual interpretations. This conclusion echoes the results of the protocol analyses of experienced and novice architects discussed earlier (Suwa & Tversky, 1997). In that study, a major difference between practicing and novice architects was in facility of seeing functional implications of designs. Constructive Perception. Designers appear to deliberately adopt perceptual strategies for reorganizing parts of ambiguous sketches in the service of generating ideas, a process we call constructive perception.. We believe that constructive perception can be fostered, and are experimenting with how to do it. Can constructive perception serve as a model for creativity in other domains? It seems likely. Even abstract domains that cannot be sketched have parts, which can be reorganized into new configurations and reconceived from new points of view. Cognitive Design Principles For design, beauty is not sufficient. The ideas must serve a user. Teapots should be easy to fill and easy to pour, and they should not burn the user. Instructions should be easy to apprehend and follow. To design a better teapot, we need to study how people use them. Similarly for instructions. Careful investigations into human cognition can provide guidelines for effective design. The domain we have chosen is visual instructions. Within those, we have selected two common and familiar cases: route instructions and assembly instructions. Visual instructions are a challenging domain because realism is not paramount. Effective visualizations omit irrelevant information and highlight, even distort, the relevant information. What is relevant depends on how people think of the task. For this reason, the contributions of cognitive experimentation go beyond traditional user testing. The cognitive experiments give insights into how people conceive of routes and assembly and how depictions and language can compatibly convey those conceptions. The conceptions as well as their diagrammatic and linguistic expression are principled. These principles serve as design principles. Let us illustrate how this happens in practice.