Tools and methods to analyze multimodal data in collaborative design ideation

Ramaswamy Senthil Kumaran Chandrasegaran, Purdue University

Abstract

Collaborative design ideation is typically characterized by informal acts of sketching, annotation, and discussion. Designers have always used the pencil-and-paper medium for this activity, partly because of the flexibility of the medium, and partly because the ambiguous and ill-defined nature of conceptual design cannot easily be supported by computers. However, recent computational tools for conceptual design have leveraged the availability of hand-held computing devices for creating and sharing ideas. In order to provide computer support for collaborative ideation in a way that augments traditional media rather than imitates it, it is necessary to study the affordances made available by digital media for this process, and to study designers' cognitive and collaborative processes when using such media. In this thesis, we present tools and methods to help make sense of unstructured verbal and sketch data generated during collaborative design, with a view to better understand these collaborative and cognitive processes. This thesis has three main contributions. We first present use text visualization techniques to analyze verbal data generated during and about a brainstorming session, and use concept maps and lexical dispersion plots to provide an overview of the design process through these representations. We also use concordance views to disambiguate concepts of interest. Secondly, we identify requirements for a digital framework that supports collaborative sketching in early design, and use skWiki, a framework designed with these requirements in mind, to understand the physical and cognitive processes involved in using such a framework. We identify the value of group awareness—to both the design team and the design researcher—afforded by skWiki's representation of collaborative sketching in the form of ``paths'' of sketches. We identify, through skWiki's branching operation the associated cognitive processes of lateral and vertical transformations that the former involves. We also propose the cognitive processes of reuse, refactor, repurpose, and reinterpret associated with skWiki's merging operation. Finally, we present a web-based visual analytics platform called VizScribe that supports the representation of the design process through the traditional medium of video and transcript, augmented by the text visualization and ``paths'' representation discussed earlier. We introduce interaction paradigms that help researchers query, filter, and code such data to make sense of collaborative design activity. We illustrate the extensibility of this framework to represent other temporal data forms, and to support collaborative coding.

Degree

Ph.D.

Advisors

Ramani, Purdue University.

Subject Area

Mechanical engineering|Web Studies|Information science

COinS